Struct Device 

pub struct Device { /* private fields */ }

Wrapper around a VkDevice handle and its loaded command table.

Owns a Box<DeviceCommands> containing all device-level function pointers, loaded at construction via vkGetDeviceProcAddr. Using the real device handle gives the ICD’s direct function pointers, bypassing the loader trampoline, this is the fastest dispatch path in Vulkan.

Holds an optional reference to the Vulkan shared library so that function pointers remain valid even if the originating Entry is dropped. When created via from_raw_parts, the caller manages the library lifetime and this field is None.

Does not implement Drop, the caller must explicitly call destroy_device when done. This avoids double-destroy bugs when wrapping externally managed handles via from_raw_parts.

Guide: Most device-level operations are covered across Memory Management, Command Buffers, and Synchronization.

Examples

use vulkan_rust::vk::structs::*;

// Use the device to create Vulkan objects.
let fence_info = FenceCreateInfo::builder();
let fence = unsafe { device.create_fence(&fence_info, None) }
    .expect("create_fence failed");

// Clean up (reverse creation order).
unsafe {
    device.destroy_fence(fence, None);
    device.destroy_device(None);
    instance.destroy_instance(None);
};

Implementations

impl Device

pub unsafe fn from_raw_parts( handle: Device, get_device_proc_addr: PFN_vkGetDeviceProcAddr, ) -> Self

Wrap a raw handle created externally (OpenXR, middleware, testing).

Safety

• handle must be a valid VkDevice.
• get_device_proc_addr must resolve commands for this device.
• The caller owns the device lifetime.

Examples

use vulkan_rust::Device;

// Given a raw device handle and proc addr from an external source:
let device = unsafe {
    Device::from_raw_parts(raw_device, entry.get_device_proc_addr())
};

pub fn handle(&self) -> Device

Returns the raw VkDevice handle.

pub fn commands(&self) -> &DeviceCommands

Returns the loaded device-level command table.

Use this to call any of the ~200 device-level commands directly, including those without hand-written ergonomic wrappers.

pub unsafe fn create_graphics_pipeline( &self, pipeline_cache: PipelineCache, create_info: &GraphicsPipelineCreateInfo, allocator: Option<&AllocationCallbacks>, ) -> VkResult<Pipeline>

Create a single graphics pipeline.

Convenience wrapper around create_graphics_pipelines for the common single-pipeline case.

Safety

Same requirements as create_graphics_pipelines.

pub unsafe fn create_compute_pipeline( &self, pipeline_cache: PipelineCache, create_info: &ComputePipelineCreateInfo, allocator: Option<&AllocationCallbacks>, ) -> VkResult<Pipeline>

Create a single compute pipeline.

Convenience wrapper around create_compute_pipelines for the common single-pipeline case.

Safety

Same requirements as create_compute_pipelines.

pub unsafe fn map_memory( &self, memory: DeviceMemory, offset: u64, size: u64, flags: MemoryMapFlags, ) -> VkResult<*mut c_void>

Map a device memory object into host address space.

Returns a pointer to the mapped region. Wraps vkMapMemory.

Safety

• memory must be a valid, non-mapped DeviceMemory.
• offset + size must not exceed the allocation size (WHOLE_SIZE is valid for size).
• The returned pointer is only valid until unmap_memory is called.

pub unsafe fn map_memory2( &self, p_memory_map_info: &MemoryMapInfo, ) -> VkResult<*mut c_void>

Map a device memory object into host address space (Vulkan 1.4+).

Returns a pointer to the mapped region. Wraps vkMapMemory2.

Safety

• p_memory_map_info must describe a valid, non-mapped memory range.
• The returned pointer is only valid until the memory is unmapped.

impl Device

pub unsafe fn get_device_proc_addr(&self, p_name: *const c_char)

Wraps vkGetDeviceProcAddr. Provided by VK_BASE_VERSION_1_0.

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetDeviceProcAddr was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Returns a function pointer for a device-level command. This is the device-specific equivalent of get_instance_proc_addr and returns pointers dispatched directly through the device’s driver, bypassing the loader trampoline.

In normal usage you do not need to call this yourself, Device loads all function pointers automatically at creation time. Use this only if you need a command that is not yet exposed as a wrapper method, or for raw interop scenarios.

The returned pointer is only valid for the device it was queried from. Passing a command name that the device does not support returns a null pointer.

pub unsafe fn destroy_device(&self, allocator: Option<&AllocationCallbacks>)

Wraps vkDestroyDevice. Provided by VK_BASE_VERSION_1_0.

Safety

• device (self) must be valid and not destroyed.
• device must be externally synchronized.

Panics

Panics if vkDestroyDevice was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Destroys a logical device and frees its resources. All objects created from this device (buffers, images, pipelines, command pools, etc.) must be destroyed before calling this.

A safe teardown order:

1. device_wait_idle, ensure no GPU work is in flight.
2. Destroy all device-child objects (pipelines, buffers, images, views, descriptor pools, command pools, fences, semaphores, etc.).
3. free_memory for all device memory allocations.
4. destroy_device.

After this call the Device handle is invalid. Do not use it or any object created from it.

Guide

See The Vulkan Object Model in the vulkan_rust guide.

pub unsafe fn get_device_queue( &self, queue_family_index: u32, queue_index: u32, ) -> Queue

Wraps vkGetDeviceQueue. Provided by VK_BASE_VERSION_1_0.

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetDeviceQueue was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Retrieves a queue handle for a queue that was requested at device creation time. The queue_family_index and queue_index must match a family and index that was included in the DeviceCreateInfo’s queue_create_infos.

Queue handles are implicitly owned by the device, they do not need to be destroyed and become invalid when the device is destroyed.

Queues retrieved this way have no special flags. If you created queues with DeviceQueueCreateFlags (e.g. protected queues), use get_device_queue2 instead.

It is common to retrieve queues once after device creation and store them for the lifetime of the device.

Guide

See Hello Triangle, Part 1 in the vulkan_rust guide.

pub unsafe fn queue_submit( &self, queue: Queue, p_submits: &[SubmitInfo], fence: Fence, ) -> VkResult<()>

Wraps vkQueueSubmit. Provided by VK_BASE_VERSION_1_0.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_DEVICE_LOST
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• queue (self) must be valid and not destroyed.
• queue must be externally synchronized.
• fence must be externally synchronized.

Panics

Panics if vkQueueSubmit was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

queue_submit is the primary way to send recorded command buffers to the GPU. Each SubmitInfo specifies:

• Wait semaphores + stage masks: the submission waits on these semaphores at the given pipeline stages before executing.
• Command buffers: executed in array order within the submission.
• Signal semaphores: signalled when all command buffers complete.

Fence: pass a fence to know when the entire batch of submissions completes on the CPU side. Passing Fence::null() means there is no CPU-visible signal, you must use semaphores or queue_wait_idle instead.

Minimize queue_submit calls. Each call has driver overhead; batching multiple SubmitInfo entries into one call is cheaper than separate calls.

Thread safety: a Queue must be externally synchronized. If multiple threads submit to the same queue, you need a mutex.

Guide

See Synchronization in the vulkan_rust guide.

pub unsafe fn queue_wait_idle(&self, queue: Queue) -> VkResult<()>

Wraps vkQueueWaitIdle. Provided by VK_BASE_VERSION_1_0.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_DEVICE_LOST
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• queue (self) must be valid and not destroyed.
• queue must be externally synchronized.

Panics

Panics if vkQueueWaitIdle was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Blocks the calling thread until all commands submitted to this queue have completed. Equivalent to submitting a fence and immediately waiting on it, but simpler.

Use this for quick synchronization in non-performance-critical paths (e.g. during teardown or after a one-shot transfer). In a render loop, prefer fences or timeline semaphores for finer-grained control, queue_wait_idle stalls the CPU and prevents overlap between CPU and GPU work.

The queue must be externally synchronized: do not call this while another thread is submitting to the same queue.

pub unsafe fn device_wait_idle(&self) -> VkResult<()>

Wraps vkDeviceWaitIdle. Provided by VK_BASE_VERSION_1_0.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_DEVICE_LOST
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkDeviceWaitIdle was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Blocks the calling thread until all queues on this device are idle. This is the nuclear option for synchronization, it drains every queue completely.

Typical uses:

• Before destroying the device: ensures no GPU work is in flight before you start tearing down resources.
• Before a swapchain resize: guarantees all frames are done so image views and framebuffers can be safely recreated.

Avoid calling this in a render loop. It forces a full CPU–GPU round-trip and prevents any overlap. Use per-frame fences or timeline semaphores instead.

pub unsafe fn allocate_memory( &self, p_allocate_info: &MemoryAllocateInfo, allocator: Option<&AllocationCallbacks>, ) -> VkResult<DeviceMemory>

Wraps vkAllocateMemory. Provided by VK_BASE_VERSION_1_0.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_INVALID_EXTERNAL_HANDLE
• VK_ERROR_INVALID_OPAQUE_CAPTURE_ADDRESS_KHR
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkAllocateMemory was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Memory allocation in Vulkan is expensive. Prefer sub-allocating from large blocks using a memory allocator (e.g., gpu-allocator) rather than calling this for every buffer or image.

The returned DeviceMemory must be freed with free_memory when no longer needed. Vulkan does not garbage-collect device memory.

Guide

See Memory Management in the vulkan_rust guide.

pub unsafe fn free_memory( &self, memory: DeviceMemory, allocator: Option<&AllocationCallbacks>, )

Wraps vkFreeMemory. Provided by VK_BASE_VERSION_1_0.

Safety

• device (self) must be valid and not destroyed.
• memory must be externally synchronized.

Panics

Panics if vkFreeMemory was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Frees a device memory allocation. All buffers and images bound to this memory must already be destroyed, freeing memory while objects are still bound is undefined behaviour.

If the memory is currently mapped, it is implicitly unmapped before being freed. You do not need to call unmap_memory first, although doing so explicitly is a common defensive practice.

Vulkan has a per-device allocation limit (max_memory_allocation_count, often 4096). Sub-allocating from large blocks and freeing them as a group keeps you well within this limit.

pub unsafe fn unmap_memory(&self, memory: DeviceMemory)

Wraps vkUnmapMemory. Provided by VK_BASE_VERSION_1_0.

Safety

• device (self) must be valid and not destroyed.
• memory must be externally synchronized.

Panics

Panics if vkUnmapMemory was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Unmaps a previously mapped device memory region. After this call the host pointer returned by map_memory is invalid.

Unmapping is only strictly necessary before free_memory if you want to be explicit. Freeing memory implicitly unmaps it.

For persistently mapped memory (the recommended pattern), you typically map once after allocation and unmap only during teardown. There is no performance penalty for keeping memory mapped.

If the memory type is not HOST_COHERENT, make sure to call flush_mapped_memory_ranges after your final writes before unmapping, to ensure the GPU sees the latest data.

pub unsafe fn flush_mapped_memory_ranges( &self, p_memory_ranges: &[MappedMemoryRange], ) -> VkResult<()>

Wraps vkFlushMappedMemoryRanges. Provided by VK_BASE_VERSION_1_0.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkFlushMappedMemoryRanges was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Flushes CPU writes to mapped non-coherent memory so the GPU can see them. Only needed when the memory type does not include MEMORY_PROPERTY_HOST_COHERENT.

Call this after writing data through the mapped pointer and before the GPU reads it (i.e. before the relevant queue_submit).

Alignment: the offset and size of each range must be multiples of non_coherent_atom_size (from physical device limits), or offset must be zero and size must be VK_WHOLE_SIZE. Failing to align causes undefined behaviour on some implementations.

Multiple ranges can be flushed in a single call. Batch them when updating several sub-allocations within the same memory object.

If you are using host-coherent memory, this call is unnecessary and can be skipped entirely.

pub unsafe fn invalidate_mapped_memory_ranges( &self, p_memory_ranges: &[MappedMemoryRange], ) -> VkResult<()>

Wraps vkInvalidateMappedMemoryRanges. Provided by VK_BASE_VERSION_1_0.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkInvalidateMappedMemoryRanges was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Invalidates CPU caches for mapped non-coherent memory so the CPU can see data written by the GPU. The counterpart to flush_mapped_memory_ranges.

Call this after the GPU has finished writing (e.g. after wait_for_fences on the relevant submission) and before reading the data through the mapped pointer.

The same alignment rules apply: offset and size must be multiples of non_coherent_atom_size, or use offset zero with VK_WHOLE_SIZE.

If you are using host-coherent memory, this call is unnecessary, GPU writes are automatically visible to the CPU. Most desktop GPUs offer host-coherent memory types for host-visible heaps.

pub unsafe fn get_device_memory_commitment(&self, memory: DeviceMemory) -> u64

Wraps vkGetDeviceMemoryCommitment. Provided by VK_BASE_VERSION_1_0.

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetDeviceMemoryCommitment was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Queries how many bytes of a lazily-allocated memory object are currently backed by physical storage. Only meaningful for memory allocated with MEMORY_PROPERTY_LAZILY_ALLOCATED.

Lazily-allocated memory is primarily used for transient framebuffer attachments on tile-based GPUs (mobile). The driver may not back the full allocation with physical memory until tiles actually need it.

On desktop GPUs this typically returns the full allocation size since lazy allocation is rarely supported. Check memory_properties.memory_types for the LAZILY_ALLOCATED flag before relying on this.

pub unsafe fn get_buffer_memory_requirements( &self, buffer: Buffer, ) -> MemoryRequirements

Wraps vkGetBufferMemoryRequirements. Provided by VK_BASE_VERSION_1_0.

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetBufferMemoryRequirements was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Returns the memory requirements (size, alignment, compatible memory type bits) for a buffer. Must be called before bind_buffer_memory.

The returned memory_type_bits is a bitmask where bit i is set if memory type i (from get_physical_device_memory_properties) is compatible. Pick a type that satisfies both this mask and your desired properties (HOST_VISIBLE, DEVICE_LOCAL, etc.).

The alignment value must be respected when sub-allocating: the offset passed to bind_buffer_memory must be a multiple of it.

For Vulkan 1.1+, prefer get_buffer_memory_requirements2 which supports dedicated allocation queries via MemoryDedicatedRequirements.

pub unsafe fn bind_buffer_memory( &self, buffer: Buffer, memory: DeviceMemory, memory_offset: u64, ) -> VkResult<()>

Wraps vkBindBufferMemory. Provided by VK_BASE_VERSION_1_0.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_INVALID_OPAQUE_CAPTURE_ADDRESS_KHR
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.
• buffer must be externally synchronized.

Panics

Panics if vkBindBufferMemory was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Binds a DeviceMemory allocation (or a region of one) to a buffer. Must be called before the buffer is used in any command.

Requirements:

1. Call get_buffer_memory_requirements first.
2. The memory_type_index used for allocation must have a bit set in the returned memory_type_bits mask.
3. memory_offset must be a multiple of the returned alignment.
4. memory_offset + size must not exceed the allocation size.

Sub-allocation: multiple buffers can share one DeviceMemory allocation at different offsets. This is strongly recommended, drivers have a per-allocation limit (max_memory_allocation_count, often 4096) and each allocation has overhead.

Once bound, the memory binding cannot be changed for the lifetime of the buffer. Destroy the buffer before freeing its backing memory.

pub unsafe fn get_image_memory_requirements( &self, image: Image, ) -> MemoryRequirements

Wraps vkGetImageMemoryRequirements. Provided by VK_BASE_VERSION_1_0.

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetImageMemoryRequirements was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Returns the memory requirements (size, alignment, compatible memory type bits) for an image. Must be called before bind_image_memory.

Image memory requirements can differ significantly based on tiling, format, and usage flags. An IMAGE_TILING_OPTIMAL image typically requires DEVICE_LOCAL memory and has stricter alignment than a linear image.

When sub-allocating linear and optimal images from the same memory object, the buffer_image_granularity device limit applies. You may need extra padding between the two to satisfy this constraint.

For Vulkan 1.1+, prefer get_image_memory_requirements2 which supports dedicated allocation queries.

pub unsafe fn bind_image_memory( &self, image: Image, memory: DeviceMemory, memory_offset: u64, ) -> VkResult<()>

Wraps vkBindImageMemory. Provided by VK_BASE_VERSION_1_0.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.
• image must be externally synchronized.

Panics

Panics if vkBindImageMemory was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Binds a DeviceMemory allocation (or a region of one) to an image. Must be called after create_image and before the image is used.

Requirements:

1. Call get_image_memory_requirements first.
2. The memory_type_index used for allocation must have a bit set in the returned memory_type_bits mask.
3. memory_offset must be a multiple of the returned alignment.

Dedicated allocations: some drivers perform better when certain images (especially swapchain-sized color or depth targets) have their own allocation. Query get_image_memory_requirements2 with MemoryDedicatedRequirements to check whether the driver prefers or requires a dedicated allocation.

Sub-allocation: like buffers, multiple images can share one allocation at different offsets. Respect alignment from the memory requirements, and note that linear and optimal-tiling images may need buffer_image_granularity spacing between them.

Once bound, the memory binding is permanent. Destroy the image before freeing its backing memory.

pub unsafe fn get_image_sparse_memory_requirements( &self, image: Image, ) -> Vec<SparseImageMemoryRequirements>

Wraps vkGetImageSparseMemoryRequirements. Provided by VK_BASE_VERSION_1_0.

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetImageSparseMemoryRequirements was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Queries the sparse memory requirements for an image created with one of the IMAGE_CREATE_SPARSE_* flags. Returns a list of sparse image format properties describing the memory layout for each image aspect (color, depth, stencil, metadata).

Sparse resources allow partially-resident textures where only some mip levels or regions are backed by physical memory. This is an advanced feature primarily used for virtual texturing and terrain streaming.

If the image was not created with sparse flags, this returns an empty list. Check physical_device_features.sparse_binding before using sparse resources.

pub unsafe fn queue_bind_sparse( &self, queue: Queue, p_bind_info: &[BindSparseInfo], fence: Fence, ) -> VkResult<()>

Wraps vkQueueBindSparse. Provided by VK_BASE_VERSION_1_0.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_DEVICE_LOST
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• queue (self) must be valid and not destroyed.
• queue must be externally synchronized.
• fence must be externally synchronized.

Panics

Panics if vkQueueBindSparse was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Binds sparse memory regions to sparse resources (buffers or images). This is the only way to change the memory backing of a sparse resource after creation.

Sparse binding supports:

• Partial residency: bind memory to individual mip tail regions or image tiles, leaving others unbound.
• Aliasing: multiple sparse resources can alias the same memory region (with IMAGE_CREATE_SPARSE_ALIASED).
• Dynamic re-binding: swap memory pages at runtime for virtual texturing or streaming.

The bind operation is asynchronous and can synchronize with semaphores, similar to queue_submit. The queue must support sparse binding (check QUEUE_SPARSE_BINDING).

This is an advanced feature. Most applications use fully-bound resources with bind_buffer_memory / bind_image_memory instead.

pub unsafe fn create_fence( &self, p_create_info: &FenceCreateInfo, allocator: Option<&AllocationCallbacks>, ) -> VkResult<Fence>

Wraps vkCreateFence. Provided by VK_BASE_VERSION_1_0.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkCreateFence was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Fences are the primary CPU–GPU synchronization primitive. The CPU blocks on wait_for_fences until the GPU signals the fence.

Initial state: create with FENCE_CREATE_SIGNALED when the fence is used in a frame loop that waits before the first submit. Without this flag the first wait_for_fences would block forever.

Typical frame loop pattern:

1. wait_for_fences, block until the previous frame’s GPU work completes.
2. reset_fences, reset back to unsignaled.
3. Record and submit commands, passing the fence to queue_submit.

A fence can only be associated with one submission at a time. Submitting with a fence that is already pending is an error.

For GPU–GPU synchronization (between queue submissions) use semaphores instead. Fences are strictly for CPU-visible signalling.

Guide

See Synchronization in the vulkan_rust guide.

pub unsafe fn destroy_fence( &self, fence: Fence, allocator: Option<&AllocationCallbacks>, )

Wraps vkDestroyFence. Provided by VK_BASE_VERSION_1_0.

Safety

• device (self) must be valid and not destroyed.
• fence must be externally synchronized.

Panics

Panics if vkDestroyFence was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Destroys a fence object. The fence must not be in use by any pending queue_submit call, wait on it or call device_wait_idle before destroying.

Fences are lightweight objects but are still tracked by the driver. Destroy them during teardown or when they are no longer part of your synchronization scheme.

pub unsafe fn reset_fences(&self, p_fences: &[Fence]) -> VkResult<()>

Wraps vkResetFences. Provided by VK_BASE_VERSION_1_0.

Errors

• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.
• pFences must be externally synchronized.

Panics

Panics if vkResetFences was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Resets one or more fences to the unsignaled state. Must be called before reusing a fence in a new queue_submit call.

The fence must not be currently waited on by wait_for_fences from another thread. A common safe pattern:

1. wait_for_fences, blocks until signaled.
2. reset_fences, immediately reset after the wait returns.
3. Submit new work with the fence.

Resetting a fence that is already unsignaled is valid but wasteful. Resetting a fence that is pending (submitted but not yet signaled) is an error.

pub unsafe fn get_fence_status(&self, fence: Fence) -> VkResult<()>

Wraps vkGetFenceStatus. Provided by VK_BASE_VERSION_1_0.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_DEVICE_LOST
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetFenceStatus was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Non-blocking check of whether a fence is signaled. Returns VK_SUCCESS if signaled, VK_NOT_READY if still pending.

Use this for polling patterns where you want to do other work while waiting:

loop {
    if get_fence_status(fence) == VK_SUCCESS { break; }
    // do other work...
}

For blocking waits, prefer wait_for_fences which is more efficient than a spin loop, it lets the CPU sleep until the driver signals.

This call can also return device-lost errors, so check the result even in non-error paths.

pub unsafe fn wait_for_fences( &self, p_fences: &[Fence], wait_all: bool, timeout: u64, ) -> VkResult<()>

Wraps vkWaitForFences. Provided by VK_BASE_VERSION_1_0.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_DEVICE_LOST
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkWaitForFences was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Blocks the calling thread until one or all of the given fences are signaled, or until the timeout expires.

wait_all: when true, the call returns only after every fence in the list is signaled. When false, it returns as soon as any one fence is signaled.

Timeout: specified in nanoseconds. u64::MAX means wait indefinitely. A timeout of zero performs a non-blocking check (equivalent to polling get_fence_status on each fence).

Returns VK_TIMEOUT if the timeout expires before the condition is met. This is not an error, check the return value and handle it (e.g. log a warning or retry).

Typical frame loop:

wait_for_fences(&[frame_fence], true, u64::MAX)
reset_fences(&[frame_fence])
// record and submit...

After wait_for_fences returns successfully, all GPU work associated with those fences is complete and the resources are safe to reuse.

pub unsafe fn create_semaphore( &self, p_create_info: &SemaphoreCreateInfo, allocator: Option<&AllocationCallbacks>, ) -> VkResult<Semaphore>

Wraps vkCreateSemaphore. Provided by VK_BASE_VERSION_1_0.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkCreateSemaphore was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Creates a semaphore for GPU–GPU synchronization between queue submissions. Unlike fences (CPU–GPU), semaphores are invisible to the CPU, they are signaled and waited on entirely within queue_submit or queue_present_khr.

Binary semaphores (the default) have two states: signaled and unsignaled. A submission signals the semaphore, and a later submission waits on it, which also resets it to unsignaled.

Timeline semaphores (Vulkan 1.2+) have a monotonically increasing 64-bit counter. Create one by chaining SemaphoreTypeCreateInfo with SEMAPHORE_TYPE_TIMELINE. Timeline semaphores can be waited on and signaled from the CPU as well via wait_semaphores and signal_semaphore.

Common uses:

• Synchronize between a graphics queue submit and a present.
• Order a transfer upload before a render pass that consumes it.
• Coordinate work across different queue families.

Guide

See Synchronization in the vulkan_rust guide.

pub unsafe fn destroy_semaphore( &self, semaphore: Semaphore, allocator: Option<&AllocationCallbacks>, )

Wraps vkDestroySemaphore. Provided by VK_BASE_VERSION_1_0.

Safety

• device (self) must be valid and not destroyed.
• semaphore must be externally synchronized.

Panics

Panics if vkDestroySemaphore was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Destroys a semaphore. The semaphore must not be referenced by any pending queue submission, either as a wait or signal semaphore.

Wait for all submissions that use this semaphore to complete (via fences or device_wait_idle) before destroying it.

pub unsafe fn create_event( &self, p_create_info: &EventCreateInfo, allocator: Option<&AllocationCallbacks>, ) -> VkResult<Event>

Wraps vkCreateEvent. Provided by VK_COMPUTE_VERSION_1_0.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkCreateEvent was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Creates an event, a fine-grained synchronisation primitive that can be signaled and waited on from both the host (CPU) and the device (GPU).

Events are most useful for split barriers: signal an event at one point in a command buffer, do other work, then wait on it later. This gives the GPU more flexibility to overlap execution compared to a single cmd_pipeline_barrier.

Host-side usage: set_event and reset_event signal and reset from the CPU. get_event_status polls the current state. However, host-signaled events cannot be reliably waited on by the GPU on all implementations, use them primarily for GPU–GPU sync within a queue.

Events are lightweight and cheap to create.

pub unsafe fn destroy_event( &self, event: Event, allocator: Option<&AllocationCallbacks>, )

Wraps vkDestroyEvent. Provided by VK_COMPUTE_VERSION_1_0.

Safety

• device (self) must be valid and not destroyed.
• event must be externally synchronized.

Panics

Panics if vkDestroyEvent was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Destroys an event. The event must not be referenced by any pending command buffer. Wait for all relevant submissions to complete before destroying.

pub unsafe fn get_event_status(&self, event: Event) -> VkResult<()>

Wraps vkGetEventStatus. Provided by VK_COMPUTE_VERSION_1_0.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_DEVICE_LOST
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetEventStatus was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Returns whether an event is currently signaled or unsignaled. Returns VK_EVENT_SET if signaled, VK_EVENT_RESET if not.

This is a non-blocking host-side query. Use it to poll for GPU-signaled events when you need to know the result without blocking. For blocking synchronisation, use fences instead.

pub unsafe fn set_event(&self, event: Event) -> VkResult<()>

Wraps vkSetEvent. Provided by VK_COMPUTE_VERSION_1_0.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.
• event must be externally synchronized.

Panics

Panics if vkSetEvent was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Signals an event from the host (CPU). After this call, get_event_status returns VK_EVENT_SET.

Host-signaled events are primarily useful for host–host synchronisation or as a manual control mechanism. For GPU–GPU synchronisation, prefer cmd_set_event recorded in a command buffer.

pub unsafe fn reset_event(&self, event: Event) -> VkResult<()>

Wraps vkResetEvent. Provided by VK_COMPUTE_VERSION_1_0.

Errors

• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.
• event must be externally synchronized.

Panics

Panics if vkResetEvent was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Resets an event to the unsignaled state from the host (CPU). The event must not be waited on by any pending cmd_wait_events call.

After resetting, the event can be signaled again by set_event or cmd_set_event.

pub unsafe fn create_query_pool( &self, p_create_info: &QueryPoolCreateInfo, allocator: Option<&AllocationCallbacks>, ) -> VkResult<QueryPool>

Wraps vkCreateQueryPool. Provided by VK_BASE_VERSION_1_0.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkCreateQueryPool was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Creates a pool of query slots. Queries let you measure GPU performance and gather statistics without stalling the pipeline.

Query types:

• OCCLUSION: counts how many samples pass the depth test. Useful for visibility culling, render a bounding box, check the count.
• PIPELINE_STATISTICS: counts shader invocations, primitives, clipping, etc. Must be enabled via pipeline_statistics_query device feature.
• TIMESTAMP: records a GPU timestamp. Use two timestamps and the timestamp_period device property to measure elapsed time.

Queries must be reset before use with cmd_reset_query_pool (or reset_query_pool on Vulkan 1.2+). Results are retrieved with get_query_pool_results or copied into a buffer with cmd_copy_query_pool_results.

pub unsafe fn destroy_query_pool( &self, query_pool: QueryPool, allocator: Option<&AllocationCallbacks>, )

Wraps vkDestroyQueryPool. Provided by VK_BASE_VERSION_1_0.

Safety

• device (self) must be valid and not destroyed.
• queryPool must be externally synchronized.

Panics

Panics if vkDestroyQueryPool was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Destroys a query pool and frees its resources. All command buffers that reference this pool must have completed execution before destroying it.

pub unsafe fn get_query_pool_results( &self, query_pool: QueryPool, first_query: u32, query_count: u32, data_size: usize, p_data: *mut c_void, stride: u64, flags: QueryResultFlags, ) -> VkResult<()>

Wraps vkGetQueryPoolResults. Provided by VK_BASE_VERSION_1_0.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_DEVICE_LOST
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetQueryPoolResults was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Reads query results from a query pool into a host buffer. This is the CPU-side retrieval path, for GPU-side copies into a device buffer, use cmd_copy_query_pool_results instead.

Key flags:

• QUERY_RESULT_64: return 64-bit results. Always use this for timestamp and pipeline statistics queries to avoid overflow.
• QUERY_RESULT_WAIT: block until all requested queries are available. Without this flag, unavailable queries return VK_NOT_READY and their slots are left untouched.
• QUERY_RESULT_WITH_AVAILABILITY: append an availability value after each result (non-zero if available). Useful for polling without blocking.
• QUERY_RESULT_PARTIAL: return whatever data is available even for incomplete queries. Only meaningful for occlusion queries.

Stride: the stride parameter is the byte distance between successive query results in your output buffer. It must be at least large enough to hold the result plus the optional availability value.

Queries that have not been started or not yet completed return VK_NOT_READY unless QUERY_RESULT_WAIT is set.

pub unsafe fn reset_query_pool( &self, query_pool: QueryPool, first_query: u32, query_count: u32, )

Wraps vkResetQueryPool. Provided by VK_BASE_VERSION_1_2.

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkResetQueryPool was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Resets a range of queries in a pool from the host (CPU). This is the Vulkan 1.2 host-side alternative to cmd_reset_query_pool, which resets queries from a command buffer.

Host-side reset is simpler, call it directly without recording a command buffer. Requires the host_query_reset feature (core in Vulkan 1.2).

Queries must be reset before use. Resetting a query that is in use by a pending command buffer is an error.

pub unsafe fn create_buffer( &self, p_create_info: &BufferCreateInfo, allocator: Option<&AllocationCallbacks>, ) -> VkResult<Buffer>

Wraps vkCreateBuffer. Provided by VK_BASE_VERSION_1_0.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_INVALID_OPAQUE_CAPTURE_ADDRESS_KHR
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkCreateBuffer was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Examples

use vulkan_rust::vk::structs::*;
use vulkan_rust::vk::bitmasks::*;

let info = BufferCreateInfo::builder()
    .size(1024)
    .usage(BufferUsageFlagBits::VERTEX_BUFFER)
    .sharing_mode(vulkan_rust::vk::enums::SharingMode::EXCLUSIVE);
let buffer = unsafe { device.create_buffer(&info, None) }
    .expect("buffer creation failed");
// Use buffer...
unsafe { device.destroy_buffer(buffer, None) };

Guide

See Memory Management in the vulkan_rust guide.

pub unsafe fn destroy_buffer( &self, buffer: Buffer, allocator: Option<&AllocationCallbacks>, )

Wraps vkDestroyBuffer. Provided by VK_BASE_VERSION_1_0.

Safety

• device (self) must be valid and not destroyed.
• buffer must be externally synchronized.

Panics

Panics if vkDestroyBuffer was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Destroys a buffer object. The buffer must not be in use by any pending GPU work, wait on the relevant fences or call device_wait_idle before destroying.

Destroying a buffer does not free its backing memory. Call free_memory separately (or let your sub-allocator reclaim the region).

Destroy order: destroy the buffer first, then free the memory. Not the reverse, freeing memory while a buffer is still bound to it is undefined behaviour.

pub unsafe fn create_buffer_view( &self, p_create_info: &BufferViewCreateInfo, allocator: Option<&AllocationCallbacks>, ) -> VkResult<BufferView>

Wraps vkCreateBufferView. Provided by VK_COMPUTE_VERSION_1_0.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkCreateBufferView was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Creates a view into a buffer that interprets its contents as a typed array of texels. Buffer views are used with:

• Uniform texel buffers (BUFFER_USAGE_UNIFORM_TEXEL_BUFFER): read-only typed access from shaders via samplerBuffer / textureBuffer in GLSL.
• Storage texel buffers (BUFFER_USAGE_STORAGE_TEXEL_BUFFER): read-write typed access from shaders via imageBuffer in GLSL.

The format, offset, and range define the view window into the buffer. The format must be supported for the buffer view usage, check format_properties.buffer_features.

Buffer views are less common than image views. They are mainly used for large, flat data arrays (e.g. particle attributes, lookup tables) that benefit from format conversion on read.

pub unsafe fn destroy_buffer_view( &self, buffer_view: BufferView, allocator: Option<&AllocationCallbacks>, )

Wraps vkDestroyBufferView. Provided by VK_COMPUTE_VERSION_1_0.

Safety

• device (self) must be valid and not destroyed.
• bufferView must be externally synchronized.

Panics

Panics if vkDestroyBufferView was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Destroys a buffer view. The view must not be referenced by any descriptor set that is bound in a pending command buffer.

Destroy buffer views before the underlying buffer.

pub unsafe fn create_image( &self, p_create_info: &ImageCreateInfo, allocator: Option<&AllocationCallbacks>, ) -> VkResult<Image>

Wraps vkCreateImage. Provided by VK_BASE_VERSION_1_0.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_COMPRESSION_EXHAUSTED_EXT
• VK_ERROR_INVALID_OPAQUE_CAPTURE_ADDRESS_KHR
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkCreateImage was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

After creating an image you must bind memory to it before use:

1. Query requirements with get_image_memory_requirements.
2. Allocate from a compatible memory type with allocate_memory.
3. Bind with bind_image_memory.

Choose IMAGE_TILING_OPTIMAL for GPU-side textures and render targets. Use IMAGE_TILING_LINEAR only when you need direct CPU access to the texel layout (e.g. CPU readback), and check format support first with get_physical_device_image_format_properties.

The initial_layout must be UNDEFINED or PREINITIALIZED. Most applications use UNDEFINED and transition via a pipeline barrier.

Destroy with destroy_image when no longer needed. Do not destroy an image that is still referenced by a framebuffer or image view.

pub unsafe fn destroy_image( &self, image: Image, allocator: Option<&AllocationCallbacks>, )

Wraps vkDestroyImage. Provided by VK_BASE_VERSION_1_0.

Safety

• device (self) must be valid and not destroyed.
• image must be externally synchronized.

Panics

Panics if vkDestroyImage was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Destroys an image object. The image must not be in use by any pending GPU work, and all image views referencing this image must already be destroyed.

Destroying an image does not free its backing memory. Call free_memory separately after destroying the image.

Safe teardown order for an image:

1. Wait for all GPU work using the image to complete.
2. Destroy all ImageView objects referencing the image.
3. Destroy any Framebuffer objects that included those views.
4. destroy_image.
5. Free or reclaim the backing memory.

pub unsafe fn get_image_subresource_layout( &self, image: Image, p_subresource: &ImageSubresource, ) -> SubresourceLayout

Wraps vkGetImageSubresourceLayout. Provided by VK_BASE_VERSION_1_0.

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetImageSubresourceLayout was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Queries the memory layout (offset, size, row pitch, array pitch, depth pitch) of a specific subresource within a linear-tiling image. Only valid for images created with IMAGE_TILING_LINEAR. For optimal-tiling images, use get_image_subresource_layout2.

pub unsafe fn create_image_view( &self, p_create_info: &ImageViewCreateInfo, allocator: Option<&AllocationCallbacks>, ) -> VkResult<ImageView>

Wraps vkCreateImageView. Provided by VK_BASE_VERSION_1_0.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_INVALID_OPAQUE_CAPTURE_ADDRESS_KHR
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkCreateImageView was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

An image view selects a subset of an image’s subresources and reinterprets them for a specific use (sampling, color attachment, etc.).

Common pitfalls:

• Aspect mask: use COLOR for color formats, DEPTH and/or STENCIL for depth/stencil formats. Getting this wrong causes validation errors that are not always obvious.
• Format compatibility: the view format must be compatible with the image’s format. Using IMAGE_CREATE_MUTABLE_FORMAT on the image relaxes this to any format in the same size-compatibility class.
• View type vs image type: a 2D image can back a VIEW_TYPE_2D or VIEW_TYPE_2D_ARRAY. A 3D image cannot be viewed as 2D without VK_EXT_image_2d_view_of_3d.

Destroy with destroy_image_view when no longer needed. Destroy image views before destroying the underlying image.

pub unsafe fn destroy_image_view( &self, image_view: ImageView, allocator: Option<&AllocationCallbacks>, )

Wraps vkDestroyImageView. Provided by VK_BASE_VERSION_1_0.

Safety

• device (self) must be valid and not destroyed.
• imageView must be externally synchronized.

Panics

Panics if vkDestroyImageView was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Destroys an image view. The view must not be referenced by any pending GPU work or by any framebuffer that is still in use.

Destroy image views before the underlying image. Destroy any framebuffers that reference the view before destroying the view itself.

pub unsafe fn create_shader_module( &self, p_create_info: &ShaderModuleCreateInfo, allocator: Option<&AllocationCallbacks>, ) -> VkResult<ShaderModule>

Wraps vkCreateShaderModule. Provided by VK_COMPUTE_VERSION_1_0.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_INVALID_SHADER_NV
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkCreateShaderModule was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

The input must be valid SPIR-V bytecode. The code slice is &[u32] and must be aligned to 4 bytes, which Rust’s &[u32] guarantees.

Use the bytecode::read_spv helper to load a .spv file from disk with correct alignment.

Shader modules can be destroyed immediately after pipeline creation; the driver copies what it needs during create_graphics_pipelines or create_compute_pipelines. Destroying early keeps the handle count low.

Guide

See Pipelines in the vulkan_rust guide.

pub unsafe fn destroy_shader_module( &self, shader_module: ShaderModule, allocator: Option<&AllocationCallbacks>, )

Wraps vkDestroyShaderModule. Provided by VK_COMPUTE_VERSION_1_0.

Safety

• device (self) must be valid and not destroyed.
• shaderModule must be externally synchronized.

Panics

Panics if vkDestroyShaderModule was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Destroys a shader module. The module must not be in use by any pipeline. Shader modules can be safely destroyed after pipeline creation since the driver copies the SPIR-V at creation time.

pub unsafe fn create_pipeline_cache( &self, p_create_info: &PipelineCacheCreateInfo, allocator: Option<&AllocationCallbacks>, ) -> VkResult<PipelineCache>

Wraps vkCreatePipelineCache. Provided by VK_COMPUTE_VERSION_1_0.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkCreatePipelineCache was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Creates a pipeline cache that stores compiled pipeline state to speed up future pipeline creation.

Initial data: pass previously serialized cache data (from get_pipeline_cache_data) to warm the cache on startup. The driver validates the header and silently ignores data from incompatible driver versions or hardware, it is always safe to pass stale data.

A single cache can be shared across all pipeline creation calls in the application. Multiple threads can use the same cache concurrently, the driver handles internal synchronization.

Recommended workflow:

1. On startup, load cache data from disk and create the cache.
2. Pass the cache to every create_graphics_pipelines and create_compute_pipelines call.
3. On shutdown, serialize with get_pipeline_cache_data and write to disk.

pub unsafe fn destroy_pipeline_cache( &self, pipeline_cache: PipelineCache, allocator: Option<&AllocationCallbacks>, )

Wraps vkDestroyPipelineCache. Provided by VK_COMPUTE_VERSION_1_0.

Safety

• device (self) must be valid and not destroyed.
• pipelineCache must be externally synchronized.

Panics

Panics if vkDestroyPipelineCache was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Destroys a pipeline cache. Pipelines that were created using this cache remain valid, the cache is only needed during creation, not at runtime.

Serialize the cache with get_pipeline_cache_data before destroying if you want to persist it to disk for the next session.

pub unsafe fn get_pipeline_cache_data( &self, pipeline_cache: PipelineCache, p_data: *mut c_void, ) -> VkResult<usize>

Wraps vkGetPipelineCacheData. Provided by VK_COMPUTE_VERSION_1_0.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetPipelineCacheData was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Serializes the contents of a pipeline cache into a byte buffer for storage on disk. The data includes a vendor-specific header that the driver uses to validate compatibility on reload.

Call this with a null data pointer first to query the required buffer size, then allocate and call again. The wrapper handles this two-call pattern for you.

The cache data is not portable across different GPU vendors, driver versions, or pipeline cache UUIDs. Always check the header or let the driver reject incompatible data silently on reload via create_pipeline_cache.

Write the data to a file (e.g. pipeline_cache.bin) and load it on the next application start to avoid redundant shader compilation.

pub unsafe fn merge_pipeline_caches( &self, dst_cache: PipelineCache, p_src_caches: &[PipelineCache], ) -> VkResult<()>

Wraps vkMergePipelineCaches. Provided by VK_COMPUTE_VERSION_1_0.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.
• dstCache must be externally synchronized.

Panics

Panics if vkMergePipelineCaches was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Merges one or more source caches into a destination cache. Useful when multiple threads each use their own cache during parallel pipeline creation, merge them into a single cache before serializing to disk.

The source caches are not modified or destroyed by this call. The destination cache receives all entries from the sources that it does not already contain. Duplicate entries are ignored.

After merging, you can destroy the source caches if they are no longer needed.

pub unsafe fn create_pipeline_binaries_khr( &self, p_create_info: &PipelineBinaryCreateInfoKHR, allocator: Option<&AllocationCallbacks>, p_binaries: &mut PipelineBinaryHandlesInfoKHR, ) -> VkResult<()>

Wraps vkCreatePipelineBinariesKHR. Provided by VK_KHR_pipeline_binary.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_INITIALIZATION_FAILED
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkCreatePipelineBinariesKHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Creates pipeline binary objects from either a pipeline create info or previously serialized binary data. Pipeline binaries capture compiled shader code in a device-specific format, enabling fast pipeline recreation without recompilation.

Two creation paths via PipelineBinaryCreateInfoKHR:

• From pipeline create info + key: compiles shaders and produces binaries. Use get_pipeline_key_khr to obtain the key first.
• From serialized data: restores binaries saved with get_pipeline_binary_data_khr from a prior run. This skips compilation entirely.

The output is written to PipelineBinaryHandlesInfoKHR. Call once with a null pipelines pointer to query the count, then again with an allocated array.

Pipeline binaries are more portable than pipeline caches, they can be validated, versioned, and stored in application-managed files rather than opaque blobs.

pub unsafe fn destroy_pipeline_binary_khr( &self, pipeline_binary: PipelineBinaryKHR, allocator: Option<&AllocationCallbacks>, )

Wraps vkDestroyPipelineBinaryKHR. Provided by VK_KHR_pipeline_binary.

Safety

• device (self) must be valid and not destroyed.
• pipelineBinary must be externally synchronized.

Panics

Panics if vkDestroyPipelineBinaryKHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Destroys a pipeline binary handle. After destruction, the binary cannot be used to create pipelines or retrieve data.

Pipeline binaries are independent of the pipelines created from them, destroying a binary does not affect any pipeline that was already created using it.

pub unsafe fn get_pipeline_key_khr( &self, p_pipeline_create_info: Option<&PipelineCreateInfoKHR>, p_pipeline_key: &mut PipelineBinaryKeyKHR, ) -> VkResult<()>

Wraps vkGetPipelineKeyKHR. Provided by VK_KHR_pipeline_binary.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetPipelineKeyKHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Computes a pipeline key that identifies the pipeline configuration for use with pipeline binaries. The key is a hash of the pipeline create info that lets you look up previously cached binaries.

Pass a PipelineCreateInfoKHR referencing the pipeline create info (graphics, compute, or ray tracing) to get its key. Pass None to get an empty key structure for use as an output parameter.

Store the key alongside serialized binary data from get_pipeline_binary_data_khr. On subsequent runs, compute the key for the current pipeline configuration and check if a matching binary exists before falling back to full compilation.

pub unsafe fn get_pipeline_binary_data_khr( &self, p_info: &PipelineBinaryDataInfoKHR, p_pipeline_binary_key: &mut PipelineBinaryKeyKHR, p_pipeline_binary_data: *mut c_void, ) -> VkResult<usize>

Wraps vkGetPipelineBinaryDataKHR. Provided by VK_KHR_pipeline_binary.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_NOT_ENOUGH_SPACE_KHR
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetPipelineBinaryDataKHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Serializes a pipeline binary into a byte buffer for offline storage. The data can be saved to disk and later passed to create_pipeline_binaries_khr to skip shader compilation on subsequent application launches.

Uses the two-call pattern: call with a null p_pipeline_binary_data to query the required data_size, allocate a buffer, then call again to fill it.

The output also includes a PipelineBinaryKeyKHR that identifies the binary. Store the key alongside the data, it is required when recreating the binary.

Serialized data is device-specific and may become invalid after driver updates. Applications should handle creation failure gracefully by falling back to full recompilation.

pub unsafe fn release_captured_pipeline_data_khr( &self, p_info: &ReleaseCapturedPipelineDataInfoKHR, allocator: Option<&AllocationCallbacks>, ) -> VkResult<()>

Wraps vkReleaseCapturedPipelineDataKHR. Provided by VK_KHR_pipeline_binary.

Errors

• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkReleaseCapturedPipelineDataKHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Releases internal data captured during pipeline creation that was retained for binary extraction. Call this after you have finished calling create_pipeline_binaries_khr for a pipeline created with PIPELINE_CREATE_CAPTURE_DATA.

The pipeline itself remains valid after this call, only the captured internal data is freed. This reduces memory usage when you no longer need to extract binaries from the pipeline.

pub unsafe fn create_graphics_pipelines( &self, pipeline_cache: PipelineCache, p_create_infos: &[GraphicsPipelineCreateInfo], allocator: Option<&AllocationCallbacks>, ) -> VkResult<Vec<Pipeline>>

Wraps vkCreateGraphicsPipelines. Provided by VK_GRAPHICS_VERSION_1_0.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_INVALID_SHADER_NV
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.
• pipelineCache must be externally synchronized.

Panics

Panics if vkCreateGraphicsPipelines was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Graphics pipeline creation is the most expensive Vulkan object creation call. Batch multiple pipelines into a single call when possible,the driver can often parallelise compilation internally.

Pipeline cache: always pass a PipelineCache. Even an empty cache helps on the first run; on subsequent runs it avoids redundant shader compilation entirely. Serialize the cache to disk between application sessions with get_pipeline_cache_data.

Dynamic state: mark states like viewport, scissor, and blend constants as dynamic to reduce the number of pipeline permutations. Vulkan 1.3 makes viewport and scissor dynamic by default via VK_DYNAMIC_STATE_VIEWPORT_WITH_COUNT and VK_DYNAMIC_STATE_SCISSOR_WITH_COUNT.

If creation fails for one pipeline in a batch, the call returns an error but may still populate some output handles. Check VK_PIPELINE_COMPILE_REQUIRED when using VK_PIPELINE_CREATE_FAIL_ON_PIPELINE_COMPILE_REQUIRED_BIT.

Guide

See Pipelines in the vulkan_rust guide.

pub unsafe fn create_compute_pipelines( &self, pipeline_cache: PipelineCache, p_create_infos: &[ComputePipelineCreateInfo], allocator: Option<&AllocationCallbacks>, ) -> VkResult<Vec<Pipeline>>

Wraps vkCreateComputePipelines. Provided by VK_COMPUTE_VERSION_1_0.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_INVALID_SHADER_NV
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.
• pipelineCache must be externally synchronized.

Panics

Panics if vkCreateComputePipelines was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Creates one or more compute pipelines. Compute pipelines are simpler than graphics pipelines, they only need a single shader stage and a pipeline layout.

Pipeline cache: pass a PipelineCache to speed up creation, the same way as with graphics pipelines. The cache is shared across both pipeline types.

Specialisation constants: use SpecializationInfo on the shader stage to bake compile-time constants into the shader (e.g. workgroup size, algorithm variant). This produces optimised code without duplicating shader source.

Batch multiple compute pipelines in a single call when possible.

Compute pipelines can be created at any time and are not tied to a render pass. They are bound with cmd_bind_pipeline using PIPELINE_BIND_POINT_COMPUTE and dispatched with cmd_dispatch.

pub unsafe fn get_device_subpass_shading_max_workgroup_size_huawei( &self, renderpass: RenderPass, p_max_workgroup_size: *mut Extent2D, ) -> VkResult<()>

Wraps vkGetDeviceSubpassShadingMaxWorkgroupSizeHUAWEI. Provided by VK_HUAWEI_subpass_shading.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_SURFACE_LOST_KHR
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetDeviceSubpassShadingMaxWorkgroupSizeHUAWEI was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Queries the maximum workgroup size supported for subpass shading on the given render pass. Returns an Extent2D with the max width and height. Use this to configure subpass shading dispatch parameters.

Requires VK_HUAWEI_subpass_shading.

pub unsafe fn destroy_pipeline( &self, pipeline: Pipeline, allocator: Option<&AllocationCallbacks>, )

Wraps vkDestroyPipeline. Provided by VK_COMPUTE_VERSION_1_0.

Safety

• device (self) must be valid and not destroyed.
• pipeline must be externally synchronized.

Panics

Panics if vkDestroyPipeline was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Destroys a graphics, compute, or ray tracing pipeline. The pipeline must not be bound in any command buffer that is still pending execution.

Pipelines are expensive to create but cheap to keep around. Only destroy them when you are certain they will not be needed again (e.g. during level transitions or application shutdown).

Shader modules used to create the pipeline can be destroyed independently, the pipeline retains its own copy of the compiled state.

pub unsafe fn create_pipeline_layout( &self, p_create_info: &PipelineLayoutCreateInfo, allocator: Option<&AllocationCallbacks>, ) -> VkResult<PipelineLayout>

Wraps vkCreatePipelineLayout. Provided by VK_COMPUTE_VERSION_1_0.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkCreatePipelineLayout was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

A pipeline layout defines the interface between shader stages and the descriptor sets and push constants that feed them. It specifies:

• Descriptor set layouts: which bindings are available at each set index (0, 1, 2, …).
• Push constant ranges: byte ranges per shader stage for small, frequently-updated data.

Set layout ordering convention: a common pattern is:

• Set 0: per-frame data (camera, time).
• Set 1: per-material data (textures, material params).
• Set 2: per-object data (transforms).

This lets you bind set 0 once per frame and only rebind sets 1–2 as materials and objects change, minimising descriptor set switches.

Pipeline layouts are immutable after creation. Two pipelines that share the same layout can share descriptor sets without rebinding.

Push constants are limited to max_push_constants_size bytes (guaranteed at least 128). Use them for small per-draw data like transform matrices or material indices.

Guide

See Pipelines in the vulkan_rust guide.

pub unsafe fn destroy_pipeline_layout( &self, pipeline_layout: PipelineLayout, allocator: Option<&AllocationCallbacks>, )

Wraps vkDestroyPipelineLayout. Provided by VK_COMPUTE_VERSION_1_0.

Safety

• device (self) must be valid and not destroyed.
• pipelineLayout must be externally synchronized.

Panics

Panics if vkDestroyPipelineLayout was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Destroys a pipeline layout. All pipelines and descriptor sets that were created with this layout must no longer be in use.

In practice, pipeline layouts are typically created once and live for the duration of the application or a major rendering context. There is little reason to destroy them early.

pub unsafe fn create_sampler( &self, p_create_info: &SamplerCreateInfo, allocator: Option<&AllocationCallbacks>, ) -> VkResult<Sampler>

Wraps vkCreateSampler. Provided by VK_COMPUTE_VERSION_1_0.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_INVALID_OPAQUE_CAPTURE_ADDRESS_KHR
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkCreateSampler was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Creates a sampler that controls how shaders read image data: filtering, addressing, mip level selection, and anisotropy.

Common configurations:

• Nearest/point: MIN_FILTER_NEAREST, MAG_FILTER_NEAREST. No interpolation, pixel art, data textures, or shadow map comparison.
• Bilinear: MIN_FILTER_LINEAR, MAG_FILTER_LINEAR, MIPMAP_MODE_NEAREST. Smooth within a mip level but snaps between levels.
• Trilinear: same as bilinear but with MIPMAP_MODE_LINEAR. Smooth transitions between mip levels. The default choice for most 3D textures.
• Anisotropic: enable anisotropy_enable and set max_anisotropy (commonly 4–16). Improves quality at oblique viewing angles at a small GPU cost.

Address modes (REPEAT, MIRRORED_REPEAT, CLAMP_TO_EDGE, CLAMP_TO_BORDER) control what happens when UVs go outside [0, 1].

Samplers are immutable after creation and can be shared across any number of descriptor sets. Most applications need only a handful of samplers.

pub unsafe fn destroy_sampler( &self, sampler: Sampler, allocator: Option<&AllocationCallbacks>, )

Wraps vkDestroySampler. Provided by VK_COMPUTE_VERSION_1_0.

Safety

• device (self) must be valid and not destroyed.
• sampler must be externally synchronized.

Panics

Panics if vkDestroySampler was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Destroys a sampler. The sampler must not be referenced by any descriptor set that is bound in a pending command buffer.

Since most applications use a small fixed set of samplers, they are typically created once at startup and destroyed only during application shutdown.

pub unsafe fn create_descriptor_set_layout( &self, p_create_info: &DescriptorSetLayoutCreateInfo, allocator: Option<&AllocationCallbacks>, ) -> VkResult<DescriptorSetLayout>

Wraps vkCreateDescriptorSetLayout. Provided by VK_COMPUTE_VERSION_1_0.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkCreateDescriptorSetLayout was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

A descriptor set layout defines the shape of a descriptor set: which binding numbers exist, what descriptor type each binding holds, and at which shader stages each binding is visible.

Binding tips:

• Keep stage_flags as narrow as possible. Declaring a binding visible to all stages when only the fragment shader uses it wastes driver resources on some implementations.
• Use DESCRIPTOR_TYPE_COMBINED_IMAGE_SAMPLER for simple texture sampling. Separate sampler + sampled image bindings offer more flexibility when you want to reuse samplers across many textures.
• Array descriptors (descriptor_count > 1) map to GLSL arrays. Useful for bindless or material-table patterns.

Layouts are immutable after creation and can be shared across multiple pipeline layouts and descriptor set allocations.

Destroy with destroy_descriptor_set_layout when no pipeline layout or pending descriptor set allocation still references it.

pub unsafe fn destroy_descriptor_set_layout( &self, descriptor_set_layout: DescriptorSetLayout, allocator: Option<&AllocationCallbacks>, )

Wraps vkDestroyDescriptorSetLayout. Provided by VK_COMPUTE_VERSION_1_0.

Safety

• device (self) must be valid and not destroyed.
• descriptorSetLayout must be externally synchronized.

Panics

Panics if vkDestroyDescriptorSetLayout was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Destroys a descriptor set layout. The layout must not be referenced by any pipeline layout or pending descriptor set allocation that is still in use.

Descriptor set layouts are lightweight and typically long-lived. Destroy them during application shutdown after all dependent pipeline layouts and descriptor pools have been destroyed.

pub unsafe fn create_descriptor_pool( &self, p_create_info: &DescriptorPoolCreateInfo, allocator: Option<&AllocationCallbacks>, ) -> VkResult<DescriptorPool>

Wraps vkCreateDescriptorPool. Provided by VK_COMPUTE_VERSION_1_0.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_FRAGMENTATION_EXT
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkCreateDescriptorPool was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Creates a pool from which descriptor sets are allocated. The pool must be sized to accommodate all the descriptor sets and individual descriptor types your application needs.

Sizing: specify max_sets (total descriptor sets) and a list of DescriptorPoolSize entries that declare how many descriptors of each type the pool holds. Under-sizing causes VK_ERROR_OUT_OF_POOL_MEMORY at allocation time.

Flags:

• FREE_DESCRIPTOR_SET: allows individual sets to be freed with free_descriptor_sets. Without this flag, sets can only be reclaimed by resetting the entire pool.
• UPDATE_AFTER_BIND: required if any allocated set uses update-after-bind bindings.

Common pattern: create one pool per frame-in-flight. At the start of each frame, reset_descriptor_pool reclaims all sets at once, no individual tracking or freeing needed.

pub unsafe fn destroy_descriptor_pool( &self, descriptor_pool: DescriptorPool, allocator: Option<&AllocationCallbacks>, )

Wraps vkDestroyDescriptorPool. Provided by VK_COMPUTE_VERSION_1_0.

Safety

• device (self) must be valid and not destroyed.
• descriptorPool must be externally synchronized.

Panics

Panics if vkDestroyDescriptorPool was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Destroys a descriptor pool and implicitly frees all descriptor sets allocated from it. You do not need to free individual sets before destroying the pool.

Ensure no command buffer that references any set from this pool is still pending execution.

pub unsafe fn reset_descriptor_pool( &self, descriptor_pool: DescriptorPool, flags: DescriptorPoolResetFlags, ) -> VkResult<()>

Wraps vkResetDescriptorPool. Provided by VK_COMPUTE_VERSION_1_0.

Errors

• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.
• descriptorPool must be externally synchronized.

Panics

Panics if vkResetDescriptorPool was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Recycles all descriptor sets allocated from this pool back to the pool, without destroying the pool itself. After a reset, all previously allocated sets are invalid and must not be used.

This is the fastest way to reclaim descriptor sets, much cheaper than freeing them individually. Ideal for the per-frame pool pattern where you allocate fresh sets every frame and reset the pool at the start of the next frame.

No command buffer that references any set from this pool may be pending execution when you reset.

pub unsafe fn allocate_descriptor_sets( &self, p_allocate_info: &DescriptorSetAllocateInfo, ) -> VkResult<Vec<DescriptorSet>>

Wraps vkAllocateDescriptorSets. Provided by VK_COMPUTE_VERSION_1_0.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_FRAGMENTED_POOL
• VK_ERROR_OUT_OF_POOL_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkAllocateDescriptorSets was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Allocates descriptor sets from a descriptor pool. Each set is backed by one of the set_layouts in the allocate info.

Common failure causes:

• Pool exhaustion: the pool does not have enough descriptors of the required types, or the maximum set count has been reached. Returns VK_ERROR_OUT_OF_POOL_MEMORY. Pre-calculate pool sizes carefully or use VK_DESCRIPTOR_POOL_CREATE_FREE_DESCRIPTOR_SET_BIT to allow freeing and reallocation.
• Fragmentation: even with enough total descriptors, internal fragmentation can cause allocation failure. Resetting the entire pool with reset_descriptor_pool defragments it.

Descriptor sets become invalid when their parent pool is destroyed or reset. Do not submit command buffers that reference descriptor sets from a pool that has been reset.

For frequently-updated descriptors, consider VK_KHR_push_descriptor which avoids set allocation entirely.

pub unsafe fn free_descriptor_sets( &self, descriptor_pool: DescriptorPool, p_descriptor_sets: &[DescriptorSet], ) -> VkResult<()>

Wraps vkFreeDescriptorSets. Provided by VK_COMPUTE_VERSION_1_0.

Errors

• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.
• descriptorPool must be externally synchronized.
• pDescriptorSets must be externally synchronized.

Panics

Panics if vkFreeDescriptorSets was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Returns individual descriptor sets back to their parent pool. The pool must have been created with FREE_DESCRIPTOR_SET, without that flag this call is invalid.

For most applications, resetting the entire pool with reset_descriptor_pool is simpler and faster than tracking and freeing individual sets. Use free_descriptor_sets only when you need fine-grained lifetime control over specific sets.

Freed sets must not be referenced by any pending command buffer.

pub unsafe fn update_descriptor_sets( &self, p_descriptor_writes: &[WriteDescriptorSet], p_descriptor_copies: &[CopyDescriptorSet], )

Wraps vkUpdateDescriptorSets. Provided by VK_COMPUTE_VERSION_1_0.

Safety

• device (self) must be valid and not destroyed.
• pDescriptorWrites must be externally synchronized.

Panics

Panics if vkUpdateDescriptorSets was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Writes or copies resource bindings into descriptor sets. This is how you connect actual buffers, images, and samplers to the descriptor slots that shaders read from.

Writes (WriteDescriptorSet): bind concrete resources to a specific set + binding + array element. Each write targets one descriptor type (uniform buffer, combined image sampler, storage buffer, etc.).

Copies (CopyDescriptorSet): duplicate bindings from one set to another. Rarely used, writes cover nearly all cases.

Updates take effect immediately and are visible to any command buffer recorded after the update. However, updating a set that is currently bound in a pending command buffer is undefined behaviour unless the set was allocated from a pool with UPDATE_AFTER_BIND and the binding is marked as update-after-bind in the layout.

Batch multiple writes into a single call when possible, the driver can often process them more efficiently.

pub unsafe fn create_framebuffer( &self, p_create_info: &FramebufferCreateInfo, allocator: Option<&AllocationCallbacks>, ) -> VkResult<Framebuffer>

Wraps vkCreateFramebuffer. Provided by VK_GRAPHICS_VERSION_1_0.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkCreateFramebuffer was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

A framebuffer binds concrete image views to the attachment slots defined by a render pass. The number and format of attachments must match the render pass exactly, mismatches cause validation errors.

Dimensions: width, height, and layers must be less than or equal to the corresponding dimensions of every attached image view. They define the renderable area for cmd_begin_render_pass.

Lifetime: the framebuffer must stay alive for the entire duration of any render pass instance that uses it. In practice, framebuffers are typically recreated when the swapchain is resized.

Imageless framebuffers (Vulkan 1.2+): create the framebuffer with FRAMEBUFFER_CREATE_IMAGELESS and no attachments. Concrete image views are then supplied at cmd_begin_render_pass time via RenderPassAttachmentBeginInfo. This avoids recreating framebuffers on swapchain resize.

Guide

See Render Passes & Framebuffers in the vulkan_rust guide.

pub unsafe fn destroy_framebuffer( &self, framebuffer: Framebuffer, allocator: Option<&AllocationCallbacks>, )

Wraps vkDestroyFramebuffer. Provided by VK_GRAPHICS_VERSION_1_0.

Safety

• device (self) must be valid and not destroyed.
• framebuffer must be externally synchronized.

Panics

Panics if vkDestroyFramebuffer was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Destroys a framebuffer. No render pass instance using this framebuffer may be pending execution.

Framebuffers are typically recreated whenever the swapchain is resized, so they tend to have shorter lifetimes than most Vulkan objects. With imageless framebuffers (Vulkan 1.2+) you can avoid this churn entirely.

pub unsafe fn create_render_pass( &self, p_create_info: &RenderPassCreateInfo, allocator: Option<&AllocationCallbacks>, ) -> VkResult<RenderPass>

Wraps vkCreateRenderPass. Provided by VK_GRAPHICS_VERSION_1_0.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkCreateRenderPass was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

A render pass describes the attachments, subpasses, and dependencies used during rendering. It does not reference actual images, those are bound later via a framebuffer.

Key design points:

• load_op / store_op: use DONT_CARE for attachments whose prior contents are irrelevant (e.g. a transient depth buffer). This lets tile-based GPUs skip loads/stores, which is significant on mobile.
• initial_layout / final_layout: Vulkan inserts implicit layout transitions at render pass boundaries. Set these to match your actual usage to avoid unnecessary transitions. UNDEFINED for initial_layout is fine when load_op is CLEAR or DONT_CARE.
• Subpass dependencies: the implicit external dependencies (VK_SUBPASS_EXTERNAL) are often insufficient. Add explicit dependencies when subsequent passes read the output.

For dynamic rendering (Vulkan 1.3+), consider cmd_begin_rendering instead, which avoids the need for render pass and framebuffer objects.

Guide

See Render Passes & Framebuffers in the vulkan_rust guide.

pub unsafe fn destroy_render_pass( &self, render_pass: RenderPass, allocator: Option<&AllocationCallbacks>, )

Wraps vkDestroyRenderPass. Provided by VK_GRAPHICS_VERSION_1_0.

Safety

• device (self) must be valid and not destroyed.
• renderPass must be externally synchronized.

Panics

Panics if vkDestroyRenderPass was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Destroys a render pass. All framebuffers and pipelines created with this render pass must no longer be in use.

Render passes are typically long-lived, created once at startup and destroyed during shutdown. Destroying a render pass does not affect pipelines that were created with a compatible render pass (same attachment count, formats, and sample counts).

pub unsafe fn get_render_area_granularity( &self, render_pass: RenderPass, ) -> Extent2D

Wraps vkGetRenderAreaGranularity. Provided by VK_GRAPHICS_VERSION_1_0.

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetRenderAreaGranularity was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Returns the granularity (in pixels) at which the render area should be aligned for optimal performance with this render pass.

The render area passed to cmd_begin_render_pass should have its offset be a multiple of this granularity and its extent should either cover the full framebuffer or be rounded up to a multiple.

On most desktop GPUs this returns (1, 1), meaning any alignment is fine. On tile-based GPUs this may return the tile size (e.g. 32×32), and misalignment can cause partial-tile overhead.

In practice, most applications render to the full framebuffer extent and never need to worry about this.

pub unsafe fn get_rendering_area_granularity( &self, p_rendering_area_info: &RenderingAreaInfo, ) -> Extent2D

Wraps vkGetRenderingAreaGranularity. Provided by VK_GRAPHICS_VERSION_1_4.

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetRenderingAreaGranularity was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Vulkan 1.4 command that queries the optimal render area granularity for a given dynamic rendering configuration, without needing a render pass object.

This is the dynamic rendering equivalent of get_render_area_granularity. Pass a RenderingAreaInfo describing the attachment formats and sample count, and receive the granularity (in pixels) at which the render area should be aligned.

On most desktop GPUs this returns (1, 1). On tile-based GPUs the granularity may match the tile size.

pub unsafe fn create_command_pool( &self, p_create_info: &CommandPoolCreateInfo, allocator: Option<&AllocationCallbacks>, ) -> VkResult<CommandPool>

Wraps vkCreateCommandPool. Provided by VK_BASE_VERSION_1_0.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkCreateCommandPool was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

A command pool provides the memory backing for command buffers allocated from it. Pools are tied to a single queue family, command buffers allocated from the pool can only be submitted to queues of that family.

Flags:

• TRANSIENT: hint that command buffers are short-lived and reset or freed frequently. Lets the driver use a faster allocation strategy.
• RESET_COMMAND_BUFFER: allows individual command buffers to be reset via reset_command_buffer. Without this flag you must reset the entire pool with reset_command_pool.

A common pattern is one pool per frame-in-flight per thread: reset the whole pool at the start of each frame instead of managing individual command buffer lifetimes.

Command pools are not thread-safe. If multiple threads record commands concurrently, each thread needs its own pool.

Guide

See Command Buffers in the vulkan_rust guide.

pub unsafe fn destroy_command_pool( &self, command_pool: CommandPool, allocator: Option<&AllocationCallbacks>, )

Wraps vkDestroyCommandPool. Provided by VK_BASE_VERSION_1_0.

Safety

• device (self) must be valid and not destroyed.
• commandPool must be externally synchronized.

Panics

Panics if vkDestroyCommandPool was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Destroys a command pool and implicitly frees all command buffers allocated from it. You do not need to free individual command buffers before destroying the pool.

All command buffers allocated from this pool must have completed execution before the pool is destroyed. Call device_wait_idle or wait on the relevant fences first.

pub unsafe fn reset_command_pool( &self, command_pool: CommandPool, flags: CommandPoolResetFlags, ) -> VkResult<()>

Wraps vkResetCommandPool. Provided by VK_BASE_VERSION_1_0.

Errors

• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.
• commandPool must be externally synchronized.

Panics

Panics if vkResetCommandPool was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Resets a command pool, recycling all command buffers allocated from it back to the initial state. This is faster than resetting individual command buffers.

Flags:

• RELEASE_RESOURCES: return memory to the system. Without this flag, the pool keeps its internal memory for reuse by future allocations, usually what you want in a frame loop.

Per-frame pattern: reset the pool at the start of each frame (without RELEASE_RESOURCES), then re-record command buffers from the same pool. Memory is reused without reallocation overhead.

All command buffers from this pool must have completed execution before resetting.

pub unsafe fn allocate_command_buffers( &self, p_allocate_info: &CommandBufferAllocateInfo, ) -> VkResult<Vec<CommandBuffer>>

Wraps vkAllocateCommandBuffers. Provided by VK_BASE_VERSION_1_0.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkAllocateCommandBuffers was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Allocates one or more command buffers from a command pool. The command_buffer_count and output slice length must match.

Level:

• PRIMARY: submitted directly to a queue via queue_submit.
• SECONDARY: recorded separately and executed inside a primary buffer via cmd_execute_commands. Useful for pre-recording draw calls that are reused across frames.

All allocated command buffers start in the initial state and must be recorded with begin_command_buffer before submission.

Command buffers are freed either individually with free_command_buffers or implicitly when the parent pool is destroyed or reset.

Guide

See Command Buffers in the vulkan_rust guide.

pub unsafe fn free_command_buffers( &self, command_pool: CommandPool, p_command_buffers: &[CommandBuffer], )

Wraps vkFreeCommandBuffers. Provided by VK_BASE_VERSION_1_0.

Safety

• device (self) must be valid and not destroyed.
• commandPool must be externally synchronized.
• pCommandBuffers must be externally synchronized.

Panics

Panics if vkFreeCommandBuffers was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Returns individual command buffers to their parent pool. The command buffers must not be pending execution.

For most applications, resetting the entire pool with reset_command_pool is simpler. Use free_command_buffers only when you need fine-grained lifetime control, for example, freeing a one-shot transfer command buffer immediately after use while keeping other buffers from the same pool alive.

Freed command buffer handles become invalid. Do not resubmit them.

pub unsafe fn begin_command_buffer( &self, command_buffer: CommandBuffer, p_begin_info: &CommandBufferBeginInfo, ) -> VkResult<()>

Wraps vkBeginCommandBuffer. Provided by VK_BASE_VERSION_1_0.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkBeginCommandBuffer was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Begins recording commands into a command buffer. The command buffer must be in the initial state, either freshly allocated, or reset via reset_command_buffer / reset_command_pool.

Flags:

• ONE_TIME_SUBMIT: the command buffer will be submitted once and then reset or freed. Lets the driver skip internal tracking it would otherwise need for resubmission.
• SIMULTANEOUS_USE: the command buffer can be pending execution on multiple queues simultaneously. Required for secondary command buffers reused across multiple primary buffers.

Inheritance info: only required for secondary command buffers. When recording a secondary buffer that will execute inside a render pass, set render_pass, subpass, and optionally framebuffer in the CommandBufferInheritanceInfo. For primary buffers the inheritance info is ignored.

Calling begin_command_buffer on a buffer that is already recording is an error. Calling it on a buffer in the executable state implicitly resets it first (if the pool allows it).

Guide

See Command Buffers in the vulkan_rust guide.

pub unsafe fn end_command_buffer( &self, command_buffer: CommandBuffer, ) -> VkResult<()>

Wraps vkEndCommandBuffer. Provided by VK_BASE_VERSION_1_0.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_INVALID_VIDEO_STD_PARAMETERS_KHR
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkEndCommandBuffer was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Finishes recording a command buffer. After this call the command buffer moves from the recording state to the executable state and can be submitted via queue_submit.

If an error occurred during recording (e.g. out of memory), this call returns the error. Always check the return value, a failed end_command_buffer means the command buffer is in an invalid state and must be reset before reuse.

A command buffer that is inside a render pass must end the render pass with cmd_end_render_pass before calling end_command_buffer.

Guide

See Command Buffers in the vulkan_rust guide.

pub unsafe fn reset_command_buffer( &self, command_buffer: CommandBuffer, flags: CommandBufferResetFlags, ) -> VkResult<()>

Wraps vkResetCommandBuffer. Provided by VK_BASE_VERSION_1_0.

Errors

• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkResetCommandBuffer was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Resets a single command buffer back to the initial state so it can be re-recorded. The command pool must have been created with RESET_COMMAND_BUFFER for this to be valid.

Flags:

• RELEASE_RESOURCES: return the command buffer’s memory to the pool. Without this flag, memory is retained for reuse during the next recording, usually preferred in a frame loop.

For bulk resets, reset_command_pool is more efficient than resetting buffers individually.

The command buffer must not be pending execution when reset.

pub unsafe fn cmd_bind_pipeline( &self, command_buffer: CommandBuffer, pipeline_bind_point: PipelineBindPoint, pipeline: Pipeline, )

Wraps vkCmdBindPipeline. Provided by VK_COMPUTE_VERSION_1_0.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdBindPipeline was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Binds a pipeline to a command buffer for subsequent draw or dispatch calls. The pipeline_bind_point must match the pipeline type:

• PIPELINE_BIND_POINT_GRAPHICS for graphics pipelines.
• PIPELINE_BIND_POINT_COMPUTE for compute pipelines.
• PIPELINE_BIND_POINT_RAY_TRACING_KHR for ray tracing pipelines.

Binding a pipeline invalidates any incompatible dynamic state. For example, binding a new graphics pipeline that uses dynamic viewport requires you to call cmd_set_viewport again before drawing.

Pipeline binds are relatively cheap, the driver patches command state internally. Minimise binds by sorting draw calls by pipeline when possible, but do not over-optimise at the expense of code clarity.

Graphics pipelines can only be bound inside a render pass (or dynamic rendering). Compute pipelines can be bound anywhere.

Guide

See Pipelines in the vulkan_rust guide.

pub unsafe fn cmd_set_attachment_feedback_loop_enable_ext( &self, command_buffer: CommandBuffer, aspect_mask: ImageAspectFlags, )

Wraps vkCmdSetAttachmentFeedbackLoopEnableEXT. Provided by VK_EXT_attachment_feedback_loop_dynamic_state.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdSetAttachmentFeedbackLoopEnableEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Dynamically enables or disables attachment feedback loops for specific image aspects (color, depth, stencil). When enabled, shaders can both read from and write to the same attachment within a render pass.

Use with care, feedback loops create read-after-write hazards. The implementation handles coherency when this flag is set.

Requires VK_EXT_attachment_feedback_loop_dynamic_state.

pub unsafe fn cmd_set_viewport( &self, command_buffer: CommandBuffer, first_viewport: u32, p_viewports: &[Viewport], )

Wraps vkCmdSetViewport. Provided by VK_GRAPHICS_VERSION_1_0.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdSetViewport was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Sets the viewport transform dynamically. Only takes effect if the pipeline was created with DYNAMIC_STATE_VIEWPORT.

The viewport defines the mapping from normalised device coordinates to framebuffer coordinates. Most applications use a single viewport covering the full render target:

Viewport { x: 0.0, y: 0.0, width: w, height: h, min_depth: 0.0, max_depth: 1.0 }

Flipped Y: Vulkan’s clip space has Y pointing downward (unlike OpenGL). To match OpenGL conventions, use a negative height and offset Y by the framebuffer height: Viewport { y: h, height: -h, ... }. This requires Vulkan 1.1 or VK_KHR_maintenance1.

Multiple viewports are supported for multi-view rendering (e.g. VR). The first_viewport parameter selects which viewport index to start writing at.

pub unsafe fn cmd_set_scissor( &self, command_buffer: CommandBuffer, first_scissor: u32, p_scissors: &[Rect2D], )

Wraps vkCmdSetScissor. Provided by VK_GRAPHICS_VERSION_1_0.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdSetScissor was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Sets the scissor rectangle dynamically. Only takes effect if the pipeline was created with DYNAMIC_STATE_SCISSOR.

The scissor test discards fragments outside the rectangle. Unlike the viewport (which transforms coordinates), the scissor is a hard clip in framebuffer pixel coordinates.

A common default is a scissor covering the full framebuffer:

Rect2D { offset: { x: 0, y: 0 }, extent: { width: w, height: h } }

Scissor rectangles are useful for UI rendering, split-screen, or any case where you want to restrict rendering to a sub-region without changing the viewport transform.

The scissor must be set before any draw call when using dynamic scissor state, even if it covers the full framebuffer.

pub unsafe fn cmd_set_line_width( &self, command_buffer: CommandBuffer, line_width: f32, )

Wraps vkCmdSetLineWidth. Provided by VK_GRAPHICS_VERSION_1_0.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdSetLineWidth was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Sets the width of rasterised line primitives dynamically. Only takes effect if the pipeline was created with DYNAMIC_STATE_LINE_WIDTH.

The default line width is 1.0. Wide lines (width > 1.0) require the wide_lines device feature, check physical_device_features.wide_lines before using.

The supported range is device-dependent (query physical_device_limits.line_width_range). If wide_lines is not supported, only 1.0 is valid.

Line width is specified in framebuffer pixels. Anti-aliased lines may be rendered slightly wider than the specified width due to coverage calculations.

pub unsafe fn cmd_set_depth_bias( &self, command_buffer: CommandBuffer, depth_bias_constant_factor: f32, depth_bias_clamp: f32, depth_bias_slope_factor: f32, )

Wraps vkCmdSetDepthBias. Provided by VK_GRAPHICS_VERSION_1_0.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdSetDepthBias was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Sets depth bias parameters dynamically. Only takes effect if the pipeline was created with DYNAMIC_STATE_DEPTH_BIAS and depth bias is enabled in the rasterisation state.

Depth bias adds a computed offset to each fragment’s depth value before the depth test. The primary use case is shadow mapping, biasing shadow caster geometry slightly away from the light prevents self-shadowing artifacts (shadow acne).

The final bias is computed as:

bias = constant_factor * r + slope_factor * max_slope

where r is the minimum resolvable depth difference and max_slope is the maximum depth slope of the triangle.

depth_bias_clamp limits the maximum bias value. Requires the depth_bias_clamp device feature. A clamp of 0.0 disables clamping.

Typical shadow map values: constant_factor = 1.25, slope_factor = 1.75, clamp = 0.0. Tune per scene.

pub unsafe fn cmd_set_blend_constants( &self, command_buffer: CommandBuffer, blend_constants: f32, )

Wraps vkCmdSetBlendConstants. Provided by VK_GRAPHICS_VERSION_1_0.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdSetBlendConstants was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Sets the constant blend colour used when a blend factor references BLEND_FACTOR_CONSTANT_COLOR, BLEND_FACTOR_CONSTANT_ALPHA, or their one-minus variants. Only takes effect if the pipeline was created with DYNAMIC_STATE_BLEND_CONSTANTS.

The four values are RGBA in [0.0, 1.0]. A common use is fading geometry by setting a constant alpha and blending with BLEND_FACTOR_CONSTANT_ALPHA.

If your pipeline does not use any constant blend factors, you do not need to set this state. The values are ignored for blend modes that do not reference them.

pub unsafe fn cmd_set_depth_bounds( &self, command_buffer: CommandBuffer, min_depth_bounds: f32, max_depth_bounds: f32, )

Wraps vkCmdSetDepthBounds. Provided by VK_GRAPHICS_VERSION_1_0.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdSetDepthBounds was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Sets the depth bounds test range dynamically. Only takes effect if the pipeline was created with DYNAMIC_STATE_DEPTH_BOUNDS and depth bounds testing is enabled in the depth-stencil state.

The depth bounds test discards fragments whose depth buffer value falls outside [min_depth_bounds, max_depth_bounds]. Note that this tests the existing depth buffer value, not the fragment’s incoming depth.

Use cases are niche:

• Stencil shadow volumes: reject fragments that are clearly outside the shadow volume’s depth range.
• Deferred shading light volumes: skip fragments outside the light’s depth range.

Requires the depth_bounds device feature. Not supported on all hardware, check physical_device_features.depth_bounds before enabling.

pub unsafe fn cmd_set_stencil_compare_mask( &self, command_buffer: CommandBuffer, face_mask: StencilFaceFlags, compare_mask: u32, )

Wraps vkCmdSetStencilCompareMask. Provided by VK_GRAPHICS_VERSION_1_0.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdSetStencilCompareMask was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Sets the stencil compare mask dynamically for front-facing, back-facing, or both face sets. Only takes effect if the pipeline was created with DYNAMIC_STATE_STENCIL_COMPARE_MASK.

The compare mask is ANDed with both the reference value and the stencil buffer value before the stencil comparison. This lets you use individual bits of the stencil buffer for different purposes (e.g. bit 0 for portals, bits 1–3 for decal layers).

A mask of 0xFF (the default) uses all 8 bits of the stencil buffer. Narrower masks isolate specific bit planes for multi-purpose stencil schemes.

pub unsafe fn cmd_set_stencil_write_mask( &self, command_buffer: CommandBuffer, face_mask: StencilFaceFlags, write_mask: u32, )

Wraps vkCmdSetStencilWriteMask. Provided by VK_GRAPHICS_VERSION_1_0.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdSetStencilWriteMask was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Sets the stencil write mask dynamically for front-facing, back-facing, or both face sets. Only takes effect if the pipeline was created with DYNAMIC_STATE_STENCIL_WRITE_MASK.

The write mask controls which bits of the stencil buffer are updated by stencil operations (KEEP, REPLACE, INCREMENT, etc.). Bits that are zero in the mask are left unchanged.

A mask of 0xFF writes all 8 bits. Use narrower masks when different rendering passes own different stencil bit planes.

pub unsafe fn cmd_set_stencil_reference( &self, command_buffer: CommandBuffer, face_mask: StencilFaceFlags, reference: u32, )

Wraps vkCmdSetStencilReference. Provided by VK_GRAPHICS_VERSION_1_0.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdSetStencilReference was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Sets the stencil reference value dynamically for front-facing, back-facing, or both face sets. Only takes effect if the pipeline was created with DYNAMIC_STATE_STENCIL_REFERENCE.

The reference value is used in stencil comparison operations (e.g. COMPARE_OP_EQUAL compares the masked buffer value against the masked reference) and as the source value for STENCIL_OP_REPLACE.

Common patterns:

• Portal/mirror rendering: set reference to a unique ID per portal, write it with STENCIL_OP_REPLACE, then test with COMPARE_OP_EQUAL to mask subsequent draws to that portal’s region.
• Decal layering: increment the reference per layer.

pub unsafe fn cmd_bind_descriptor_sets( &self, command_buffer: CommandBuffer, pipeline_bind_point: PipelineBindPoint, layout: PipelineLayout, first_set: u32, p_descriptor_sets: &[DescriptorSet], p_dynamic_offsets: &[u32], )

Wraps vkCmdBindDescriptorSets. Provided by VK_COMPUTE_VERSION_1_0.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdBindDescriptorSets was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Binds one or more descriptor sets to a command buffer at specified set indices. Subsequent draw or dispatch calls read resources from these bound sets.

first_set: the set index at which binding starts. Sets at lower indices are not disturbed. This lets you bind per-frame data at set 0 once and rebind only per-material or per-object sets at higher indices.

Dynamic offsets: for descriptors of type UNIFORM_BUFFER_DYNAMIC or STORAGE_BUFFER_DYNAMIC, the dynamic_offsets slice provides byte offsets applied at bind time. This lets multiple draw calls share a single large buffer with different sub-regions without updating the descriptor set.

The bound pipeline layout and the descriptor set layouts must be compatible, same binding layout at each set index. Binding a set with an incompatible layout is undefined behaviour.

pub unsafe fn cmd_bind_index_buffer( &self, command_buffer: CommandBuffer, buffer: Buffer, offset: u64, index_type: IndexType, )

Wraps vkCmdBindIndexBuffer. Provided by VK_GRAPHICS_VERSION_1_0.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdBindIndexBuffer was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Binds an index buffer for subsequent indexed draw calls (cmd_draw_indexed, cmd_draw_indexed_indirect).

Index type:

• INDEX_TYPE_UINT16: 2 bytes per index. Good for meshes with fewer than 65536 vertices, saves memory bandwidth.
• INDEX_TYPE_UINT32: 4 bytes per index. Required for large meshes.
• INDEX_TYPE_UINT8_KHR (extension): 1 byte per index for very small meshes.

The buffer must have been created with BUFFER_USAGE_INDEX_BUFFER. The offset must be a multiple of the index type size (2 for UINT16, 4 for UINT32).

Only one index buffer can be bound at a time, binding a new one replaces the previous binding.

pub unsafe fn cmd_bind_vertex_buffers( &self, command_buffer: CommandBuffer, first_binding: u32, p_buffers: &[Buffer], p_offsets: &u64, )

Wraps vkCmdBindVertexBuffers. Provided by VK_GRAPHICS_VERSION_1_0.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdBindVertexBuffers was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Binds one or more vertex buffers to input binding slots for subsequent draw calls.

first_binding: the binding slot index to start at. Binding slots are defined in the pipeline’s vertex input state. Multiple buffers can be bound to consecutive slots in a single call.

Interleaved vs separate: a single buffer with interleaved attributes (position + normal + UV) uses one binding slot. Separate attribute streams (one buffer per attribute) use multiple slots. Interleaved is generally more cache-friendly.

Buffers must have been created with BUFFER_USAGE_VERTEX_BUFFER. The offsets array specifies the byte offset within each buffer where vertex data starts.

For Vulkan 1.3+, cmd_bind_vertex_buffers2 also lets you set buffer sizes and strides dynamically.

pub unsafe fn cmd_draw( &self, command_buffer: CommandBuffer, vertex_count: u32, instance_count: u32, first_vertex: u32, first_instance: u32, )

Wraps vkCmdDraw. Provided by VK_GRAPHICS_VERSION_1_0.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdDraw was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Records a non-indexed draw call. Vertices are generated sequentially from first_vertex to first_vertex + vertex_count - 1.

Parameters:

• vertex_count: number of vertices to draw.
• instance_count: number of instances. Use 1 for non-instanced drawing.
• first_vertex: offset into the vertex buffer (added to gl_VertexIndex in the shader).
• first_instance: offset into instance data (added to gl_InstanceIndex). Requires the first_instance feature if non-zero.

For indexed geometry (the common case for meshes), use cmd_draw_indexed instead. cmd_draw is typically used for full-screen quads, procedural geometry, or particle systems where vertices are generated in the shader.

Guide

See Command Buffers in the vulkan_rust guide.

pub unsafe fn cmd_draw_indexed( &self, command_buffer: CommandBuffer, index_count: u32, instance_count: u32, first_index: u32, vertex_offset: i32, first_instance: u32, )

Wraps vkCmdDrawIndexed. Provided by VK_GRAPHICS_VERSION_1_0.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdDrawIndexed was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Records an indexed draw call, the standard way to draw meshes. Indices are read from the bound index buffer, and each index is used to fetch vertex attributes from the bound vertex buffers.

Parameters:

• index_count: number of indices to read.
• instance_count: number of instances. Use 1 for non-instanced.
• first_index: offset into the index buffer (in units of indices, not bytes).
• vertex_offset: added to each index value before fetching the vertex. Useful for packing multiple meshes into a single vertex buffer at different offsets.
• first_instance: offset into instance data.

Indexed drawing reuses vertices via the index buffer, saving memory and bandwidth compared to non-indexed draws for any mesh with shared vertices (which is nearly all of them).

pub unsafe fn cmd_draw_multi_ext( &self, command_buffer: CommandBuffer, p_vertex_info: &[MultiDrawInfoEXT], instance_count: u32, first_instance: u32, stride: u32, )

Wraps vkCmdDrawMultiEXT. Provided by VK_EXT_multi_draw.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdDrawMultiEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Draws multiple non-indexed draw calls from an array of MultiDrawInfoEXT (first_vertex, vertex_count pairs). More efficient than issuing separate cmd_draw calls because the driver can batch them.

Requires VK_EXT_multi_draw and the multiDraw feature.

pub unsafe fn cmd_draw_multi_indexed_ext( &self, command_buffer: CommandBuffer, p_index_info: &[MultiDrawIndexedInfoEXT], instance_count: u32, first_instance: u32, stride: u32, p_vertex_offset: *const i32, )

Wraps vkCmdDrawMultiIndexedEXT. Provided by VK_EXT_multi_draw.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdDrawMultiIndexedEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Draws multiple indexed draw calls from an array of MultiDrawIndexedInfoEXT (first_index, index_count, vertex_offset triples). More efficient than separate cmd_draw_indexed calls.

An optional p_vertex_offset overrides all per-draw vertex offsets with a single value.

Requires VK_EXT_multi_draw and the multiDraw feature.

pub unsafe fn cmd_draw_indirect( &self, command_buffer: CommandBuffer, buffer: Buffer, offset: u64, draw_count: u32, stride: u32, )

Wraps vkCmdDrawIndirect. Provided by VK_GRAPHICS_VERSION_1_0.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdDrawIndirect was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Records one or more non-indexed draw calls whose parameters are read from a GPU buffer at execution time rather than baked into the command buffer.

Each DrawIndirectCommand in the buffer contains vertex_count, instance_count, first_vertex, and first_instance, the same parameters as cmd_draw.

Use cases:

• GPU-driven rendering: a compute shader fills the indirect buffer with draw parameters (e.g. after culling), and the GPU draws without CPU round-trips.
• Conditional draw counts: pair with cmd_draw_indirect_count (Vulkan 1.2) to have the GPU also determine how many draws to execute.

The buffer must have been created with BUFFER_USAGE_INDIRECT_BUFFER. The stride between commands must be at least sizeof(DrawIndirectCommand) (16 bytes) and a multiple of 4.

pub unsafe fn cmd_draw_indexed_indirect( &self, command_buffer: CommandBuffer, buffer: Buffer, offset: u64, draw_count: u32, stride: u32, )

Wraps vkCmdDrawIndexedIndirect. Provided by VK_GRAPHICS_VERSION_1_0.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdDrawIndexedIndirect was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Records one or more indexed draw calls whose parameters are read from a GPU buffer. Each DrawIndexedIndirectCommand contains index_count, instance_count, first_index, vertex_offset, and first_instance.

This is the indexed counterpart to cmd_draw_indirect and the most common indirect draw call for GPU-driven rendering pipelines. A compute shader performs culling and writes surviving draw commands into the buffer; the GPU then draws them without CPU involvement.

The buffer must have BUFFER_USAGE_INDIRECT_BUFFER. The stride must be at least sizeof(DrawIndexedIndirectCommand) (20 bytes) and a multiple of 4.

For dynamic draw counts, use cmd_draw_indexed_indirect_count (Vulkan 1.2).

pub unsafe fn cmd_dispatch( &self, command_buffer: CommandBuffer, group_count_x: u32, group_count_y: u32, group_count_z: u32, )

Wraps vkCmdDispatch. Provided by VK_COMPUTE_VERSION_1_0.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdDispatch was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Launches a compute shader with the given number of workgroups in X, Y, and Z dimensions. The total number of invocations is group_count * local_size (defined in the shader’s local_size_x, local_size_y, local_size_z).

A compute pipeline must be bound before calling this. Compute dispatches can be recorded outside a render pass.

Sizing: to process an image of width × height pixels with a local size of 16×16, dispatch ceil(width / 16) × ceil(height / 16) × 1 workgroups.

Limits: each dimension is capped by max_compute_work_group_count (at least 65535 per axis). The total invocations per workgroup are capped by max_compute_work_group_invocations (at least 128).

For GPU-driven dispatch counts, use cmd_dispatch_indirect.

pub unsafe fn cmd_dispatch_indirect( &self, command_buffer: CommandBuffer, buffer: Buffer, offset: u64, )

Wraps vkCmdDispatchIndirect. Provided by VK_COMPUTE_VERSION_1_0.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdDispatchIndirect was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Launches a compute shader with workgroup counts read from a GPU buffer. The buffer contains a DispatchIndirectCommand with group_count_x, group_count_y, group_count_z.

Use this when a prior compute pass determines how much work to do, for example, a culling pass writes the surviving workgroup count for a subsequent processing pass.

The buffer must have BUFFER_USAGE_INDIRECT_BUFFER. The offset must be a multiple of 4.

The same workgroup count limits apply as for cmd_dispatch.

pub unsafe fn cmd_subpass_shading_huawei(&self, command_buffer: CommandBuffer)

Wraps vkCmdSubpassShadingHUAWEI. Provided by VK_HUAWEI_subpass_shading.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdSubpassShadingHUAWEI was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Dispatches subpass shading work within the current subpass. Subpass shading runs compute-like shaders that have access to input attachments at the fragment’s location, combining the efficiency of compute with the data locality of subpasses.

Requires VK_HUAWEI_subpass_shading.

pub unsafe fn cmd_draw_cluster_huawei( &self, command_buffer: CommandBuffer, group_count_x: u32, group_count_y: u32, group_count_z: u32, )

Wraps vkCmdDrawClusterHUAWEI. Provided by VK_HUAWEI_cluster_culling_shader.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdDrawClusterHUAWEI was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Dispatches cluster culling shader workgroups using the Huawei cluster culling shader pipeline. The group counts specify the 3D dispatch dimensions, similar to cmd_dispatch.

Requires VK_HUAWEI_cluster_culling_shader.

pub unsafe fn cmd_draw_cluster_indirect_huawei( &self, command_buffer: CommandBuffer, buffer: Buffer, offset: u64, )

Wraps vkCmdDrawClusterIndirectHUAWEI. Provided by VK_HUAWEI_cluster_culling_shader.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdDrawClusterIndirectHUAWEI was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Indirect variant of cmd_draw_cluster_huawei. Reads the cluster dispatch parameters from a buffer at the given offset, allowing GPU-driven cluster culling without CPU readback.

Requires VK_HUAWEI_cluster_culling_shader.

pub unsafe fn cmd_update_pipeline_indirect_buffer_nv( &self, command_buffer: CommandBuffer, pipeline_bind_point: PipelineBindPoint, pipeline: Pipeline, )

Wraps vkCmdUpdatePipelineIndirectBufferNV. Provided by VK_NV_device_generated_commands_compute.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdUpdatePipelineIndirectBufferNV was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Updates the indirect dispatch parameters for a compute pipeline in a GPU buffer. Used with device-generated compute commands so the GPU can dispatch compute pipelines indirectly.

Requires VK_NV_device_generated_commands_compute.

pub unsafe fn cmd_copy_buffer( &self, command_buffer: CommandBuffer, src_buffer: Buffer, dst_buffer: Buffer, p_regions: &[BufferCopy], )

Wraps vkCmdCopyBuffer. Provided by VK_BASE_VERSION_1_0.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdCopyBuffer was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Copies data between two buffers. Multiple regions can be copied in a single call. Must be recorded outside a render pass.

Common patterns:

• Staging upload: copy from a host-visible staging buffer to a device-local buffer. This is the standard way to get vertex, index, and uniform data onto the GPU.
• Buffer-to-buffer transfers: defragment or reorganise GPU data.

Source and destination regions must not overlap within the same buffer. Use a temporary staging buffer if you need to shift data within a single buffer.

For Vulkan 1.3+, prefer cmd_copy_buffer2 which uses an extensible CopyBufferInfo2 struct.

pub unsafe fn cmd_copy_image( &self, command_buffer: CommandBuffer, src_image: Image, src_image_layout: ImageLayout, dst_image: Image, dst_image_layout: ImageLayout, p_regions: &[ImageCopy], )

Wraps vkCmdCopyImage. Provided by VK_BASE_VERSION_1_0.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdCopyImage was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Copies texel data between two images. Both images must have been created with TRANSFER_SRC and TRANSFER_DST usage respectively, and must be in compatible layouts (TRANSFER_SRC_OPTIMAL / TRANSFER_DST_OPTIMAL or GENERAL).

The source and destination formats must be identical, or both must be in the same size-compatibility class. For format conversion, use cmd_blit_image instead.

Copy operates on raw texel blocks, no filtering or scaling. The extent must be aligned to the texel block size for compressed formats.

Must be recorded outside a render pass. For Vulkan 1.3+, prefer cmd_copy_image2.

pub unsafe fn cmd_blit_image( &self, command_buffer: CommandBuffer, src_image: Image, src_image_layout: ImageLayout, dst_image: Image, dst_image_layout: ImageLayout, p_regions: &[ImageBlit], filter: Filter, )

Wraps vkCmdBlitImage. Provided by VK_GRAPHICS_VERSION_1_0.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdBlitImage was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Copies a region between two images with optional scaling and format conversion. Unlike cmd_copy_image, blit supports different source and destination extents (scaling) and applies a filter.

Filters:

• FILTER_NEAREST: no interpolation. Fast, but produces blocky results when scaling.
• FILTER_LINEAR: bilinear interpolation. Smooth scaling. Requires the format to support FORMAT_FEATURE_SAMPLED_IMAGE_FILTER_LINEAR.

Common uses:

• Mipmap generation: blit each mip level from the previous one, halving dimensions each step.
• Resolve or downscale: blit a high-resolution offscreen image to a smaller swapchain image.

Both images must be in appropriate transfer layouts. Must be recorded outside a render pass. For Vulkan 1.3+, prefer cmd_blit_image2.

Not supported for depth/stencil or compressed formats. Use cmd_copy_image or cmd_resolve_image for those.

pub unsafe fn cmd_copy_buffer_to_image( &self, command_buffer: CommandBuffer, src_buffer: Buffer, dst_image: Image, dst_image_layout: ImageLayout, p_regions: &[BufferImageCopy], )

Wraps vkCmdCopyBufferToImage. Provided by VK_BASE_VERSION_1_0.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdCopyBufferToImage was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Copies data from a buffer to an image, the primary way to upload texture data from CPU to GPU.

Typical upload workflow:

1. Write pixel data into a host-visible staging buffer.
2. Transition the target image to TRANSFER_DST_OPTIMAL.
3. cmd_copy_buffer_to_image with the appropriate BufferImageCopy regions.
4. Transition the image to SHADER_READ_ONLY_OPTIMAL for sampling.

Buffer layout: buffer_row_length and buffer_image_height control the row and slice pitch of the source data in the buffer. Set both to zero to use a tightly packed layout matching the image extent.

Multiple regions can be copied in a single call (e.g. all mip levels of a texture). Must be recorded outside a render pass.

For Vulkan 1.3+, prefer cmd_copy_buffer_to_image2.

pub unsafe fn cmd_copy_image_to_buffer( &self, command_buffer: CommandBuffer, src_image: Image, src_image_layout: ImageLayout, dst_buffer: Buffer, p_regions: &[BufferImageCopy], )

Wraps vkCmdCopyImageToBuffer. Provided by VK_BASE_VERSION_1_0.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdCopyImageToBuffer was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Copies data from an image to a buffer, used for GPU readback of rendered images, screenshots, or compute shader output.

Typical readback workflow:

1. Transition the source image to TRANSFER_SRC_OPTIMAL.
2. cmd_copy_image_to_buffer into a host-visible buffer.
3. Submit and wait for the fence.
4. Map the buffer and read the pixel data on the CPU.

The buffer_row_length and buffer_image_height fields in BufferImageCopy control the destination layout. Set both to zero for tightly packed output.

Be aware that readback is not instantaneous, it requires a full GPU round-trip. Avoid reading back in the render loop unless you are double- or triple-buffering the readback to hide latency.

Must be recorded outside a render pass. For Vulkan 1.3+, prefer cmd_copy_image_to_buffer2.

pub unsafe fn cmd_copy_memory_indirect_nv( &self, command_buffer: CommandBuffer, copy_buffer_address: u64, copy_count: u32, stride: u32, )

Wraps vkCmdCopyMemoryIndirectNV. Provided by VK_NV_copy_memory_indirect.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdCopyMemoryIndirectNV was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Copies memory regions using indirect parameters stored in a GPU buffer at the given device address. Enables GPU-driven memory copy workflows where the copy descriptors are generated on the device.

Requires VK_NV_copy_memory_indirect.

pub unsafe fn cmd_copy_memory_indirect_khr( &self, command_buffer: CommandBuffer, p_copy_memory_indirect_info: &CopyMemoryIndirectInfoKHR, )

Wraps vkCmdCopyMemoryIndirectKHR. Provided by VK_KHR_copy_memory_indirect.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdCopyMemoryIndirectKHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Copies memory regions using parameters read from a device-side info structure. Enables GPU-driven memory copies without CPU involvement in specifying source/destination addresses.

Requires VK_KHR_copy_memory_indirect.

pub unsafe fn cmd_copy_memory_to_image_indirect_nv( &self, command_buffer: CommandBuffer, copy_buffer_address: u64, stride: u32, dst_image: Image, dst_image_layout: ImageLayout, p_image_subresources: &[ImageSubresourceLayers], )

Wraps vkCmdCopyMemoryToImageIndirectNV. Provided by VK_NV_copy_memory_indirect.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdCopyMemoryToImageIndirectNV was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Copies data from memory to an image using indirect parameters stored at a device address. Enables GPU-driven texture uploads where the copy regions are generated on the device.

Requires VK_NV_copy_memory_indirect.

pub unsafe fn cmd_copy_memory_to_image_indirect_khr( &self, command_buffer: CommandBuffer, p_copy_memory_to_image_indirect_info: &CopyMemoryToImageIndirectInfoKHR, )

Wraps vkCmdCopyMemoryToImageIndirectKHR. Provided by VK_KHR_copy_memory_indirect.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdCopyMemoryToImageIndirectKHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Copies data from memory to an image using parameters read from a device-side info structure. Enables GPU-driven memory-to-image transfers without CPU readback of copy parameters.

Requires VK_KHR_copy_memory_indirect.

pub unsafe fn cmd_update_buffer( &self, command_buffer: CommandBuffer, dst_buffer: Buffer, dst_offset: u64, data_size: u64, p_data: *const c_void, )

Wraps vkCmdUpdateBuffer. Provided by VK_BASE_VERSION_1_0.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdUpdateBuffer was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Writes a small amount of data inline into a buffer from the command stream. The data is embedded directly in the command buffer, no staging buffer needed.

Size limit: the data size must be ≤ 65536 bytes and a multiple of 4. For larger uploads, use cmd_copy_buffer with a staging buffer instead.

This is convenient for small per-frame updates (e.g. a uniform buffer with a single matrix) but should not be used for bulk data, it inflates command buffer size and the data is uploaded through the command stream, which is slower than a DMA transfer.

Must be recorded outside a render pass. The destination buffer must have BUFFER_USAGE_TRANSFER_DST.

pub unsafe fn cmd_fill_buffer( &self, command_buffer: CommandBuffer, dst_buffer: Buffer, dst_offset: u64, size: u64, data: u32, )

Wraps vkCmdFillBuffer. Provided by VK_BASE_VERSION_1_0.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdFillBuffer was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Fills a region of a buffer with a repeating 4-byte value. Useful for clearing GPU buffers to zero (or any uniform value) without a staging upload.

Common uses:

• Zero-initialise an indirect draw count buffer before a compute culling pass.
• Clear a storage buffer used as a histogram or counter.

The offset and size must be multiples of 4. Use VK_WHOLE_SIZE to fill from the offset to the end of the buffer.

Must be recorded outside a render pass. The buffer must have BUFFER_USAGE_TRANSFER_DST.

pub unsafe fn cmd_clear_color_image( &self, command_buffer: CommandBuffer, image: Image, image_layout: ImageLayout, p_color: &ClearColorValue, p_ranges: &[ImageSubresourceRange], )

Wraps vkCmdClearColorImage. Provided by VK_COMPUTE_VERSION_1_0.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdClearColorImage was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Clears one or more regions of a colour image to a specified value. The image must be in TRANSFER_DST_OPTIMAL or GENERAL layout.

This is an explicit clear outside a render pass. For clears inside a render pass, use load_op = CLEAR in the attachment description or cmd_clear_attachments, both are typically faster because the driver can integrate them with tile-based rendering.

The clear value is a ClearColorValue union: either four float32, int32, or uint32 values depending on the image format.

Must be recorded outside a render pass.

pub unsafe fn cmd_clear_depth_stencil_image( &self, command_buffer: CommandBuffer, image: Image, image_layout: ImageLayout, p_depth_stencil: &ClearDepthStencilValue, p_ranges: &[ImageSubresourceRange], )

Wraps vkCmdClearDepthStencilImage. Provided by VK_GRAPHICS_VERSION_1_0.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdClearDepthStencilImage was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Clears one or more regions of a depth/stencil image to a specified depth and stencil value. The image must be in TRANSFER_DST_OPTIMAL or GENERAL layout.

For most rendering, clearing via load_op = CLEAR on the depth/stencil attachment is preferred, it lets tile-based GPUs avoid a separate clear pass. Use this command only when you need to clear a depth/stencil image outside a render pass.

The image_subresource_range must reference the appropriate aspect flags (DEPTH, STENCIL, or both).

Must be recorded outside a render pass.

pub unsafe fn cmd_clear_attachments( &self, command_buffer: CommandBuffer, p_attachments: &[ClearAttachment], p_rects: &[ClearRect], )

Wraps vkCmdClearAttachments. Provided by VK_GRAPHICS_VERSION_1_0.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdClearAttachments was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Clears one or more attachment regions inside an active render pass. Unlike load_op = CLEAR (which clears the entire attachment at render pass begin), this clears arbitrary rectangular regions mid-render-pass.

Use cases:

• Clear a sub-region of a colour attachment (e.g. a UI panel background).
• Clear the stencil buffer for a specific screen region.

Each ClearAttachment specifies which attachment to clear (colour index, depth, or stencil) and the clear value. Each ClearRect defines the pixel rectangle and layer range.

For whole-attachment clears, prefer load_op = CLEAR, it is always at least as fast and often faster on tile-based hardware.

pub unsafe fn cmd_resolve_image( &self, command_buffer: CommandBuffer, src_image: Image, src_image_layout: ImageLayout, dst_image: Image, dst_image_layout: ImageLayout, p_regions: &[ImageResolve], )

Wraps vkCmdResolveImage. Provided by VK_GRAPHICS_VERSION_1_0.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdResolveImage was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Resolves (downsamples) a multisample image into a single-sample image. Typically used to produce the final single-sample result from a multisampled colour attachment.

Both images must be in appropriate transfer layouts (TRANSFER_SRC_OPTIMAL and TRANSFER_DST_OPTIMAL respectively). The source must be multisampled; the destination must be single-sample. Formats must be identical.

For resolving inside a render pass, use resolve_attachment in the subpass description instead, it is more efficient on tile-based GPUs because the resolve happens in-tile.

Must be recorded outside a render pass. For Vulkan 1.3+, prefer cmd_resolve_image2.

pub unsafe fn cmd_set_event( &self, command_buffer: CommandBuffer, event: Event, stage_mask: PipelineStageFlags, )

Wraps vkCmdSetEvent. Provided by VK_COMPUTE_VERSION_1_0.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdSetEvent was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Signals an event from the GPU at a specific pipeline stage. A later cmd_wait_events call can wait for this signal to synchronise work within the same queue.

Events provide finer-grained synchronisation than pipeline barriers when you want to split a dependency into a “signal” point and a “wait” point separated by other commands. This lets the GPU execute interleaving work between the signal and wait.

The stage_mask specifies at which pipeline stage the event is signaled. The event becomes signaled once all commands prior to this call have completed that stage.

Events must only be used within a single queue. For cross-queue synchronisation, use semaphores.

For Vulkan 1.3+, prefer cmd_set_event2 which supports more precise stage and access masks.

pub unsafe fn cmd_reset_event( &self, command_buffer: CommandBuffer, event: Event, stage_mask: PipelineStageFlags, )

Wraps vkCmdResetEvent. Provided by VK_COMPUTE_VERSION_1_0.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdResetEvent was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Resets an event to the unsignaled state from the GPU at a specific pipeline stage. The event can then be signaled again by a subsequent cmd_set_event.

Must not be called while a cmd_wait_events that waits on this event is between its wait and the completion of the dependent work.

For Vulkan 1.3+, prefer cmd_reset_event2.

pub unsafe fn cmd_wait_events( &self, command_buffer: CommandBuffer, p_events: &[Event], src_stage_mask: PipelineStageFlags, dst_stage_mask: PipelineStageFlags, p_memory_barriers: &[MemoryBarrier], p_buffer_memory_barriers: &[BufferMemoryBarrier], p_image_memory_barriers: &[ImageMemoryBarrier], )

Wraps vkCmdWaitEvents. Provided by VK_COMPUTE_VERSION_1_0.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdWaitEvents was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Waits for one or more events to be signaled and then inserts memory and execution dependencies. This is the “wait” half of the signal/wait pattern started by cmd_set_event.

The wait blocks execution at the specified destination pipeline stages until all events are signaled. Memory barriers provided in this call make the specified source writes visible to the destination stages.

Split barriers: the main advantage over cmd_pipeline_barrier is that you can interleave unrelated commands between the signal and wait, giving the GPU more opportunity for parallel execution.

Events must not be waited on across different queues. For cross-queue synchronisation, use semaphores.

For Vulkan 1.3+, prefer cmd_wait_events2.

pub unsafe fn cmd_pipeline_barrier( &self, command_buffer: CommandBuffer, src_stage_mask: PipelineStageFlags, dst_stage_mask: PipelineStageFlags, dependency_flags: DependencyFlags, p_memory_barriers: &[MemoryBarrier], p_buffer_memory_barriers: &[BufferMemoryBarrier], p_image_memory_barriers: &[ImageMemoryBarrier], )

Wraps vkCmdPipelineBarrier. Provided by VK_BASE_VERSION_1_0.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdPipelineBarrier was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Inserts an execution and memory dependency between commands recorded before and after the barrier. This is the primary synchronisation tool inside a command buffer.

Three types of barrier:

• Memory barriers (MemoryBarrier): global, affects all resources. Rarely needed, prefer the more specific variants.
• Buffer memory barriers (BufferMemoryBarrier): targets a specific buffer region. Use for storage buffer read-after-write.
• Image memory barriers (ImageMemoryBarrier): targets a specific image subresource range. Also performs layout transitions.

Layout transitions: image memory barriers are the primary way to transition images between layouts (e.g. TRANSFER_DST_OPTIMAL → SHADER_READ_ONLY_OPTIMAL). The old and new layouts in the barrier define the transition.

Stage masks: src_stage_mask is the set of stages that must complete before the barrier. dst_stage_mask is the set of stages that must wait for the barrier. Choose the narrowest stages possible to minimise stalls.

Inside a render pass, only self-dependencies are allowed (barriers within a single subpass). Outside a render pass, there are no restrictions.

For Vulkan 1.3+, prefer cmd_pipeline_barrier2 which uses extensible structs.

pub unsafe fn cmd_begin_query( &self, command_buffer: CommandBuffer, query_pool: QueryPool, query: u32, flags: QueryControlFlags, )

Wraps vkCmdBeginQuery. Provided by VK_BASE_VERSION_1_0.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdBeginQuery was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Begins a query at the specified index in a query pool. All rendering or compute commands recorded between cmd_begin_query and cmd_end_query are measured by the query.

Flags:

• QUERY_CONTROL_PRECISE: for occlusion queries, return an exact sample count instead of a boolean. More expensive on some hardware. Requires the occlusion_query_precise device feature.

The query slot must have been reset with cmd_reset_query_pool (or reset_query_pool on Vulkan 1.2+) before beginning.

Pipeline statistics queries must be begun and ended outside a render pass. Occlusion queries can span draw calls within a render pass.

pub unsafe fn cmd_end_query( &self, command_buffer: CommandBuffer, query_pool: QueryPool, query: u32, )

Wraps vkCmdEndQuery. Provided by VK_BASE_VERSION_1_0.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdEndQuery was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Ends an active query at the specified index. The query results become available for retrieval via get_query_pool_results or cmd_copy_query_pool_results once the command buffer has completed execution.

Must be paired with a preceding cmd_begin_query on the same query index. Beginning a query without ending it, or ending one that was not begun, is an error.

pub unsafe fn cmd_begin_conditional_rendering_ext( &self, command_buffer: CommandBuffer, p_conditional_rendering_begin: &ConditionalRenderingBeginInfoEXT, )

Wraps vkCmdBeginConditionalRenderingEXT. Provided by VK_EXT_conditional_rendering.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdBeginConditionalRenderingEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Begins a conditional rendering block. Subsequent rendering and dispatch commands are discarded if the 32-bit value at the specified buffer offset is zero (or non-zero if INVERTED is set).

End with cmd_end_conditional_rendering_ext.

Useful for GPU-driven occlusion culling, write visibility results to a buffer, then conditionally skip draw calls.

Requires VK_EXT_conditional_rendering.

pub unsafe fn cmd_end_conditional_rendering_ext( &self, command_buffer: CommandBuffer, )

Wraps vkCmdEndConditionalRenderingEXT. Provided by VK_EXT_conditional_rendering.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdEndConditionalRenderingEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Ends a conditional rendering block started with cmd_begin_conditional_rendering_ext. Commands after this call execute unconditionally.

Requires VK_EXT_conditional_rendering.

pub unsafe fn cmd_begin_custom_resolve_ext( &self, command_buffer: CommandBuffer, p_begin_custom_resolve_info: Option<&BeginCustomResolveInfoEXT>, )

Wraps vkCmdBeginCustomResolveEXT. Provided by VK_EXT_custom_resolve.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdBeginCustomResolveEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Begins a custom resolve region, allowing the application to use its own fragment shader for MSAA resolve instead of the fixed- function resolve. End with cmd_end_custom_resolve_ext.

Useful for tone-mapped or weighted resolves that the built-in resolve operations cannot express.

Requires VK_EXT_custom_resolve.

pub unsafe fn cmd_reset_query_pool( &self, command_buffer: CommandBuffer, query_pool: QueryPool, first_query: u32, query_count: u32, )

Wraps vkCmdResetQueryPool. Provided by VK_BASE_VERSION_1_0.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdResetQueryPool was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Resets a range of queries in a pool from the GPU command stream. Queries must be reset before they can be used in cmd_begin_query or cmd_write_timestamp.

This is the pre-1.2 way to reset queries. For Vulkan 1.2+, reset_query_pool (host-side) is often more convenient and avoids adding the reset to the command buffer.

Must be recorded outside a render pass.

pub unsafe fn cmd_write_timestamp( &self, command_buffer: CommandBuffer, pipeline_stage: PipelineStageFlagBits, query_pool: QueryPool, query: u32, )

Wraps vkCmdWriteTimestamp. Provided by VK_BASE_VERSION_1_0.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdWriteTimestamp was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Writes the current GPU timestamp into a query pool slot when the specified pipeline stage completes. Use two timestamps to measure elapsed GPU time:

cmd_write_timestamp(PIPELINE_STAGE_TOP_OF_PIPE, pool, 0);
// ... commands to measure ...
cmd_write_timestamp(PIPELINE_STAGE_BOTTOM_OF_PIPE, pool, 1);

After the command buffer completes, read the values with get_query_pool_results (with QUERY_RESULT_64) and compute:

elapsed_ns = (timestamp[1] - timestamp[0]) * timestamp_period

timestamp_period is in nanoseconds per tick, available from physical_device_limits.

Not all queue families support timestamps, check timestamp_valid_bits in the queue family properties. A value of 0 means timestamps are not supported on that queue.

For Vulkan 1.3+, prefer cmd_write_timestamp2.

pub unsafe fn cmd_copy_query_pool_results( &self, command_buffer: CommandBuffer, query_pool: QueryPool, first_query: u32, query_count: u32, dst_buffer: Buffer, dst_offset: u64, stride: u64, flags: QueryResultFlags, )

Wraps vkCmdCopyQueryPoolResults. Provided by VK_BASE_VERSION_1_0.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdCopyQueryPoolResults was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Copies query results directly into a GPU buffer. This is the GPU-side counterpart to get_query_pool_results and avoids a CPU round-trip when the results are consumed by subsequent GPU work (e.g. conditional rendering or indirect dispatch).

The same flags apply as for get_query_pool_results: QUERY_RESULT_64, QUERY_RESULT_WAIT, QUERY_RESULT_WITH_AVAILABILITY, and QUERY_RESULT_PARTIAL.

The destination buffer must have BUFFER_USAGE_TRANSFER_DST. The stride must be large enough to hold the result (and availability value, if requested).

Must be recorded outside a render pass.

pub unsafe fn cmd_push_constants( &self, command_buffer: CommandBuffer, layout: PipelineLayout, stage_flags: ShaderStageFlags, offset: u32, p_values: &[c_void], )

Wraps vkCmdPushConstants. Provided by VK_COMPUTE_VERSION_1_0.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdPushConstants was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Updates push constant values for the bound pipeline layout. Push constants are a fast path for small, frequently-changing data that avoids descriptor set updates entirely.

Size limit: the total push constant range is at least 128 bytes (device limit max_push_constants_size). Use push constants for per-draw data like transform matrices, material indices, or time values.

Stage flags: the stage_flags parameter must match the stages declared in the pipeline layout’s push constant range. You can update different stage ranges separately (e.g. update the vertex shader’s range without touching the fragment shader’s range).

Push constant data persists across draw/dispatch calls until the pipeline layout is changed or the values are overwritten.

For Vulkan 1.4+, cmd_push_constants2 uses an extensible struct.

pub unsafe fn cmd_begin_render_pass( &self, command_buffer: CommandBuffer, p_render_pass_begin: &RenderPassBeginInfo, contents: SubpassContents, )

Wraps vkCmdBeginRenderPass. Provided by VK_GRAPHICS_VERSION_1_0.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdBeginRenderPass was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Begins a render pass instance. All subsequent drawing commands are recorded within this render pass until cmd_end_render_pass.

render_pass_begin_info specifies:

• Render pass and framebuffer: which render pass to use and which concrete image views are bound.
• Render area: the pixel region to render. Should match the framebuffer extent for best performance. Misalignment with the render area granularity can cause overhead on tile-based GPUs.
• Clear values: one per attachment with load_op = CLEAR. The array must include entries for all attachments (use a dummy value for non-cleared attachments).

contents:

• SUBPASS_CONTENTS_INLINE: draw commands are recorded directly in this command buffer.
• SUBPASS_CONTENTS_SECONDARY_COMMAND_BUFFERS: draw commands come from secondary command buffers via cmd_execute_commands.

For Vulkan 1.2+, cmd_begin_render_pass2 accepts a SubpassBeginInfo. For Vulkan 1.3+, consider dynamic rendering (cmd_begin_rendering) which avoids render pass and framebuffer objects entirely.

Guide

See Render Passes & Framebuffers in the vulkan_rust guide.

pub unsafe fn cmd_next_subpass( &self, command_buffer: CommandBuffer, contents: SubpassContents, )

Wraps vkCmdNextSubpass. Provided by VK_GRAPHICS_VERSION_1_0.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdNextSubpass was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Advances to the next subpass within a render pass. Subpass transitions allow the driver to resolve dependencies between subpasses, for example, reading a colour attachment written in the previous subpass as an input attachment.

The contents parameter has the same meaning as in cmd_begin_render_pass: INLINE or SECONDARY_COMMAND_BUFFERS.

Multi-subpass render passes are an optimisation for tile-based GPUs where they can keep data on-chip between subpasses. On desktop GPUs the benefit is smaller. Many applications use a single subpass and handle inter-pass dependencies with explicit pipeline barriers.

For Vulkan 1.2+, prefer cmd_next_subpass2.

pub unsafe fn cmd_end_render_pass(&self, command_buffer: CommandBuffer)

Wraps vkCmdEndRenderPass. Provided by VK_GRAPHICS_VERSION_1_0.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdEndRenderPass was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Ends the current render pass instance. After this call, the implicit layout transitions specified by each attachment’s final_layout are applied.

No draw commands may be recorded after this until a new render pass is begun (or dynamic rendering is started).

For Vulkan 1.2+, prefer cmd_end_render_pass2.

pub unsafe fn cmd_execute_commands( &self, command_buffer: CommandBuffer, p_command_buffers: &[CommandBuffer], )

Wraps vkCmdExecuteCommands. Provided by VK_BASE_VERSION_1_0.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdExecuteCommands was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Executes one or more secondary command buffers from a primary command buffer. The secondary buffers are inlined into the primary buffer’s execution stream in array order.

Use cases:

• Multi-threaded recording: each thread records a secondary command buffer, and the main thread assembles them with a single cmd_execute_commands call. This is the primary scaling strategy for CPU-bound recording.
• Reusable draw sequences: record a secondary buffer once and execute it in multiple frames or from multiple primary buffers (requires SIMULTANEOUS_USE on the secondary buffer).

Secondary command buffers inherit certain state from the primary buffer (viewport, scissor, etc.) only if declared in the CommandBufferInheritanceInfo. The render pass and subpass must match what the primary buffer is currently in.

This command can only be called from a primary command buffer.

pub unsafe fn create_shared_swapchains_khr( &self, p_create_infos: &[SwapchainCreateInfoKHR], allocator: Option<&AllocationCallbacks>, ) -> VkResult<Vec<SwapchainKHR>>

Wraps vkCreateSharedSwapchainsKHR. Provided by VK_KHR_display_swapchain.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_INCOMPATIBLE_DISPLAY_KHR
• VK_ERROR_DEVICE_LOST
• VK_ERROR_SURFACE_LOST_KHR
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkCreateSharedSwapchainsKHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Creates one or more swapchains that share presentable images with their old_swapchain. Provided by VK_KHR_display_swapchain.

This is primarily used for direct-to-display rendering where multiple swapchains share the same display plane. For window-based rendering, use create_swapchain_khr instead.

pub unsafe fn create_swapchain_khr( &self, p_create_info: &SwapchainCreateInfoKHR, allocator: Option<&AllocationCallbacks>, ) -> VkResult<SwapchainKHR>

Wraps vkCreateSwapchainKHR. Provided by VK_KHR_swapchain.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_DEVICE_LOST
• VK_ERROR_SURFACE_LOST_KHR
• VK_ERROR_NATIVE_WINDOW_IN_USE_KHR
• VK_ERROR_INITIALIZATION_FAILED
• VK_ERROR_COMPRESSION_EXHAUSTED_EXT
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkCreateSwapchainKHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Creates a swapchain, a queue of presentable images tied to a surface (window). The swapchain is the bridge between rendering and display.

Key parameters:

• min_image_count: request at least this many images. For double buffering use 2, for triple buffering use 3. Query get_physical_device_surface_capabilities_khr for the supported range.
• image_format / image_color_space: pick a pair from get_physical_device_surface_formats_khr. B8G8R8A8_SRGB + SRGB_NONLINEAR is the most portable.
• present_mode: FIFO (vsync, always supported), MAILBOX (low-latency triple buffering), IMMEDIATE (no vsync, tearing).
• pre_transform: set to current_transform from surface capabilities to avoid an extra composition blit.
• old_swapchain: when recreating after a resize, pass the old swapchain here. The driver can reuse internal resources.

Swapchain recreation is required when the surface size changes (window resize) or when acquire_next_image_khr returns VK_ERROR_OUT_OF_DATE_KHR. Destroy the old swapchain after creating the new one.

Guide

See Hello Triangle, Part 2 in the vulkan_rust guide.

pub unsafe fn destroy_swapchain_khr( &self, swapchain: SwapchainKHR, allocator: Option<&AllocationCallbacks>, )

Wraps vkDestroySwapchainKHR. Provided by VK_KHR_swapchain.

Safety

• device (self) must be valid and not destroyed.
• swapchain must be externally synchronized.

Panics

Panics if vkDestroySwapchainKHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Destroys a swapchain. All images obtained from get_swapchain_images_khr become invalid, destroy any image views and framebuffers referencing them first.

Wait for all rendering to complete (device_wait_idle) before destroying. Do not destroy a swapchain while an acquired image has not been presented.

When recreating a swapchain (e.g. on resize), create the new one first (passing the old as old_swapchain), then destroy the old one.

Guide

See Hello Triangle, Part 2 in the vulkan_rust guide.

pub unsafe fn get_swapchain_images_khr( &self, swapchain: SwapchainKHR, ) -> VkResult<Vec<Image>>

Wraps vkGetSwapchainImagesKHR. Provided by VK_KHR_swapchain.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetSwapchainImagesKHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Returns the array of presentable images owned by the swapchain. You do not create or destroy these images, they are managed by the swapchain.

The returned image count may be greater than min_image_count requested at swapchain creation.

Create an ImageView for each swapchain image to use them as render targets. These views (and any framebuffers using them) must be destroyed before the swapchain is destroyed.

The images start in an undefined layout. Transition them to the appropriate layout (e.g. COLOR_ATTACHMENT_OPTIMAL) during the first render pass or via a pipeline barrier.

Guide

See Hello Triangle, Part 2 in the vulkan_rust guide.

pub unsafe fn acquire_next_image_khr( &self, swapchain: SwapchainKHR, timeout: u64, semaphore: Semaphore, fence: Fence, ) -> VkResult<u32>

Wraps vkAcquireNextImageKHR. Provided by VK_KHR_swapchain.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_DEVICE_LOST
• VK_ERROR_OUT_OF_DATE_KHR
• VK_ERROR_SURFACE_LOST_KHR
• VK_ERROR_FULL_SCREEN_EXCLUSIVE_MODE_LOST_EXT
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.
• swapchain must be externally synchronized.
• semaphore must be externally synchronized.
• fence must be externally synchronized.

Panics

Panics if vkAcquireNextImageKHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Acquires the next available image from the swapchain for rendering. Returns the index into the array from get_swapchain_images_khr.

Synchronisation: pass a semaphore, a fence, or both. The semaphore/fence is signaled when the image is ready to be rendered to. Do not start rendering until the semaphore is waited on in queue_submit.

Timeout: in nanoseconds. u64::MAX blocks indefinitely.

Special return values:

• VK_SUBOPTIMAL_KHR: the swapchain still works but no longer matches the surface perfectly (e.g. after a resize). You can continue rendering but should recreate the swapchain soon.
• VK_ERROR_OUT_OF_DATE_KHR: the swapchain is incompatible with the surface and must be recreated before rendering. Do not present the acquired image.

A common frame loop:

acquire_next_image_khr(swapchain, u64::MAX, image_available_sem, null)
// wait on image_available_sem in queue_submit
// render to swapchain_images[index]
// signal render_finished_sem in queue_submit
queue_present_khr(render_finished_sem, swapchain, index)

Guide

See Synchronization in the vulkan_rust guide.

pub unsafe fn queue_present_khr( &self, queue: Queue, p_present_info: &PresentInfoKHR, ) -> VkResult<()>

Wraps vkQueuePresentKHR. Provided by VK_KHR_swapchain.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_DEVICE_LOST
• VK_ERROR_OUT_OF_DATE_KHR
• VK_ERROR_SURFACE_LOST_KHR
• VK_ERROR_FULL_SCREEN_EXCLUSIVE_MODE_LOST_EXT
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED
• VK_ERROR_PRESENT_TIMING_QUEUE_FULL_EXT

Safety

• queue (self) must be valid and not destroyed.
• queue must be externally synchronized.

Panics

Panics if vkQueuePresentKHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Presents a rendered swapchain image to the display. This is the final step in the frame loop, after rendering is complete, present the image to make it visible.

Wait semaphores: the present waits on these semaphores before presenting. Pass the semaphore that your render submission signals to ensure the image is fully rendered before it goes to the display.

Multiple swapchains: a single present call can present to multiple swapchains simultaneously (e.g. for multi-window or multi-monitor rendering).

Return values:

• VK_SUBOPTIMAL_KHR: presented successfully but the swapchain should be recreated.
• VK_ERROR_OUT_OF_DATE_KHR: presentation failed, the swapchain must be recreated.

The present queue does not need to be the same as the graphics queue, but the semaphore synchronisation must be correct if they differ.

Guide

See Hello Triangle, Part 4 in the vulkan_rust guide.

pub unsafe fn debug_marker_set_object_name_ext( &self, p_name_info: &DebugMarkerObjectNameInfoEXT, ) -> VkResult<()>

Wraps vkDebugMarkerSetObjectNameEXT. Provided by VK_EXT_debug_marker.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkDebugMarkerSetObjectNameEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Assigns a name to a Vulkan object for debugging. This is the legacy VK_EXT_debug_marker equivalent of set_debug_utils_object_name_ext.

DebugMarkerObjectNameInfoEXT uses the old DebugReportObjectTypeEXT enum to identify object types.

Superseded by VK_EXT_debug_utils. Prefer set_debug_utils_object_name_ext for new code.

pub unsafe fn debug_marker_set_object_tag_ext( &self, p_tag_info: &DebugMarkerObjectTagInfoEXT, ) -> VkResult<()>

Wraps vkDebugMarkerSetObjectTagEXT. Provided by VK_EXT_debug_marker.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkDebugMarkerSetObjectTagEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Attaches arbitrary binary data to a Vulkan object. This is the legacy VK_EXT_debug_marker equivalent of set_debug_utils_object_tag_ext.

Superseded by VK_EXT_debug_utils. Prefer set_debug_utils_object_tag_ext for new code.

pub unsafe fn cmd_debug_marker_begin_ext( &self, command_buffer: CommandBuffer, p_marker_info: &DebugMarkerMarkerInfoEXT, )

Wraps vkCmdDebugMarkerBeginEXT. Provided by VK_EXT_debug_marker.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdDebugMarkerBeginEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Opens a debug marker region in a command buffer. This is the legacy VK_EXT_debug_marker equivalent of cmd_begin_debug_utils_label_ext.

DebugMarkerMarkerInfoEXT specifies a name and optional RGBA color. Close with cmd_debug_marker_end_ext.

Superseded by VK_EXT_debug_utils. Prefer cmd_begin_debug_utils_label_ext for new code.

pub unsafe fn cmd_debug_marker_end_ext(&self, command_buffer: CommandBuffer)

Wraps vkCmdDebugMarkerEndEXT. Provided by VK_EXT_debug_marker.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdDebugMarkerEndEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Closes the most recently opened debug marker region in the command buffer. Must be paired with cmd_debug_marker_begin_ext.

This is the legacy VK_EXT_debug_marker equivalent of cmd_end_debug_utils_label_ext. Prefer VK_EXT_debug_utils for new code.

pub unsafe fn cmd_debug_marker_insert_ext( &self, command_buffer: CommandBuffer, p_marker_info: &DebugMarkerMarkerInfoEXT, )

Wraps vkCmdDebugMarkerInsertEXT. Provided by VK_EXT_debug_marker.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdDebugMarkerInsertEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Inserts a single-point debug marker into the command buffer. This is the legacy VK_EXT_debug_marker equivalent of cmd_insert_debug_utils_label_ext.

DebugMarkerMarkerInfoEXT specifies a name and optional RGBA color.

Superseded by VK_EXT_debug_utils. Prefer cmd_insert_debug_utils_label_ext for new code.

pub unsafe fn get_memory_win32_handle_nv( &self, memory: DeviceMemory, handle_type: ExternalMemoryHandleTypeFlagsNV, ) -> VkResult<isize>

Wraps vkGetMemoryWin32HandleNV. Provided by VK_NV_external_memory_win32.

Errors

• VK_ERROR_TOO_MANY_OBJECTS
• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetMemoryWin32HandleNV was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Exports a Vulkan device memory allocation as a Win32 handle (HANDLE) for sharing with other APIs or processes. This is the legacy NV path; prefer get_memory_win32_handle_khr for new code.

Requires VK_NV_external_memory_win32. Windows only.

pub unsafe fn cmd_execute_generated_commands_nv( &self, command_buffer: CommandBuffer, is_preprocessed: bool, p_generated_commands_info: &GeneratedCommandsInfoNV, )

Wraps vkCmdExecuteGeneratedCommandsNV. Provided by VK_NV_device_generated_commands.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdExecuteGeneratedCommandsNV was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Executes commands that were generated on the GPU. If is_preprocessed is set, the commands must have been preprocessed with cmd_preprocess_generated_commands_nv first.

Requires VK_NV_device_generated_commands.

pub unsafe fn cmd_preprocess_generated_commands_nv( &self, command_buffer: CommandBuffer, p_generated_commands_info: &GeneratedCommandsInfoNV, )

Wraps vkCmdPreprocessGeneratedCommandsNV. Provided by VK_NV_device_generated_commands.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdPreprocessGeneratedCommandsNV was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Preprocesses device-generated commands into a form suitable for fast execution. The preprocessing result is stored in a preprocess buffer and later consumed by cmd_execute_generated_commands_nv with is_preprocessed set.

Requires VK_NV_device_generated_commands.

pub unsafe fn cmd_bind_pipeline_shader_group_nv( &self, command_buffer: CommandBuffer, pipeline_bind_point: PipelineBindPoint, pipeline: Pipeline, group_index: u32, )

Wraps vkCmdBindPipelineShaderGroupNV. Provided by VK_NV_device_generated_commands.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdBindPipelineShaderGroupNV was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Binds a shader group from a pipeline that was created with multiple shader groups. Used with device-generated commands to switch between pre-compiled shader variants without rebinding the entire pipeline.

Requires VK_NV_device_generated_commands.

pub unsafe fn get_generated_commands_memory_requirements_nv( &self, p_info: &GeneratedCommandsMemoryRequirementsInfoNV, p_memory_requirements: &mut MemoryRequirements2, )

Wraps vkGetGeneratedCommandsMemoryRequirementsNV. Provided by VK_NV_device_generated_commands.

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetGeneratedCommandsMemoryRequirementsNV was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Queries the memory requirements for preprocessing device-generated commands. The returned size determines how large the preprocess buffer must be for cmd_preprocess_generated_commands_nv.

Requires VK_NV_device_generated_commands.

pub unsafe fn create_indirect_commands_layout_nv( &self, p_create_info: &IndirectCommandsLayoutCreateInfoNV, allocator: Option<&AllocationCallbacks>, ) -> VkResult<IndirectCommandsLayoutNV>

Wraps vkCreateIndirectCommandsLayoutNV. Provided by VK_NV_device_generated_commands.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkCreateIndirectCommandsLayoutNV was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Creates a layout that describes the structure of indirect command sequences for device-generated commands. The layout defines which tokens (draw, dispatch, push constants, etc.) appear in the command stream and their order.

Destroy with destroy_indirect_commands_layout_nv.

Requires VK_NV_device_generated_commands.

pub unsafe fn destroy_indirect_commands_layout_nv( &self, indirect_commands_layout: IndirectCommandsLayoutNV, allocator: Option<&AllocationCallbacks>, )

Wraps vkDestroyIndirectCommandsLayoutNV. Provided by VK_NV_device_generated_commands.

Safety

• device (self) must be valid and not destroyed.
• indirectCommandsLayout must be externally synchronized.

Panics

Panics if vkDestroyIndirectCommandsLayoutNV was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Destroys an indirect commands layout created with create_indirect_commands_layout_nv.

Requires VK_NV_device_generated_commands.

pub unsafe fn cmd_execute_generated_commands_ext( &self, command_buffer: CommandBuffer, is_preprocessed: bool, p_generated_commands_info: &GeneratedCommandsInfoEXT, )

Wraps vkCmdExecuteGeneratedCommandsEXT. Provided by VK_EXT_device_generated_commands.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdExecuteGeneratedCommandsEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Executes commands generated on the GPU from an indirect commands layout. The GeneratedCommandsInfoEXT specifies the indirect commands layout, pipeline/shader objects, and the buffer containing the generated command data.

If is_preprocessed is true, the command data was prepared by a prior cmd_preprocess_generated_commands_ext call. Otherwise, preprocessing and execution happen in one step.

Requires VK_EXT_device_generated_commands.

pub unsafe fn cmd_preprocess_generated_commands_ext( &self, command_buffer: CommandBuffer, p_generated_commands_info: &GeneratedCommandsInfoEXT, state_command_buffer: CommandBuffer, )

Wraps vkCmdPreprocessGeneratedCommandsEXT. Provided by VK_EXT_device_generated_commands.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.
• stateCommandBuffer must be externally synchronized.

Panics

Panics if vkCmdPreprocessGeneratedCommandsEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Preprocesses device-generated commands into an intermediate format. This can be done in a separate command buffer or pass, then executed later with cmd_execute_generated_commands_ext (with is_preprocessed = true).

Separating preprocessing from execution allows overlapping the preprocessing work with other GPU tasks.

Requires VK_EXT_device_generated_commands.

pub unsafe fn get_generated_commands_memory_requirements_ext( &self, p_info: &GeneratedCommandsMemoryRequirementsInfoEXT, p_memory_requirements: &mut MemoryRequirements2, )

Wraps vkGetGeneratedCommandsMemoryRequirementsEXT. Provided by VK_EXT_device_generated_commands.

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetGeneratedCommandsMemoryRequirementsEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Queries the memory requirements for preprocessing and executing device-generated commands. Returns a MemoryRequirements2 with the size and alignment needed for the preprocess buffer.

Call this before allocating the preprocess buffer used by cmd_preprocess_generated_commands_ext and cmd_execute_generated_commands_ext.

Requires VK_EXT_device_generated_commands.

pub unsafe fn create_indirect_commands_layout_ext( &self, p_create_info: &IndirectCommandsLayoutCreateInfoEXT, allocator: Option<&AllocationCallbacks>, ) -> VkResult<IndirectCommandsLayoutEXT>

Wraps vkCreateIndirectCommandsLayoutEXT. Provided by VK_EXT_device_generated_commands.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkCreateIndirectCommandsLayoutEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Creates an indirect commands layout that defines the structure of GPU-generated command sequences. Each token in the layout describes one command element (draw, dispatch, push constant, vertex buffer bind, index buffer bind, etc.).

The layout is used with cmd_execute_generated_commands_ext.

Destroy with destroy_indirect_commands_layout_ext.

Requires VK_EXT_device_generated_commands.

pub unsafe fn destroy_indirect_commands_layout_ext( &self, indirect_commands_layout: IndirectCommandsLayoutEXT, allocator: Option<&AllocationCallbacks>, )

Wraps vkDestroyIndirectCommandsLayoutEXT. Provided by VK_EXT_device_generated_commands.

Safety

• device (self) must be valid and not destroyed.
• indirectCommandsLayout must be externally synchronized.

Panics

Panics if vkDestroyIndirectCommandsLayoutEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Destroys an indirect commands layout created with create_indirect_commands_layout_ext.

Requires VK_EXT_device_generated_commands.

pub unsafe fn create_indirect_execution_set_ext( &self, p_create_info: &IndirectExecutionSetCreateInfoEXT, allocator: Option<&AllocationCallbacks>, ) -> VkResult<IndirectExecutionSetEXT>

Wraps vkCreateIndirectExecutionSetEXT. Provided by VK_EXT_device_generated_commands.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkCreateIndirectExecutionSetEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Creates an indirect execution set, a table of pipelines or shader objects that can be indexed at execution time by device-generated commands.

The GPU selects which pipeline/shader to use based on an index in the generated command stream, enabling fully GPU-driven material/shader selection.

Destroy with destroy_indirect_execution_set_ext.

Requires VK_EXT_device_generated_commands.

pub unsafe fn destroy_indirect_execution_set_ext( &self, indirect_execution_set: IndirectExecutionSetEXT, allocator: Option<&AllocationCallbacks>, )

Wraps vkDestroyIndirectExecutionSetEXT. Provided by VK_EXT_device_generated_commands.

Safety

• device (self) must be valid and not destroyed.
• indirectExecutionSet must be externally synchronized.

Panics

Panics if vkDestroyIndirectExecutionSetEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Destroys an indirect execution set created with create_indirect_execution_set_ext.

Requires VK_EXT_device_generated_commands.

pub unsafe fn update_indirect_execution_set_pipeline_ext( &self, indirect_execution_set: IndirectExecutionSetEXT, p_execution_set_writes: &[WriteIndirectExecutionSetPipelineEXT], )

Wraps vkUpdateIndirectExecutionSetPipelineEXT. Provided by VK_EXT_device_generated_commands.

Safety

• device (self) must be valid and not destroyed.
• indirectExecutionSet must be externally synchronized.

Panics

Panics if vkUpdateIndirectExecutionSetPipelineEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Updates entries in an indirect execution set that holds pipelines. Each WriteIndirectExecutionSetPipelineEXT maps an index to a pipeline handle.

The pipelines must be compatible with the initial pipeline used to create the execution set.

Requires VK_EXT_device_generated_commands.

pub unsafe fn update_indirect_execution_set_shader_ext( &self, indirect_execution_set: IndirectExecutionSetEXT, p_execution_set_writes: &[WriteIndirectExecutionSetShaderEXT], )

Wraps vkUpdateIndirectExecutionSetShaderEXT. Provided by VK_EXT_device_generated_commands.

Safety

• device (self) must be valid and not destroyed.
• indirectExecutionSet must be externally synchronized.

Panics

Panics if vkUpdateIndirectExecutionSetShaderEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Updates entries in an indirect execution set that holds shader objects. Each WriteIndirectExecutionSetShaderEXT maps an index to a shader object handle.

The shaders must be compatible with the initial shader used to create the execution set.

Requires VK_EXT_device_generated_commands.

pub unsafe fn cmd_push_descriptor_set( &self, command_buffer: CommandBuffer, pipeline_bind_point: PipelineBindPoint, layout: PipelineLayout, set: u32, p_descriptor_writes: &[WriteDescriptorSet], )

Wraps vkCmdPushDescriptorSet. Provided by VK_COMPUTE_VERSION_1_4.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdPushDescriptorSet was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Pushes descriptor updates directly into the command buffer without allocating a descriptor set from a pool. The descriptors are embedded in the command stream and only live for the duration of the current command buffer recording.

Advantages:

• No descriptor pool allocation or management.
• No need to track descriptor set lifetimes.
• Ideal for per-draw data that changes every frame.

Trade-offs:

• Inflates command buffer size (descriptors are stored inline).
• Not suitable for large descriptor sets, use conventional allocated sets for sets with many bindings.

The pipeline layout must have been created with DESCRIPTOR_SET_LAYOUT_CREATE_PUSH_DESCRIPTOR on the target set index.

Core in Vulkan 1.4. Previously available via VK_KHR_push_descriptor.

pub unsafe fn trim_command_pool( &self, command_pool: CommandPool, flags: CommandPoolTrimFlags, )

Wraps vkTrimCommandPool. Provided by VK_BASE_VERSION_1_1.

Safety

• device (self) must be valid and not destroyed.
• commandPool must be externally synchronized.

Panics

Panics if vkTrimCommandPool was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Returns unused memory from a command pool back to the system. This is a hint to the driver, it may or may not actually release memory.

Call this after a period of high command buffer allocation followed by a return to lower usage (e.g. after loading screens or level transitions). It does not affect any allocated command buffers.

Unlike reset_command_pool, trimming does not reset or invalidate command buffers. It only reclaims excess internal memory that the pool pre-allocated.

In a steady-state frame loop where you reset the pool every frame, trimming is unnecessary, the pool reuses its memory naturally.

pub unsafe fn get_memory_win32_handle_khr( &self, p_get_win32_handle_info: &MemoryGetWin32HandleInfoKHR, ) -> VkResult<isize>

Wraps vkGetMemoryWin32HandleKHR. Provided by VK_KHR_external_memory_win32.

Errors

• VK_ERROR_TOO_MANY_OBJECTS
• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetMemoryWin32HandleKHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Exports a device memory allocation as a Windows HANDLE. The handle can be shared with other processes or APIs (D3D12, CUDA) for GPU memory interop.

MemoryGetWin32HandleInfoKHR specifies the DeviceMemory and handle type (OPAQUE_WIN32 or OPAQUE_WIN32_KMT).

For OPAQUE_WIN32, the handle must be closed with CloseHandle when done. OPAQUE_WIN32_KMT handles are kernel-managed and do not need explicit cleanup.

Windows only. Use get_memory_fd_khr on Linux.

pub unsafe fn get_memory_win32_handle_properties_khr( &self, handle_type: ExternalMemoryHandleTypeFlagBits, handle: isize, p_memory_win32_handle_properties: &mut MemoryWin32HandlePropertiesKHR, ) -> VkResult<()>

Wraps vkGetMemoryWin32HandlePropertiesKHR. Provided by VK_KHR_external_memory_win32.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_INVALID_EXTERNAL_HANDLE
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetMemoryWin32HandlePropertiesKHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Queries which memory types are compatible with an external Windows HANDLE. Use this when importing memory from a handle received from another process or API (e.g., D3D11 shared textures).

Returns MemoryWin32HandlePropertiesKHR with memory_type_bits indicating compatible memory type indices for import.

Not valid for OPAQUE_WIN32 or OPAQUE_WIN32_KMT handle types, those have their memory type determined by the exporting allocation.

pub unsafe fn get_memory_fd_khr( &self, p_get_fd_info: &MemoryGetFdInfoKHR, ) -> VkResult<c_int>

Wraps vkGetMemoryFdKHR. Provided by VK_KHR_external_memory_fd.

Errors

• VK_ERROR_TOO_MANY_OBJECTS
• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetMemoryFdKHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Exports a device memory allocation as a POSIX file descriptor. The fd can be sent to another process (via Unix domain sockets) or another Vulkan device to share GPU memory.

MemoryGetFdInfoKHR specifies the DeviceMemory and the handle type (OPAQUE_FD or DMA_BUF). The memory must have been allocated with ExportMemoryAllocateInfo requesting the corresponding handle type.

The caller owns the returned fd and must close it when done. Each call returns a new fd, duplicates are independent.

Linux/Android only. Use get_memory_win32_handle_khr on Windows.

pub unsafe fn get_memory_fd_properties_khr( &self, handle_type: ExternalMemoryHandleTypeFlagBits, fd: c_int, p_memory_fd_properties: &mut MemoryFdPropertiesKHR, ) -> VkResult<()>

Wraps vkGetMemoryFdPropertiesKHR. Provided by VK_KHR_external_memory_fd.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_INVALID_EXTERNAL_HANDLE
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetMemoryFdPropertiesKHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Queries which memory types are compatible with an external file descriptor. Use this when importing memory from an fd received from another process or API.

Returns MemoryFdPropertiesKHR with a memory_type_bits bitmask indicating which memory type indices can be used when allocating memory to import this fd.

Only valid for DMA_BUF handle types. OPAQUE_FD handles don’t need this query, their memory type is determined by the exporting allocation.

pub unsafe fn get_memory_zircon_handle_fuchsia( &self, p_get_zircon_handle_info: &MemoryGetZirconHandleInfoFUCHSIA, ) -> VkResult<u32>

Wraps vkGetMemoryZirconHandleFUCHSIA. Provided by VK_FUCHSIA_external_memory.

Errors

• VK_ERROR_TOO_MANY_OBJECTS
• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetMemoryZirconHandleFUCHSIA was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Exports a Vulkan device memory allocation as a Fuchsia Zircon VMO handle. The returned handle can be shared with other Fuchsia processes. Fuchsia OS only.

Requires VK_FUCHSIA_external_memory.

pub unsafe fn get_memory_zircon_handle_properties_fuchsia( &self, handle_type: ExternalMemoryHandleTypeFlagBits, zircon_handle: u32, p_memory_zircon_handle_properties: &mut MemoryZirconHandlePropertiesFUCHSIA, ) -> VkResult<()>

Wraps vkGetMemoryZirconHandlePropertiesFUCHSIA. Provided by VK_FUCHSIA_external_memory.

Errors

• VK_ERROR_INVALID_EXTERNAL_HANDLE
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetMemoryZirconHandlePropertiesFUCHSIA was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Queries the memory properties (compatible memory type bits) for a Zircon VMO handle. Use before importing external memory to determine which memory type to allocate. Fuchsia OS only.

Requires VK_FUCHSIA_external_memory.

pub unsafe fn get_memory_remote_address_nv( &self, p_memory_get_remote_address_info: &MemoryGetRemoteAddressInfoNV, ) -> VkResult<*mut c_void>

Wraps vkGetMemoryRemoteAddressNV. Provided by VK_NV_external_memory_rdma.

Errors

• VK_ERROR_INVALID_EXTERNAL_HANDLE
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetMemoryRemoteAddressNV was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Retrieves a remote device address for a Vulkan memory allocation, enabling RDMA (Remote Direct Memory Access) between devices. The returned address can be used by another device to directly access this memory over a high-speed interconnect.

Requires VK_NV_external_memory_rdma.

pub unsafe fn get_memory_sci_buf_nv( &self, p_get_sci_buf_info: &MemoryGetSciBufInfoNV, ) -> VkResult<c_void>

Wraps vkGetMemorySciBufNV.

Errors

• VK_ERROR_INITIALIZATION_FAILED
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetMemorySciBufNV was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Exports a Vulkan device memory allocation as a QNX SCI buffer for cross-process or cross-API sharing on QNX platforms.

Requires VK_NV_external_memory_sci_buf. QNX only.

pub unsafe fn get_semaphore_win32_handle_khr( &self, p_get_win32_handle_info: &SemaphoreGetWin32HandleInfoKHR, ) -> VkResult<isize>

Wraps vkGetSemaphoreWin32HandleKHR. Provided by VK_KHR_external_semaphore_win32.

Errors

• VK_ERROR_TOO_MANY_OBJECTS
• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetSemaphoreWin32HandleKHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Exports a semaphore’s synchronization state as a Windows HANDLE. The handle can be shared with other processes or APIs for cross-process GPU synchronization.

SemaphoreGetWin32HandleInfoKHR specifies the semaphore and handle type (OPAQUE_WIN32, OPAQUE_WIN32_KMT, or D3D12_FENCE).

For OPAQUE_WIN32 and D3D12_FENCE, close the handle with CloseHandle when done. OPAQUE_WIN32_KMT handles are kernel-managed.

Windows only. Use get_semaphore_fd_khr on Linux.

pub unsafe fn import_semaphore_win32_handle_khr( &self, p_import_semaphore_win32_handle_info: &ImportSemaphoreWin32HandleInfoKHR, ) -> VkResult<()>

Wraps vkImportSemaphoreWin32HandleKHR. Provided by VK_KHR_external_semaphore_win32.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_INVALID_EXTERNAL_HANDLE
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkImportSemaphoreWin32HandleKHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Imports a synchronization payload from a Windows HANDLE into a semaphore. After import, the semaphore uses the external synchronization state.

ImportSemaphoreWin32HandleInfoKHR specifies the target semaphore, handle type, handle (or name for named handles), and whether the import is temporary or permanent.

The handle is duplicated internally, the caller retains ownership and should close it when no longer needed.

Windows only. Use import_semaphore_fd_khr on Linux.

pub unsafe fn get_semaphore_fd_khr( &self, p_get_fd_info: &SemaphoreGetFdInfoKHR, ) -> VkResult<c_int>

Wraps vkGetSemaphoreFdKHR. Provided by VK_KHR_external_semaphore_fd.

Errors

• VK_ERROR_TOO_MANY_OBJECTS
• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetSemaphoreFdKHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Exports a semaphore’s synchronization state as a POSIX file descriptor. The fd can be transferred to another process for cross-process GPU synchronization.

SemaphoreGetFdInfoKHR specifies the semaphore and handle type (OPAQUE_FD or SYNC_FD). For SYNC_FD, the semaphore must be signaled or have a pending signal operation, exporting transfers ownership and resets the semaphore to unsignaled.

The caller owns the returned fd. For OPAQUE_FD, each export creates a new reference. For SYNC_FD, the export is a move, the semaphore payload is transferred.

Linux/Android only. Use get_semaphore_win32_handle_khr on Windows.

pub unsafe fn import_semaphore_fd_khr( &self, p_import_semaphore_fd_info: &ImportSemaphoreFdInfoKHR, ) -> VkResult<()>

Wraps vkImportSemaphoreFdKHR. Provided by VK_KHR_external_semaphore_fd.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_INVALID_EXTERNAL_HANDLE
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkImportSemaphoreFdKHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Imports a synchronization payload from a POSIX file descriptor into a semaphore. After import, the semaphore uses the external synchronization state.

ImportSemaphoreFdInfoKHR specifies the target semaphore, handle type, fd, and whether the import is temporary (payload consumed on first wait) or permanent.

For SYNC_FD, the import takes ownership of the fd, do not close it afterward. For OPAQUE_FD, the fd is duplicated internally and can be closed after the call.

Linux/Android only. Use import_semaphore_win32_handle_khr on Windows.

pub unsafe fn get_semaphore_zircon_handle_fuchsia( &self, p_get_zircon_handle_info: &SemaphoreGetZirconHandleInfoFUCHSIA, ) -> VkResult<u32>

Wraps vkGetSemaphoreZirconHandleFUCHSIA. Provided by VK_FUCHSIA_external_semaphore.

Errors

• VK_ERROR_TOO_MANY_OBJECTS
• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetSemaphoreZirconHandleFUCHSIA was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Exports a Vulkan semaphore as a Fuchsia Zircon event handle for cross-process synchronisation. Fuchsia OS only.

Requires VK_FUCHSIA_external_semaphore.

pub unsafe fn import_semaphore_zircon_handle_fuchsia( &self, p_import_semaphore_zircon_handle_info: &ImportSemaphoreZirconHandleInfoFUCHSIA, ) -> VkResult<()>

Wraps vkImportSemaphoreZirconHandleFUCHSIA. Provided by VK_FUCHSIA_external_semaphore.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_INVALID_EXTERNAL_HANDLE
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkImportSemaphoreZirconHandleFUCHSIA was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Imports a Fuchsia Zircon event handle into an existing Vulkan semaphore for cross-process synchronisation. Fuchsia OS only.

Requires VK_FUCHSIA_external_semaphore.

pub unsafe fn get_fence_win32_handle_khr( &self, p_get_win32_handle_info: &FenceGetWin32HandleInfoKHR, ) -> VkResult<isize>

Wraps vkGetFenceWin32HandleKHR. Provided by VK_KHR_external_fence_win32.

Errors

• VK_ERROR_TOO_MANY_OBJECTS
• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetFenceWin32HandleKHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Exports a fence’s synchronization state as a Windows HANDLE for cross-process fence synchronization.

FenceGetWin32HandleInfoKHR specifies the fence and handle type (OPAQUE_WIN32 or OPAQUE_WIN32_KMT).

For OPAQUE_WIN32, close the handle with CloseHandle when done. OPAQUE_WIN32_KMT handles are kernel-managed.

Windows only. Use get_fence_fd_khr on Linux.

pub unsafe fn import_fence_win32_handle_khr( &self, p_import_fence_win32_handle_info: &ImportFenceWin32HandleInfoKHR, ) -> VkResult<()>

Wraps vkImportFenceWin32HandleKHR. Provided by VK_KHR_external_fence_win32.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_INVALID_EXTERNAL_HANDLE
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkImportFenceWin32HandleKHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Imports a synchronization payload from a Windows HANDLE into a fence.

ImportFenceWin32HandleInfoKHR specifies the target fence, handle type, handle (or name), and whether the import is temporary or permanent.

The handle is duplicated internally, the caller retains ownership and should close it when no longer needed.

Windows only. Use import_fence_fd_khr on Linux.

pub unsafe fn get_fence_fd_khr( &self, p_get_fd_info: &FenceGetFdInfoKHR, ) -> VkResult<c_int>

Wraps vkGetFenceFdKHR. Provided by VK_KHR_external_fence_fd.

Errors

• VK_ERROR_TOO_MANY_OBJECTS
• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetFenceFdKHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Exports a fence’s synchronization state as a POSIX file descriptor. The fd enables cross-process fence synchronization.

FenceGetFdInfoKHR specifies the fence and handle type (OPAQUE_FD or SYNC_FD). For SYNC_FD, the fence must be signaled or have a pending signal, exporting transfers the payload and resets the fence.

The caller owns the returned fd and must close it when done.

Linux/Android only. Use get_fence_win32_handle_khr on Windows.

pub unsafe fn import_fence_fd_khr( &self, p_import_fence_fd_info: &ImportFenceFdInfoKHR, ) -> VkResult<()>

Wraps vkImportFenceFdKHR. Provided by VK_KHR_external_fence_fd.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_INVALID_EXTERNAL_HANDLE
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkImportFenceFdKHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Imports a synchronization payload from a POSIX file descriptor into a fence.

ImportFenceFdInfoKHR specifies the target fence, handle type, fd, and whether the import is temporary (payload consumed on first wait/reset) or permanent.

For SYNC_FD, ownership of the fd transfers to the implementation, do not close it. For OPAQUE_FD, the fd is duplicated and can be closed after the call.

Linux/Android only. Use import_fence_win32_handle_khr on Windows.

pub unsafe fn get_fence_sci_sync_fence_nv( &self, p_get_sci_sync_handle_info: &FenceGetSciSyncInfoNV, ) -> VkResult<c_void>

Wraps vkGetFenceSciSyncFenceNV.

Errors

• VK_ERROR_INVALID_EXTERNAL_HANDLE
• VK_ERROR_NOT_PERMITTED
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetFenceSciSyncFenceNV was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Exports a Vulkan fence as a SciSync fence handle for cross-process synchronisation on NV safety-critical platforms. QNX/NVIDIA Safety only.

Requires VK_NV_external_sci_sync.

pub unsafe fn get_fence_sci_sync_obj_nv( &self, p_get_sci_sync_handle_info: &FenceGetSciSyncInfoNV, ) -> VkResult<c_void>

Wraps vkGetFenceSciSyncObjNV.

Errors

• VK_ERROR_INVALID_EXTERNAL_HANDLE
• VK_ERROR_NOT_PERMITTED
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetFenceSciSyncObjNV was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Exports a Vulkan fence as a SciSync object handle for cross-process synchronisation on NV safety-critical platforms. QNX/NVIDIA Safety only.

Requires VK_NV_external_sci_sync.

pub unsafe fn import_fence_sci_sync_fence_nv( &self, p_import_fence_sci_sync_info: &ImportFenceSciSyncInfoNV, ) -> VkResult<()>

Wraps vkImportFenceSciSyncFenceNV.

Errors

• VK_ERROR_INVALID_EXTERNAL_HANDLE
• VK_ERROR_NOT_PERMITTED
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkImportFenceSciSyncFenceNV was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Imports a SciSync fence handle into an existing Vulkan fence for cross-process synchronisation on NV safety-critical platforms. QNX/NVIDIA Safety only.

Requires VK_NV_external_sci_sync.

pub unsafe fn import_fence_sci_sync_obj_nv( &self, p_import_fence_sci_sync_info: &ImportFenceSciSyncInfoNV, ) -> VkResult<()>

Wraps vkImportFenceSciSyncObjNV.

Errors

• VK_ERROR_INVALID_EXTERNAL_HANDLE
• VK_ERROR_NOT_PERMITTED
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkImportFenceSciSyncObjNV was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Imports a SciSync object handle into an existing Vulkan fence for cross-process synchronisation on NV safety-critical platforms. QNX/NVIDIA Safety only.

Requires VK_NV_external_sci_sync.

pub unsafe fn get_semaphore_sci_sync_obj_nv( &self, p_get_sci_sync_info: &SemaphoreGetSciSyncInfoNV, ) -> VkResult<c_void>

Wraps vkGetSemaphoreSciSyncObjNV.

Errors

• VK_ERROR_INVALID_EXTERNAL_HANDLE
• VK_ERROR_NOT_PERMITTED
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetSemaphoreSciSyncObjNV was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Exports a Vulkan semaphore as a SciSync object handle for cross-process synchronisation on NV safety-critical platforms. QNX/NVIDIA Safety only.

Requires VK_NV_external_sci_sync.

pub unsafe fn import_semaphore_sci_sync_obj_nv( &self, p_import_semaphore_sci_sync_info: &ImportSemaphoreSciSyncInfoNV, ) -> VkResult<()>

Wraps vkImportSemaphoreSciSyncObjNV.

Errors

• VK_ERROR_INVALID_EXTERNAL_HANDLE
• VK_ERROR_NOT_PERMITTED
• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkImportSemaphoreSciSyncObjNV was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Imports a SciSync object handle into an existing Vulkan semaphore for cross-process synchronisation on NV safety-critical platforms. QNX/NVIDIA Safety only.

Requires VK_NV_external_sci_sync.

pub unsafe fn create_semaphore_sci_sync_pool_nv( &self, p_create_info: &SemaphoreSciSyncPoolCreateInfoNV, allocator: Option<&AllocationCallbacks>, ) -> VkResult<SemaphoreSciSyncPoolNV>

Wraps vkCreateSemaphoreSciSyncPoolNV.

Errors

• VK_ERROR_INITIALIZATION_FAILED
• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkCreateSemaphoreSciSyncPoolNV was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Creates a pool of SciSync-backed semaphores for efficient cross-engine synchronisation on NV safety-critical platforms. QNX/NVIDIA Safety only.

Requires VK_NV_external_sci_sync2.

pub unsafe fn destroy_semaphore_sci_sync_pool_nv( &self, semaphore_pool: SemaphoreSciSyncPoolNV, allocator: Option<&AllocationCallbacks>, )

Wraps vkDestroySemaphoreSciSyncPoolNV.

Safety

• device (self) must be valid and not destroyed.
• semaphorePool must be externally synchronized.

Panics

Panics if vkDestroySemaphoreSciSyncPoolNV was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Destroys a SciSync semaphore pool. The pool must not be in use by any pending operations. QNX/NVIDIA Safety only.

Requires VK_NV_external_sci_sync2.

pub unsafe fn display_power_control_ext( &self, display: DisplayKHR, p_display_power_info: &DisplayPowerInfoEXT, ) -> VkResult<()>

Wraps vkDisplayPowerControlEXT. Provided by VK_EXT_display_control.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkDisplayPowerControlEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Controls the power state of a display (e.g., standby, suspend, off, on). DisplayPowerInfoEXT specifies the desired power state.

Requires VK_EXT_display_control.

pub unsafe fn register_device_event_ext( &self, p_device_event_info: &DeviceEventInfoEXT, allocator: Option<&AllocationCallbacks>, ) -> VkResult<Fence>

Wraps vkRegisterDeviceEventEXT. Provided by VK_EXT_display_control.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkRegisterDeviceEventEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Registers a fence to be signaled when a device event occurs. DeviceEventInfoEXT specifies the event type (e.g., DISPLAY_HOTPLUG).

Returns a fence that will be signaled when the event fires.

Requires VK_EXT_display_control.

pub unsafe fn register_display_event_ext( &self, display: DisplayKHR, p_display_event_info: &DisplayEventInfoEXT, allocator: Option<&AllocationCallbacks>, ) -> VkResult<Fence>

Wraps vkRegisterDisplayEventEXT. Provided by VK_EXT_display_control.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkRegisterDisplayEventEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Registers a fence to be signaled when a display event occurs. DisplayEventInfoEXT specifies the event type (e.g., FIRST_PIXEL_OUT, signaled at the start of the first scanline after a present).

Returns a fence. Useful for frame pacing and display timing.

Requires VK_EXT_display_control.

pub unsafe fn get_swapchain_counter_ext( &self, swapchain: SwapchainKHR, counter: SurfaceCounterFlagBitsEXT, ) -> VkResult<u64>

Wraps vkGetSwapchainCounterEXT. Provided by VK_EXT_display_control.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_DEVICE_LOST
• VK_ERROR_OUT_OF_DATE_KHR
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetSwapchainCounterEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Queries a performance counter associated with a swapchain (e.g., vertical blanks). Returns the counter value as a u64.

The counter type is specified as a SurfaceCounterFlagBitsEXT (typically VBLANK).

Requires VK_EXT_display_control.

pub unsafe fn get_device_group_peer_memory_features( &self, heap_index: u32, local_device_index: u32, remote_device_index: u32, ) -> PeerMemoryFeatureFlags

Wraps vkGetDeviceGroupPeerMemoryFeatures. Provided by VK_BASE_VERSION_1_1.

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetDeviceGroupPeerMemoryFeatures was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Queries the memory access capabilities between two physical devices in a device group. Returns flags indicating whether memory allocated on one device can be copied to, accessed generically, or accessed natively from the other device.

Only relevant for multi-GPU device groups. On single-GPU systems this is not needed.

Use the returned flags to decide how to share resources across devices, for example, whether a texture on GPU 0 can be sampled directly by GPU 1, or whether it must be copied.

pub unsafe fn bind_buffer_memory2( &self, p_bind_infos: &[BindBufferMemoryInfo], ) -> VkResult<()>

Wraps vkBindBufferMemory2. Provided by VK_BASE_VERSION_1_1.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_INVALID_OPAQUE_CAPTURE_ADDRESS_KHR
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkBindBufferMemory2 was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Binds memory to one or more buffers in a single call. This is the Vulkan 1.1 batch version of bind_buffer_memory.

Each BindBufferMemoryInfo specifies a buffer, memory object, and offset, the same parameters as bind_buffer_memory, but batched.

Use BindBufferMemoryDeviceGroupInfo in the pNext chain to bind memory for specific devices in a device group (multi-GPU). For single-GPU usage, bind_buffer_memory and bind_buffer_memory2 are equivalent.

pub unsafe fn bind_image_memory2( &self, p_bind_infos: &[BindImageMemoryInfo], ) -> VkResult<()>

Wraps vkBindImageMemory2. Provided by VK_BASE_VERSION_1_1.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkBindImageMemory2 was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Binds memory to one or more images in a single call. This is the Vulkan 1.1 batch version of bind_image_memory.

Also required when binding memory to images created with disjoint multi-planar formats, each plane is bound separately via BindImagePlaneMemoryInfo in the pNext chain.

For device groups (multi-GPU), chain BindImageMemoryDeviceGroupInfo to assign memory per device and specify split-instance bind regions.

pub unsafe fn cmd_set_device_mask( &self, command_buffer: CommandBuffer, device_mask: u32, )

Wraps vkCmdSetDeviceMask. Provided by VK_BASE_VERSION_1_1.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdSetDeviceMask was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Sets the device mask for subsequent commands in a command buffer when using device groups (multi-GPU with VK_KHR_device_group / Vulkan 1.1).

The device mask is a bitmask where bit i indicates that subsequent commands execute on physical device i in the device group.

For single-GPU systems (the common case), the device mask is always 0x1 and this command is not needed.

Must only be called outside a render pass, or inside a render pass that was begun with DEVICE_GROUP_BEGIN_INFO and has the DEVICE_GROUP flag set.

pub unsafe fn get_device_group_present_capabilities_khr( &self, p_device_group_present_capabilities: &mut DeviceGroupPresentCapabilitiesKHR, ) -> VkResult<()>

Wraps vkGetDeviceGroupPresentCapabilitiesKHR. Provided by VK_KHR_swapchain.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetDeviceGroupPresentCapabilitiesKHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Queries the present capabilities of a device group, which physical devices can present to which surfaces, and what presentation modes are supported.

Only relevant for multi-GPU device groups. On single-GPU systems, only DEVICE_GROUP_PRESENT_MODE_LOCAL is supported (each device presents its own images).

pub unsafe fn get_device_group_surface_present_modes_khr( &self, surface: SurfaceKHR, ) -> VkResult<DeviceGroupPresentModeFlagsKHR>

Wraps vkGetDeviceGroupSurfacePresentModesKHR. Provided by VK_KHR_swapchain.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_SURFACE_LOST_KHR
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.
• surface must be externally synchronized.

Panics

Panics if vkGetDeviceGroupSurfacePresentModesKHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Queries which device group present modes a surface supports. The returned bitmask indicates whether LOCAL, REMOTE, SUM, or LOCAL_MULTI_DEVICE modes are available.

Only relevant for multi-GPU device groups. On single-GPU systems, only LOCAL is supported.

pub unsafe fn acquire_next_image2_khr( &self, p_acquire_info: &AcquireNextImageInfoKHR, ) -> VkResult<u32>

Wraps vkAcquireNextImage2KHR. Provided by VK_KHR_swapchain.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_DEVICE_LOST
• VK_ERROR_OUT_OF_DATE_KHR
• VK_ERROR_SURFACE_LOST_KHR
• VK_ERROR_FULL_SCREEN_EXCLUSIVE_MODE_LOST_EXT
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkAcquireNextImage2KHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Extended version of acquire_next_image_khr that takes an AcquireNextImageInfoKHR struct with pNext support.

The key addition is device_mask for device groups (multi-GPU), specifying which physical devices the acquired image will be used on.

For single-GPU usage, this is functionally identical to acquire_next_image_khr.

pub unsafe fn cmd_dispatch_base( &self, command_buffer: CommandBuffer, base_group_x: u32, base_group_y: u32, base_group_z: u32, group_count_x: u32, group_count_y: u32, group_count_z: u32, )

Wraps vkCmdDispatchBase. Provided by VK_COMPUTE_VERSION_1_1.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdDispatchBase was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Dispatches a compute shader with a non-zero base workgroup offset. The shader’s gl_WorkGroupID starts at (base_group_x, base_group_y, base_group_z) instead of (0, 0, 0).

This is useful for splitting a large dispatch across multiple submissions or for device groups where different physical devices handle different regions of the workgroup space.

gl_NumWorkGroups reflects the group_count parameters, not the total. The shader sees workgroup IDs in the range [base, base + count).

For single-GPU, zero-base dispatches, use cmd_dispatch instead.

pub unsafe fn create_descriptor_update_template( &self, p_create_info: &DescriptorUpdateTemplateCreateInfo, allocator: Option<&AllocationCallbacks>, ) -> VkResult<DescriptorUpdateTemplate>

Wraps vkCreateDescriptorUpdateTemplate. Provided by VK_COMPUTE_VERSION_1_1.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkCreateDescriptorUpdateTemplate was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Creates a template that describes how to update a descriptor set from a compact block of host memory. Instead of building an array of WriteDescriptorSet structs, you define the layout once in the template and then call update_descriptor_set_with_template with a raw pointer to your data.

Benefits:

• Reduces CPU overhead for frequent descriptor updates.
• Avoids allocating WriteDescriptorSet arrays every frame.
• Pairs well with cmd_push_descriptor_set_with_template for push descriptors.

Each entry in the template maps an offset in your host data block to a descriptor binding, array element, and type. The driver compiles this into an optimised update path.

Templates are immutable after creation and can be reused across frames.

pub unsafe fn destroy_descriptor_update_template( &self, descriptor_update_template: DescriptorUpdateTemplate, allocator: Option<&AllocationCallbacks>, )

Wraps vkDestroyDescriptorUpdateTemplate. Provided by VK_COMPUTE_VERSION_1_1.

Safety

• device (self) must be valid and not destroyed.
• descriptorUpdateTemplate must be externally synchronized.

Panics

Panics if vkDestroyDescriptorUpdateTemplate was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Destroys a descriptor update template. The template must not be in use by any pending update_descriptor_set_with_template or cmd_push_descriptor_set_with_template call.

pub unsafe fn update_descriptor_set_with_template( &self, descriptor_set: DescriptorSet, descriptor_update_template: DescriptorUpdateTemplate, p_data: *const c_void, )

Wraps vkUpdateDescriptorSetWithTemplate. Provided by VK_COMPUTE_VERSION_1_1.

Safety

• device (self) must be valid and not destroyed.
• descriptorSet must be externally synchronized.

Panics

Panics if vkUpdateDescriptorSetWithTemplate was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Updates a descriptor set using a descriptor update template and a raw pointer to a block of host data. The template defines the mapping from the data block to descriptor bindings.

This is faster than update_descriptor_sets for repeated updates with the same layout, the driver has pre-compiled the update path.

The data pointer must point to a block of memory laid out according to the template’s entry offsets and strides. The data is consumed immediately; the pointer does not need to remain valid after the call returns.

pub unsafe fn cmd_push_descriptor_set_with_template( &self, command_buffer: CommandBuffer, descriptor_update_template: DescriptorUpdateTemplate, layout: PipelineLayout, set: u32, p_data: *const c_void, )

Wraps vkCmdPushDescriptorSetWithTemplate. Provided by VK_COMPUTE_VERSION_1_4.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdPushDescriptorSetWithTemplate was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Combines push descriptors with descriptor update templates for maximum efficiency. Updates are pushed directly into the command buffer using the compact host data format defined by the template.

This is the fastest path for per-draw descriptor updates: no pool allocation, no WriteDescriptorSet array construction, and the driver has a pre-compiled update path from the template.

The template must have been created with DESCRIPTOR_UPDATE_TEMPLATE_TYPE_PUSH_DESCRIPTORS.

Core in Vulkan 1.4. Previously via VK_KHR_push_descriptor + VK_KHR_descriptor_update_template.

pub unsafe fn set_hdr_metadata_ext( &self, p_swapchains: &[SwapchainKHR], p_metadata: &HdrMetadataEXT, )

Wraps vkSetHdrMetadataEXT. Provided by VK_EXT_hdr_metadata.

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkSetHdrMetadataEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Sets HDR metadata (mastering display color volume, content light levels) for one or more swapchains. The compositor uses this to tone-map content appropriately for the connected display.

Call whenever the content characteristics change (e.g., switching between SDR UI and HDR scene rendering).

Requires VK_EXT_hdr_metadata.

pub unsafe fn get_swapchain_status_khr( &self, swapchain: SwapchainKHR, ) -> VkResult<()>

Wraps vkGetSwapchainStatusKHR. Provided by VK_KHR_shared_presentable_image.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_DEVICE_LOST
• VK_ERROR_OUT_OF_DATE_KHR
• VK_ERROR_SURFACE_LOST_KHR
• VK_ERROR_FULL_SCREEN_EXCLUSIVE_MODE_LOST_EXT
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.
• swapchain must be externally synchronized.

Panics

Panics if vkGetSwapchainStatusKHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Queries the current status of a shared presentable swapchain (created with PRESENT_MODE_SHARED_DEMAND_REFRESH or PRESENT_MODE_SHARED_CONTINUOUS_REFRESH).

Returns VK_SUCCESS if the swapchain is usable, or VK_SUBOPTIMAL_KHR / VK_ERROR_OUT_OF_DATE_KHR / surface-lost errors if the swapchain needs recreation.

Only relevant for shared presentable images (VK_KHR_shared_presentable_image). For regular swapchains, status is communicated through acquire_next_image_khr and queue_present_khr return values.

pub unsafe fn get_refresh_cycle_duration_google( &self, swapchain: SwapchainKHR, ) -> VkResult<RefreshCycleDurationGOOGLE>

Wraps vkGetRefreshCycleDurationGOOGLE. Provided by VK_GOOGLE_display_timing.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_DEVICE_LOST
• VK_ERROR_SURFACE_LOST_KHR
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.
• swapchain must be externally synchronized.

Panics

Panics if vkGetRefreshCycleDurationGOOGLE was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Queries the duration of a single refresh cycle (vsync interval) for the display associated with the given swapchain. Returns the period in nanoseconds. Essential for accurate frame pacing.

Requires VK_GOOGLE_display_timing.

pub unsafe fn get_past_presentation_timing_google( &self, swapchain: SwapchainKHR, ) -> VkResult<Vec<PastPresentationTimingGOOGLE>>

Wraps vkGetPastPresentationTimingGOOGLE. Provided by VK_GOOGLE_display_timing.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_DEVICE_LOST
• VK_ERROR_OUT_OF_DATE_KHR
• VK_ERROR_SURFACE_LOST_KHR
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.
• swapchain must be externally synchronized.

Panics

Panics if vkGetPastPresentationTimingGOOGLE was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Returns timing data for past presentations on a swapchain, including actual present time, earliest possible present time, and finish time. Uses the two-call idiom (call once to get the count, again to fill the buffer). Useful for frame pacing and latency analysis.

Requires VK_GOOGLE_display_timing.

pub unsafe fn cmd_set_viewport_w_scaling_nv( &self, command_buffer: CommandBuffer, first_viewport: u32, p_viewport_w_scalings: &[ViewportWScalingNV], )

Wraps vkCmdSetViewportWScalingNV. Provided by VK_NV_clip_space_w_scaling.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdSetViewportWScalingNV was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Sets per-viewport W scaling factors for lens-matched shading. The W scaling modifies the clip-space W coordinate to account for lens distortion in VR headsets, enabling more efficient shading by varying pixel density across the viewport.

Requires VK_NV_clip_space_w_scaling.

pub unsafe fn cmd_set_discard_rectangle_ext( &self, command_buffer: CommandBuffer, first_discard_rectangle: u32, p_discard_rectangles: &[Rect2D], )

Wraps vkCmdSetDiscardRectangleEXT. Provided by VK_EXT_discard_rectangles.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdSetDiscardRectangleEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Dynamically sets the discard rectangles for the current command buffer. Fragments inside (or outside, depending on mode) these rectangles are discarded before the fragment shader runs.

Useful for multi-view or split-screen rendering to cheaply cull fragments that belong to a different viewport.

Requires VK_EXT_discard_rectangles.

pub unsafe fn cmd_set_discard_rectangle_enable_ext( &self, command_buffer: CommandBuffer, discard_rectangle_enable: bool, )

Wraps vkCmdSetDiscardRectangleEnableEXT. Provided by VK_EXT_discard_rectangles.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdSetDiscardRectangleEnableEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Dynamically enables or disables discard rectangles for subsequent draw commands. When disabled, no fragments are discarded regardless of the configured rectangles.

Requires VK_EXT_discard_rectangles.

pub unsafe fn cmd_set_discard_rectangle_mode_ext( &self, command_buffer: CommandBuffer, discard_rectangle_mode: DiscardRectangleModeEXT, )

Wraps vkCmdSetDiscardRectangleModeEXT. Provided by VK_EXT_discard_rectangles.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdSetDiscardRectangleModeEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Dynamically sets whether fragments inside or outside the discard rectangles are discarded. INCLUSIVE discards fragments inside the rectangles; EXCLUSIVE discards fragments outside.

Requires VK_EXT_discard_rectangles.

pub unsafe fn cmd_set_sample_locations_ext( &self, command_buffer: CommandBuffer, p_sample_locations_info: &SampleLocationsInfoEXT, )

Wraps vkCmdSetSampleLocationsEXT. Provided by VK_EXT_sample_locations.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdSetSampleLocationsEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Sets custom sample locations for multisampled rasterization. SampleLocationsInfoEXT specifies the grid size, sample count, and per-sample (x, y) positions within each pixel.

Custom sample locations enable techniques like temporal AA jittering and programmable MSAA patterns.

Must be called when custom sample locations are enabled (via cmd_set_sample_locations_enable_ext or pipeline state).

Requires VK_EXT_sample_locations.

pub unsafe fn get_buffer_memory_requirements2( &self, p_info: &BufferMemoryRequirementsInfo2, p_memory_requirements: &mut MemoryRequirements2, )

Wraps vkGetBufferMemoryRequirements2. Provided by VK_BASE_VERSION_1_1.

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetBufferMemoryRequirements2 was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Vulkan 1.1 version of get_buffer_memory_requirements that supports extensible output structs via pNext.

Chain MemoryDedicatedRequirements to query whether the driver prefers or requires a dedicated allocation for this buffer. If prefers_dedicated_allocation is true, allocating a dedicated DeviceMemory for this buffer may improve performance.

The base MemoryRequirements (size, alignment, memory type bits) is identical to what get_buffer_memory_requirements returns.

pub unsafe fn get_image_memory_requirements2( &self, p_info: &ImageMemoryRequirementsInfo2, p_memory_requirements: &mut MemoryRequirements2, )

Wraps vkGetImageMemoryRequirements2. Provided by VK_BASE_VERSION_1_1.

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetImageMemoryRequirements2 was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Vulkan 1.1 version of get_image_memory_requirements that supports extensible output structs via pNext.

Chain MemoryDedicatedRequirements to query whether the driver prefers or requires a dedicated allocation for this image. Dedicated allocations are common for large render targets and swapchain-sized images, some drivers require them.

For multi-planar images, chain ImagePlaneMemoryRequirementsInfo in the input to query requirements for a specific plane.

The base MemoryRequirements is identical to what get_image_memory_requirements returns.

pub unsafe fn get_image_sparse_memory_requirements2( &self, p_info: &ImageSparseMemoryRequirementsInfo2, ) -> Vec<SparseImageMemoryRequirements2>

Wraps vkGetImageSparseMemoryRequirements2. Provided by VK_BASE_VERSION_1_1.

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetImageSparseMemoryRequirements2 was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Vulkan 1.1 version of get_image_sparse_memory_requirements that supports extensible output structs via pNext. Returns the sparse memory requirements for an image created with sparse flags.

For non-sparse images, returns an empty list. Only relevant if you are using sparse resources.

pub unsafe fn get_device_buffer_memory_requirements( &self, p_info: &DeviceBufferMemoryRequirements, p_memory_requirements: &mut MemoryRequirements2, )

Wraps vkGetDeviceBufferMemoryRequirements. Provided by VK_BASE_VERSION_1_3.

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetDeviceBufferMemoryRequirements was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Vulkan 1.3 command that queries memory requirements for a buffer without creating it first. Pass a DeviceBufferMemoryRequirements containing the hypothetical BufferCreateInfo.

This lets you pre-plan memory allocations before creating any objects, useful for memory allocation strategies that need to know sizes and alignments up front.

The returned requirements are identical to what get_buffer_memory_requirements2 would return for an actual buffer created with the same parameters.

pub unsafe fn get_device_image_memory_requirements( &self, p_info: &DeviceImageMemoryRequirements, p_memory_requirements: &mut MemoryRequirements2, )

Wraps vkGetDeviceImageMemoryRequirements. Provided by VK_BASE_VERSION_1_3.

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetDeviceImageMemoryRequirements was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Vulkan 1.3 command that queries memory requirements for an image without creating it first. Pass a DeviceImageMemoryRequirements containing the hypothetical ImageCreateInfo.

Useful for pre-planning memory allocations or estimating VRAM usage before committing to image creation.

For multi-planar images, set plane_aspect to query requirements for a specific plane.

The returned requirements are identical to what get_image_memory_requirements2 would return for an actual image created with the same parameters.

pub unsafe fn get_device_image_sparse_memory_requirements( &self, p_info: &DeviceImageMemoryRequirements, ) -> Vec<SparseImageMemoryRequirements2>

Wraps vkGetDeviceImageSparseMemoryRequirements. Provided by VK_BASE_VERSION_1_3.

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetDeviceImageSparseMemoryRequirements was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Vulkan 1.3 command that queries sparse memory requirements for an image without creating it first. The counterpart to get_device_image_memory_requirements for sparse images.

Only relevant if you are using sparse resources with the hypothetical image creation parameters.

pub unsafe fn create_sampler_ycbcr_conversion( &self, p_create_info: &SamplerYcbcrConversionCreateInfo, allocator: Option<&AllocationCallbacks>, ) -> VkResult<SamplerYcbcrConversion>

Wraps vkCreateSamplerYcbcrConversion. Provided by VK_COMPUTE_VERSION_1_1.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkCreateSamplerYcbcrConversion was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Creates a sampler YCBCR conversion object that describes how to convert YCBCR-encoded image data to RGB during sampling. Required for multi-planar formats like G8_B8_R8_3PLANE_420_UNORM commonly used in video decoding and camera capture.

The conversion parameters specify:

• Format: the multi-planar format being converted.
• YCBCR model: RGB_IDENTITY, YCBCR_IDENTITY, YCBCR_709, YCBCR_601, YCBCR_2020.
• Range: ITU_FULL or ITU_NARROW.
• Chroma location: where subsampled chroma samples are located relative to luma samples.
• Chroma filter: NEAREST or LINEAR for chroma upsampling.

The conversion object is attached to a sampler via SamplerYcbcrConversionInfo in the pNext chain, and that sampler must be used as an immutable sampler in the descriptor set layout.

pub unsafe fn destroy_sampler_ycbcr_conversion( &self, ycbcr_conversion: SamplerYcbcrConversion, allocator: Option<&AllocationCallbacks>, )

Wraps vkDestroySamplerYcbcrConversion. Provided by VK_COMPUTE_VERSION_1_1.

Safety

• device (self) must be valid and not destroyed.
• ycbcrConversion must be externally synchronized.

Panics

Panics if vkDestroySamplerYcbcrConversion was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Destroys a sampler YCBCR conversion object. Any sampler that was created with this conversion must already be destroyed.

After destruction, any descriptor set layout that used the associated sampler as an immutable sampler remains valid but cannot be used to allocate new descriptor sets.

pub unsafe fn get_device_queue2(&self, p_queue_info: &DeviceQueueInfo2) -> Queue

Wraps vkGetDeviceQueue2. Provided by VK_BASE_VERSION_1_1.

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetDeviceQueue2 was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Retrieves a queue handle for a queue created with specific flags. This is the Vulkan 1.1 version of get_device_queue that supports DeviceQueueInfo2 with queue creation flags.

Use this instead of get_device_queue when you created queues with non-zero DeviceQueueCreateFlags (e.g. PROTECTED for protected content processing). For queues created without flags, both get_device_queue and get_device_queue2 work.

pub unsafe fn create_validation_cache_ext( &self, p_create_info: &ValidationCacheCreateInfoEXT, allocator: Option<&AllocationCallbacks>, ) -> VkResult<ValidationCacheEXT>

Wraps vkCreateValidationCacheEXT. Provided by VK_EXT_validation_cache.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkCreateValidationCacheEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Creates a validation cache that stores the results of validation layer checks. Subsequent pipeline creations with the same shaders can skip redundant validation, improving pipeline creation time.

Provide previously saved cache data to warm-start the cache. Retrieve data with get_validation_cache_data_ext.

Destroy with destroy_validation_cache_ext.

Requires VK_EXT_validation_cache.

pub unsafe fn destroy_validation_cache_ext( &self, validation_cache: ValidationCacheEXT, allocator: Option<&AllocationCallbacks>, )

Wraps vkDestroyValidationCacheEXT. Provided by VK_EXT_validation_cache.

Safety

• device (self) must be valid and not destroyed.
• validationCache must be externally synchronized.

Panics

Panics if vkDestroyValidationCacheEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Destroys a validation cache created with create_validation_cache_ext.

Requires VK_EXT_validation_cache.

pub unsafe fn get_validation_cache_data_ext( &self, validation_cache: ValidationCacheEXT, p_data: *mut c_void, ) -> VkResult<usize>

Wraps vkGetValidationCacheDataEXT. Provided by VK_EXT_validation_cache.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetValidationCacheDataEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Retrieves the data from a validation cache for serialization to disk. Call once with a null buffer to query the size, then again with an appropriately sized buffer.

Feed the saved data back into create_validation_cache_ext on the next run to avoid redundant validation.

Requires VK_EXT_validation_cache.

pub unsafe fn merge_validation_caches_ext( &self, dst_cache: ValidationCacheEXT, p_src_caches: &[ValidationCacheEXT], ) -> VkResult<()>

Wraps vkMergeValidationCachesEXT. Provided by VK_EXT_validation_cache.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.
• dstCache must be externally synchronized.

Panics

Panics if vkMergeValidationCachesEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Merges one or more source validation caches into a destination cache. Useful for combining caches from parallel pipeline creation threads.

Requires VK_EXT_validation_cache.

pub unsafe fn get_descriptor_set_layout_support( &self, p_create_info: &DescriptorSetLayoutCreateInfo, p_support: &mut DescriptorSetLayoutSupport, )

Wraps vkGetDescriptorSetLayoutSupport. Provided by VK_COMPUTE_VERSION_1_1.

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetDescriptorSetLayoutSupport was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Queries whether a descriptor set layout with the given bindings can be created on this device, and returns information about the variable descriptor count limit if applicable.

Use this before creating layouts with very large descriptor counts or update-after-bind bindings to verify they are within device limits. The call is lightweight and does not allocate anything.

Chain DescriptorSetVariableDescriptorCountLayoutSupport in the output to query the maximum variable descriptor count for layouts that use DESCRIPTOR_BINDING_VARIABLE_DESCRIPTOR_COUNT.

pub unsafe fn get_swapchain_gralloc_usage_android( &self, format: Format, image_usage: ImageUsageFlags, ) -> VkResult<c_int>

Wraps vkGetSwapchainGrallocUsageANDROID.

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetSwapchainGrallocUsageANDROID was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Queries the Android gralloc usage flags needed for swapchain images with the given format and Vulkan image usage. Used internally by the Android WSI implementation. Android only.

Requires VK_ANDROID_native_buffer.

pub unsafe fn get_swapchain_gralloc_usage2_android( &self, format: Format, image_usage: ImageUsageFlags, swapchain_image_usage: SwapchainImageUsageFlagsANDROID, gralloc_consumer_usage: *mut u64, ) -> VkResult<u64>

Wraps vkGetSwapchainGrallocUsage2ANDROID.

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetSwapchainGrallocUsage2ANDROID was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Extended version of get_swapchain_gralloc_usage_android that additionally accepts swapchain-specific image usage flags and returns both producer and consumer gralloc usage. Android only.

Requires VK_ANDROID_native_buffer.

pub unsafe fn acquire_image_android( &self, image: Image, native_fence_fd: c_int, semaphore: Semaphore, fence: Fence, ) -> VkResult<()>

Wraps vkAcquireImageANDROID.

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkAcquireImageANDROID was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Acquires ownership of a swapchain image on Android. Takes a native fence FD for synchronisation and can signal a Vulkan semaphore or fence on completion. Android only.

Requires VK_ANDROID_native_buffer.

pub unsafe fn queue_signal_release_image_android( &self, queue: Queue, p_wait_semaphores: &[Semaphore], image: Image, ) -> VkResult<c_int>

Wraps vkQueueSignalReleaseImageANDROID.

Safety

• queue (self) must be valid and not destroyed.

Panics

Panics if vkQueueSignalReleaseImageANDROID was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Releases a swapchain image back to the Android compositor after rendering. Waits on the given semaphores and returns a native fence FD for external synchronisation. Android only.

Requires VK_ANDROID_native_buffer.

pub unsafe fn get_shader_info_amd( &self, pipeline: Pipeline, shader_stage: ShaderStageFlagBits, info_type: ShaderInfoTypeAMD, p_info: *mut c_void, ) -> VkResult<usize>

Wraps vkGetShaderInfoAMD. Provided by VK_AMD_shader_info.

Errors

• VK_ERROR_FEATURE_NOT_PRESENT
• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetShaderInfoAMD was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Queries AMD-specific shader information such as compiled binary statistics, disassembly, or resource usage. Call once with a null buffer to query the size, then again with an appropriately sized buffer.

Requires VK_AMD_shader_info.

pub unsafe fn set_local_dimming_amd( &self, swap_chain: SwapchainKHR, local_dimming_enable: bool, )

Wraps vkSetLocalDimmingAMD. Provided by VK_AMD_display_native_hdr.

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkSetLocalDimmingAMD was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Enables or disables local dimming on an AMD display with native HDR support. Local dimming improves HDR contrast by adjusting backlight zones independently.

Requires VK_AMD_display_native_hdr.

pub unsafe fn get_calibrated_timestamps_khr( &self, p_timestamp_infos: &[CalibratedTimestampInfoKHR], p_max_deviation: *mut u64, ) -> VkResult<Vec<u64>>

Wraps vkGetCalibratedTimestampsKHR. Provided by VK_KHR_calibrated_timestamps.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetCalibratedTimestampsKHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Samples multiple time domains simultaneously and returns calibrated timestamps. This allows correlating GPU timestamps (from cmd_write_timestamp2) with CPU time.

Each CalibratedTimestampInfoKHR specifies a time domain (e.g., DEVICE, CLOCK_MONOTONIC, CLOCK_MONOTONIC_RAW, QUERY_PERFORMANCE_COUNTER). All requested timestamps are sampled as close together as possible.

The returned max_deviation (in nanoseconds) bounds how far apart the samples could be, smaller is better. If deviation is too large, retry the call.

Query available time domains first with get_physical_device_calibrateable_time_domains_khr.

pub unsafe fn set_debug_utils_object_name_ext( &self, p_name_info: &DebugUtilsObjectNameInfoEXT, ) -> VkResult<()>

Wraps vkSetDebugUtilsObjectNameEXT. Provided by VK_EXT_debug_utils.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.
• pNameInfo must be externally synchronized.

Panics

Panics if vkSetDebugUtilsObjectNameEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Assigns a human-readable name to any Vulkan object. The name appears in validation layer messages, GPU debuggers (RenderDoc, Nsight), and crash reports.

DebugUtilsObjectNameInfoEXT specifies the object type, handle, and a null-terminated UTF-8 name string.

Set the name to null to remove a previously assigned name.

This is the most impactful debugging tool in Vulkan, name every object you create.

Requires VK_EXT_debug_utils.

pub unsafe fn set_debug_utils_object_tag_ext( &self, p_tag_info: &DebugUtilsObjectTagInfoEXT, ) -> VkResult<()>

Wraps vkSetDebugUtilsObjectTagEXT. Provided by VK_EXT_debug_utils.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkSetDebugUtilsObjectTagEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Attaches arbitrary binary data to a Vulkan object. Unlike set_debug_utils_object_name_ext (which sets a string), tags carry opaque byte data identified by a tag_name (u64).

Tags are consumed by debugging tools and layers that understand the specific tag_name value. Most applications only need set_debug_utils_object_name_ext, use tags for tool-specific metadata.

Requires VK_EXT_debug_utils.

pub unsafe fn queue_begin_debug_utils_label_ext( &self, queue: Queue, p_label_info: &DebugUtilsLabelEXT, )

Wraps vkQueueBeginDebugUtilsLabelEXT. Provided by VK_EXT_debug_utils.

Safety

• queue (self) must be valid and not destroyed.
• queue must be externally synchronized.

Panics

Panics if vkQueueBeginDebugUtilsLabelEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Opens a debug label region on a queue. All submissions between this call and the matching queue_end_debug_utils_label_ext are grouped under the label in GPU debuggers.

Unlike cmd_begin_debug_utils_label_ext (which operates inside a command buffer), this groups entire queue submissions.

Requires VK_EXT_debug_utils.

pub unsafe fn queue_end_debug_utils_label_ext(&self, queue: Queue)

Wraps vkQueueEndDebugUtilsLabelEXT. Provided by VK_EXT_debug_utils.

Safety

• queue (self) must be valid and not destroyed.
• queue must be externally synchronized.

Panics

Panics if vkQueueEndDebugUtilsLabelEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Closes the most recently opened debug label region on the queue. Must be paired with a prior queue_begin_debug_utils_label_ext on the same queue.

Requires VK_EXT_debug_utils.

pub unsafe fn queue_insert_debug_utils_label_ext( &self, queue: Queue, p_label_info: &DebugUtilsLabelEXT, )

Wraps vkQueueInsertDebugUtilsLabelEXT. Provided by VK_EXT_debug_utils.

Safety

• queue (self) must be valid and not destroyed.
• queue must be externally synchronized.

Panics

Panics if vkQueueInsertDebugUtilsLabelEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Inserts a single-point debug label on a queue. Marks a specific moment in the queue’s submission timeline without opening a begin/end region.

Requires VK_EXT_debug_utils.

pub unsafe fn cmd_begin_debug_utils_label_ext( &self, command_buffer: CommandBuffer, p_label_info: &DebugUtilsLabelEXT, )

Wraps vkCmdBeginDebugUtilsLabelEXT. Provided by VK_EXT_debug_utils.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdBeginDebugUtilsLabelEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Opens a debug label region in a command buffer. All commands recorded between this call and the matching cmd_end_debug_utils_label_ext are grouped under the label in GPU debuggers (RenderDoc, Nsight).

DebugUtilsLabelEXT specifies a name and optional RGBA color for the region.

Labels can nest. Every begin must have a matching end within the same command buffer.

Requires VK_EXT_debug_utils.

pub unsafe fn cmd_end_debug_utils_label_ext( &self, command_buffer: CommandBuffer, )

Wraps vkCmdEndDebugUtilsLabelEXT. Provided by VK_EXT_debug_utils.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdEndDebugUtilsLabelEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Closes the most recently opened debug label region in the command buffer. Must be paired with a prior cmd_begin_debug_utils_label_ext in the same command buffer.

Requires VK_EXT_debug_utils.

pub unsafe fn cmd_insert_debug_utils_label_ext( &self, command_buffer: CommandBuffer, p_label_info: &DebugUtilsLabelEXT, )

Wraps vkCmdInsertDebugUtilsLabelEXT. Provided by VK_EXT_debug_utils.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdInsertDebugUtilsLabelEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Inserts a single-point debug label into the command buffer (as opposed to a begin/end region). Useful for marking specific events like “shadow pass complete” or “post-process start” that don’t span a range of commands.

Appears as a marker in GPU debuggers (RenderDoc, Nsight).

Requires VK_EXT_debug_utils.

pub unsafe fn get_memory_host_pointer_properties_ext( &self, handle_type: ExternalMemoryHandleTypeFlagBits, p_host_pointer: *const c_void, p_memory_host_pointer_properties: &mut MemoryHostPointerPropertiesEXT, ) -> VkResult<()>

Wraps vkGetMemoryHostPointerPropertiesEXT. Provided by VK_EXT_external_memory_host.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_INVALID_EXTERNAL_HANDLE
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetMemoryHostPointerPropertiesEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Queries which memory types are compatible with importing a host pointer as external memory. Use this before allocating device memory backed by a host-allocated buffer to determine valid memory type bits.

Requires VK_EXT_external_memory_host.

pub unsafe fn cmd_write_buffer_marker_amd( &self, command_buffer: CommandBuffer, pipeline_stage: PipelineStageFlagBits, dst_buffer: Buffer, dst_offset: u64, marker: u32, )

Wraps vkCmdWriteBufferMarkerAMD. Provided by VK_AMD_buffer_marker.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdWriteBufferMarkerAMD was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Writes a 32-bit marker value into a buffer after a specified pipeline stage completes. Useful for fine-grained GPU progress tracking and debugging GPU hangs by identifying the last completed stage.

Requires VK_AMD_buffer_marker.

pub unsafe fn create_render_pass2( &self, p_create_info: &RenderPassCreateInfo2, allocator: Option<&AllocationCallbacks>, ) -> VkResult<RenderPass>

Wraps vkCreateRenderPass2. Provided by VK_GRAPHICS_VERSION_1_2.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkCreateRenderPass2 was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Vulkan 1.2 version of create_render_pass that uses extensible RenderPassCreateInfo2, AttachmentDescription2, SubpassDescription2, and SubpassDependency2 structs.

Key additions over the 1.0 version:

• View masks: SubpassDescription2::view_mask enables multiview rendering where a single draw call renders to multiple layers simultaneously (e.g. VR left/right eyes).
• Fragment density map: chain RenderPassFragmentDensityMapCreateInfoEXT for variable-rate shading via density maps.
• Depth/stencil resolve: SubpassDescriptionDepthStencilResolve enables automatic depth/stencil resolve at the end of a subpass.

Prefer this over create_render_pass when targeting Vulkan 1.2+. For Vulkan 1.3+, consider dynamic rendering (cmd_begin_rendering) which avoids render pass objects entirely.

pub unsafe fn cmd_begin_render_pass2( &self, command_buffer: CommandBuffer, p_render_pass_begin: &RenderPassBeginInfo, p_subpass_begin_info: &SubpassBeginInfo, )

Wraps vkCmdBeginRenderPass2. Provided by VK_GRAPHICS_VERSION_1_2.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdBeginRenderPass2 was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Vulkan 1.2 version of cmd_begin_render_pass that takes an additional SubpassBeginInfo parameter specifying the subpass contents mode.

The extensible structs allow chaining RenderPassAttachmentBeginInfo for imageless framebuffers, concrete image views are supplied at begin time rather than at framebuffer creation time.

Prefer this over cmd_begin_render_pass when targeting Vulkan 1.2+. For Vulkan 1.3+, consider cmd_begin_rendering (dynamic rendering) which eliminates render pass and framebuffer objects entirely.

pub unsafe fn cmd_next_subpass2( &self, command_buffer: CommandBuffer, p_subpass_begin_info: &SubpassBeginInfo, p_subpass_end_info: &SubpassEndInfo, )

Wraps vkCmdNextSubpass2. Provided by VK_GRAPHICS_VERSION_1_2.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdNextSubpass2 was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Vulkan 1.2 version of cmd_next_subpass that takes extensible SubpassBeginInfo and SubpassEndInfo structs.

Functionally identical to cmd_next_subpass. Prefer this when targeting Vulkan 1.2+.

pub unsafe fn cmd_end_render_pass2( &self, command_buffer: CommandBuffer, p_subpass_end_info: &SubpassEndInfo, )

Wraps vkCmdEndRenderPass2. Provided by VK_GRAPHICS_VERSION_1_2.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdEndRenderPass2 was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Vulkan 1.2 version of cmd_end_render_pass that takes an extensible SubpassEndInfo struct.

Functionally identical to cmd_end_render_pass. Prefer this when targeting Vulkan 1.2+.

pub unsafe fn get_semaphore_counter_value( &self, semaphore: Semaphore, ) -> VkResult<u64>

Wraps vkGetSemaphoreCounterValue. Provided by VK_BASE_VERSION_1_2.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_DEVICE_LOST
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetSemaphoreCounterValue was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Returns the current counter value of a timeline semaphore. Timeline semaphores (Vulkan 1.2) use a monotonically increasing 64-bit counter instead of binary signaled/unsignaled state.

Use this for non-blocking progress checks:

let value = get_semaphore_counter_value(semaphore);
if value >= expected_frame {
    // GPU has finished frame N, safe to reuse resources
}

For blocking waits, use wait_semaphores. For signaling from the CPU, use signal_semaphore.

Only valid for semaphores created with SEMAPHORE_TYPE_TIMELINE. Calling this on a binary semaphore is an error.

pub unsafe fn wait_semaphores( &self, p_wait_info: &SemaphoreWaitInfo, timeout: u64, ) -> VkResult<()>

Wraps vkWaitSemaphores. Provided by VK_BASE_VERSION_1_2.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_DEVICE_LOST
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkWaitSemaphores was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Blocks the calling thread until one or all of the specified timeline semaphores reach their target values, or until the timeout expires.

SemaphoreWaitInfo flags:

• SEMAPHORE_WAIT_ANY: return when any semaphore reaches its target. Without this flag, waits for all semaphores.

Timeout: in nanoseconds. u64::MAX for indefinite. Zero for a non-blocking poll.

Timeline semaphore waits are the CPU-side counterpart to queue_submit timeline waits. They replace many fence-based synchronisation patterns with a single, more flexible primitive.

// Wait for frame N to complete on the GPU
let info = SemaphoreWaitInfo::builder()
    .semaphores(&[timeline_sem])
    .values(&[frame_number]);
wait_semaphores(&info, u64::MAX);

Only valid for semaphores created with SEMAPHORE_TYPE_TIMELINE.

pub unsafe fn signal_semaphore( &self, p_signal_info: &SemaphoreSignalInfo, ) -> VkResult<()>

Wraps vkSignalSemaphore. Provided by VK_BASE_VERSION_1_2.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkSignalSemaphore was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Signals a timeline semaphore from the host (CPU), advancing its counter to the specified value. The value must be greater than the current counter value.

Use this to unblock GPU work that is waiting on the semaphore via queue_submit. For example, a CPU-side data preparation step can signal a timeline semaphore when data is ready, and the GPU waits on it before processing.

Only valid for semaphores created with SEMAPHORE_TYPE_TIMELINE.

Timeline semaphores replace many use cases that previously required fences, they can be waited on from both the CPU (wait_semaphores) and the GPU (queue_submit).

pub unsafe fn get_android_hardware_buffer_properties_android( &self, buffer: *const c_void, p_properties: &mut AndroidHardwareBufferPropertiesANDROID, ) -> VkResult<()>

Wraps vkGetAndroidHardwareBufferPropertiesANDROID. Provided by VK_ANDROID_external_memory_android_hardware_buffer.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_INVALID_EXTERNAL_HANDLE_KHR
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetAndroidHardwareBufferPropertiesANDROID was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Queries the Vulkan memory properties (memory type bits, size) for an Android AHardwareBuffer. Use before importing the buffer as Vulkan memory. Android only.

Requires VK_ANDROID_external_memory_android_hardware_buffer.

pub unsafe fn get_memory_android_hardware_buffer_android( &self, p_info: &MemoryGetAndroidHardwareBufferInfoANDROID, p_buffer: *mut *mut c_void, ) -> VkResult<()>

Wraps vkGetMemoryAndroidHardwareBufferANDROID. Provided by VK_ANDROID_external_memory_android_hardware_buffer.

Errors

• VK_ERROR_TOO_MANY_OBJECTS
• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetMemoryAndroidHardwareBufferANDROID was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Exports a Vulkan device memory allocation as an Android AHardwareBuffer for sharing with other Android APIs (camera, media codec, SurfaceFlinger). Android only.

Requires VK_ANDROID_external_memory_android_hardware_buffer.

pub unsafe fn cmd_draw_indirect_count( &self, command_buffer: CommandBuffer, buffer: Buffer, offset: u64, count_buffer: Buffer, count_buffer_offset: u64, max_draw_count: u32, stride: u32, )

Wraps vkCmdDrawIndirectCount. Provided by VK_GRAPHICS_VERSION_1_2.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdDrawIndirectCount was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Draws non-indexed geometry with both the draw parameters and the draw count read from GPU buffers. The non-indexed counterpart to cmd_draw_indexed_indirect_count.

See cmd_draw_indexed_indirect_count for a full explanation of the GPU-driven rendering pattern. The only difference is that this command reads DrawIndirectCommand entries instead of DrawIndexedIndirectCommand.

Core in Vulkan 1.2. Previously available via VK_KHR_draw_indirect_count.

pub unsafe fn cmd_draw_indexed_indirect_count( &self, command_buffer: CommandBuffer, buffer: Buffer, offset: u64, count_buffer: Buffer, count_buffer_offset: u64, max_draw_count: u32, stride: u32, )

Wraps vkCmdDrawIndexedIndirectCount. Provided by VK_GRAPHICS_VERSION_1_2.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdDrawIndexedIndirectCount was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Draws indexed geometry with both the draw parameters and the draw count read from GPU buffers. The count_buffer contains a u32 that limits how many DrawIndexedIndirectCommand entries from the main buffer are actually executed, up to max_draw_count.

This is the key primitive for GPU-driven rendering pipelines: a compute shader fills an indirect buffer and writes the surviving draw count after culling. The CPU does not need to know the count.

The main buffer must have BUFFER_USAGE_INDIRECT_BUFFER. The count buffer must also have BUFFER_USAGE_INDIRECT_BUFFER. The count offset must be a multiple of 4.

Core in Vulkan 1.2. Previously available via VK_KHR_draw_indirect_count.

pub unsafe fn cmd_set_checkpoint_nv( &self, command_buffer: CommandBuffer, p_checkpoint_marker: *const c_void, )

Wraps vkCmdSetCheckpointNV. Provided by VK_NV_device_diagnostic_checkpoints.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdSetCheckpointNV was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Inserts a checkpoint marker into the command buffer for diagnostic purposes. If the device is lost, the most recently executed checkpoint can be retrieved with get_queue_checkpoint_data_nv to identify which commands completed before the failure.

Requires VK_NV_device_diagnostic_checkpoints.

pub unsafe fn get_queue_checkpoint_data_nv( &self, queue: Queue, ) -> Vec<CheckpointDataNV>

Wraps vkGetQueueCheckpointDataNV. Provided by VK_NV_device_diagnostic_checkpoints.

Safety

• queue (self) must be valid and not destroyed.

Panics

Panics if vkGetQueueCheckpointDataNV was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Retrieves the checkpoint markers that were most recently executed on a queue before a device-lost event. Use with cmd_set_checkpoint_nv for post-mortem debugging.

Requires VK_NV_device_diagnostic_checkpoints.

pub unsafe fn cmd_bind_transform_feedback_buffers_ext( &self, command_buffer: CommandBuffer, first_binding: u32, p_buffers: &[Buffer], p_offsets: &u64, p_sizes: Option<&u64>, )

Wraps vkCmdBindTransformFeedbackBuffersEXT. Provided by VK_EXT_transform_feedback.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdBindTransformFeedbackBuffersEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Binds buffers for transform feedback output. Each binding slot receives vertex data streamed from the vertex or geometry shader during a transform feedback pass.

first_binding is the first binding index. Arrays of buffers, offsets, and sizes specify the output targets.

Must be called before cmd_begin_transform_feedback_ext.

Requires VK_EXT_transform_feedback.

pub unsafe fn cmd_begin_transform_feedback_ext( &self, command_buffer: CommandBuffer, first_counter_buffer: u32, p_counter_buffers: &[Buffer], p_counter_buffer_offsets: Option<&u64>, )

Wraps vkCmdBeginTransformFeedbackEXT. Provided by VK_EXT_transform_feedback.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdBeginTransformFeedbackEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Begins a transform feedback pass. Vertex shader outputs (or geometry shader outputs) are written to transform feedback buffers previously bound with cmd_bind_transform_feedback_buffers_ext.

first_counter_buffer and p_counter_buffer_offsets specify where to resume writing from (pass null offsets to start from scratch).

End with cmd_end_transform_feedback_ext.

Requires VK_EXT_transform_feedback and the transformFeedback feature.

pub unsafe fn cmd_end_transform_feedback_ext( &self, command_buffer: CommandBuffer, first_counter_buffer: u32, p_counter_buffers: &[Buffer], p_counter_buffer_offsets: Option<&u64>, )

Wraps vkCmdEndTransformFeedbackEXT. Provided by VK_EXT_transform_feedback.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdEndTransformFeedbackEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Ends a transform feedback pass started with cmd_begin_transform_feedback_ext. Counter values are written back to the counter buffers so that a subsequent pass can resume where this one left off.

Requires VK_EXT_transform_feedback.

pub unsafe fn cmd_begin_query_indexed_ext( &self, command_buffer: CommandBuffer, query_pool: QueryPool, query: u32, flags: QueryControlFlags, index: u32, )

Wraps vkCmdBeginQueryIndexedEXT. Provided by VK_EXT_transform_feedback.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdBeginQueryIndexedEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Begins an indexed query, like cmd_begin_query but with an additional index parameter that selects which vertex stream to query when used with transform feedback statistics queries.

For QUERY_TYPE_TRANSFORM_FEEDBACK_STREAM_EXT, the index selects the stream (0–3). For other query types, index must be 0.

End with cmd_end_query_indexed_ext.

Requires VK_EXT_transform_feedback.

pub unsafe fn cmd_end_query_indexed_ext( &self, command_buffer: CommandBuffer, query_pool: QueryPool, query: u32, index: u32, )

Wraps vkCmdEndQueryIndexedEXT. Provided by VK_EXT_transform_feedback.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdEndQueryIndexedEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Ends an indexed query started with cmd_begin_query_indexed_ext. The index parameter must match the one used in the begin call.

Requires VK_EXT_transform_feedback.

pub unsafe fn cmd_draw_indirect_byte_count_ext( &self, command_buffer: CommandBuffer, instance_count: u32, first_instance: u32, counter_buffer: Buffer, counter_buffer_offset: u64, counter_offset: u32, vertex_stride: u32, )

Wraps vkCmdDrawIndirectByteCountEXT. Provided by VK_EXT_transform_feedback.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdDrawIndirectByteCountEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Draws vertices using a byte count stored in a counter buffer (typically written by a transform feedback pass). The vertex count is computed as counter_value / vertex_stride.

counter_offset accounts for any header bytes before the counter value in the buffer.

Requires VK_EXT_transform_feedback.

pub unsafe fn cmd_set_exclusive_scissor_nv( &self, command_buffer: CommandBuffer, first_exclusive_scissor: u32, p_exclusive_scissors: &[Rect2D], )

Wraps vkCmdSetExclusiveScissorNV. Provided by VK_NV_scissor_exclusive.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdSetExclusiveScissorNV was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Dynamically sets the exclusive scissor rectangles for one or more viewports. Fragments outside these rectangles are discarded when exclusive scissor testing is enabled.

Requires VK_NV_scissor_exclusive.

pub unsafe fn cmd_set_exclusive_scissor_enable_nv( &self, command_buffer: CommandBuffer, first_exclusive_scissor: u32, p_exclusive_scissor_enables: &[u32], )

Wraps vkCmdSetExclusiveScissorEnableNV. Provided by VK_NV_scissor_exclusive.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdSetExclusiveScissorEnableNV was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Dynamically enables or disables the exclusive scissor test for one or more viewports. When enabled, fragments outside the exclusive scissor rectangle are discarded.

Requires VK_NV_scissor_exclusive.

pub unsafe fn cmd_bind_shading_rate_image_nv( &self, command_buffer: CommandBuffer, image_view: ImageView, image_layout: ImageLayout, )

Wraps vkCmdBindShadingRateImageNV. Provided by VK_NV_shading_rate_image.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdBindShadingRateImageNV was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Binds a shading rate image that controls per-region fragment shading rate. Each texel in the image maps to a tile of the framebuffer and specifies the coarse shading rate for that tile.

Requires VK_NV_shading_rate_image.

pub unsafe fn cmd_set_viewport_shading_rate_palette_nv( &self, command_buffer: CommandBuffer, first_viewport: u32, p_shading_rate_palettes: &[ShadingRatePaletteNV], )

Wraps vkCmdSetViewportShadingRatePaletteNV. Provided by VK_NV_shading_rate_image.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdSetViewportShadingRatePaletteNV was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Sets the shading rate palette for one or more viewports. The palette maps shading rate image texel values to actual shading rates (e.g., 1x1, 2x2, 4x4).

Requires VK_NV_shading_rate_image.

pub unsafe fn cmd_set_coarse_sample_order_nv( &self, command_buffer: CommandBuffer, sample_order_type: CoarseSampleOrderTypeNV, p_custom_sample_orders: &[CoarseSampleOrderCustomNV], )

Wraps vkCmdSetCoarseSampleOrderNV. Provided by VK_NV_shading_rate_image.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdSetCoarseSampleOrderNV was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Sets the sample ordering for coarse shading rate fragments. By default, the driver chooses sample order; use this to specify a custom order when the fragment shader relies on specific sample positions within coarse fragments.

Requires VK_NV_shading_rate_image.

pub unsafe fn cmd_draw_mesh_tasks_nv( &self, command_buffer: CommandBuffer, task_count: u32, first_task: u32, )

Wraps vkCmdDrawMeshTasksNV. Provided by VK_NV_mesh_shader.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdDrawMeshTasksNV was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Dispatches mesh shader work using the NV mesh shader model. Launches task_count task shader workgroups starting at first_task.

This is the legacy NV path; prefer cmd_draw_mesh_tasks_ext for new code.

Requires VK_NV_mesh_shader.

pub unsafe fn cmd_draw_mesh_tasks_indirect_nv( &self, command_buffer: CommandBuffer, buffer: Buffer, offset: u64, draw_count: u32, stride: u32, )

Wraps vkCmdDrawMeshTasksIndirectNV. Provided by VK_NV_mesh_shader.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdDrawMeshTasksIndirectNV was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Indirect variant of cmd_draw_mesh_tasks_nv. Reads draw parameters from a buffer, enabling GPU-driven mesh shader dispatch.

Requires VK_NV_mesh_shader.

pub unsafe fn cmd_draw_mesh_tasks_indirect_count_nv( &self, command_buffer: CommandBuffer, buffer: Buffer, offset: u64, count_buffer: Buffer, count_buffer_offset: u64, max_draw_count: u32, stride: u32, )

Wraps vkCmdDrawMeshTasksIndirectCountNV. Provided by VK_NV_mesh_shader.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdDrawMeshTasksIndirectCountNV was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Indirect-count variant of cmd_draw_mesh_tasks_nv. Reads both the draw parameters and the draw count from GPU buffers, enabling fully GPU-driven mesh shader dispatch where the number of draws is determined at runtime.

Requires VK_NV_mesh_shader.

pub unsafe fn cmd_draw_mesh_tasks_ext( &self, command_buffer: CommandBuffer, group_count_x: u32, group_count_y: u32, group_count_z: u32, )

Wraps vkCmdDrawMeshTasksEXT. Provided by VK_EXT_mesh_shader.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdDrawMeshTasksEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Dispatches mesh shader work groups. group_count_x/y/z specify the number of task or mesh shader work groups to launch.

Each work group runs the mesh shader (or task shader, if bound) which emits primitives directly without traditional vertex/index buffers.

Requires VK_EXT_mesh_shader and the meshShader feature.

pub unsafe fn cmd_draw_mesh_tasks_indirect_ext( &self, command_buffer: CommandBuffer, buffer: Buffer, offset: u64, draw_count: u32, stride: u32, )

Wraps vkCmdDrawMeshTasksIndirectEXT. Provided by VK_EXT_mesh_shader.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdDrawMeshTasksIndirectEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Indirect version of cmd_draw_mesh_tasks_ext. Reads mesh shader dispatch parameters from a buffer. Each indirect command contains group counts (x, y, z).

draw_count specifies how many indirect commands to execute. stride is the byte stride between commands in the buffer.

Requires VK_EXT_mesh_shader.

pub unsafe fn cmd_draw_mesh_tasks_indirect_count_ext( &self, command_buffer: CommandBuffer, buffer: Buffer, offset: u64, count_buffer: Buffer, count_buffer_offset: u64, max_draw_count: u32, stride: u32, )

Wraps vkCmdDrawMeshTasksIndirectCountEXT. Provided by VK_EXT_mesh_shader.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdDrawMeshTasksIndirectCountEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Count-indirect version of mesh shader dispatch. Both the dispatch parameters and the draw count are read from GPU buffers, enabling fully GPU-driven mesh shader workloads.

max_draw_count caps the count read from the count buffer.

Requires VK_EXT_mesh_shader.

pub unsafe fn compile_deferred_nv( &self, pipeline: Pipeline, shader: u32, ) -> VkResult<()>

Wraps vkCompileDeferredNV. Provided by VK_NV_ray_tracing.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkCompileDeferredNV was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Triggers compilation of a deferred shader in an NV ray tracing pipeline. When a pipeline is created with PIPELINE_CREATE_DEFER_COMPILE_BIT_NV, individual shaders can be compiled on demand using this command.

Useful for spreading compilation cost across frames or threads.

Requires VK_NV_ray_tracing.

pub unsafe fn create_acceleration_structure_nv( &self, p_create_info: &AccelerationStructureCreateInfoNV, allocator: Option<&AllocationCallbacks>, ) -> VkResult<AccelerationStructureNV>

Wraps vkCreateAccelerationStructureNV. Provided by VK_NV_ray_tracing.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkCreateAccelerationStructureNV was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Creates an NV ray tracing acceleration structure. This is the legacy NV path; prefer create_acceleration_structure_khr for new code.

The NV acceleration structure owns its memory implicitly, bind memory with bind_acceleration_structure_memory_nv. Destroy with destroy_acceleration_structure_nv.

Requires VK_NV_ray_tracing.

pub unsafe fn cmd_bind_invocation_mask_huawei( &self, command_buffer: CommandBuffer, image_view: ImageView, image_layout: ImageLayout, )

Wraps vkCmdBindInvocationMaskHUAWEI. Provided by VK_HUAWEI_invocation_mask.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdBindInvocationMaskHUAWEI was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Binds an image view as an invocation mask for ray tracing. The mask selectively enables or disables ray generation shader invocations per pixel, allowing efficient partial-screen ray tracing on Huawei GPUs.

Requires VK_HUAWEI_invocation_mask.

pub unsafe fn destroy_acceleration_structure_khr( &self, acceleration_structure: AccelerationStructureKHR, allocator: Option<&AllocationCallbacks>, )

Wraps vkDestroyAccelerationStructureKHR. Provided by VK_KHR_acceleration_structure.

Safety

• device (self) must be valid and not destroyed.
• accelerationStructure must be externally synchronized.

Panics

Panics if vkDestroyAccelerationStructureKHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Destroys an acceleration structure. The structure must not be referenced by any pending ray tracing command or TLAS build.

Destroying the acceleration structure does not free the backing buffer, destroy or reclaim it separately.

pub unsafe fn destroy_acceleration_structure_nv( &self, acceleration_structure: AccelerationStructureNV, allocator: Option<&AllocationCallbacks>, )

Wraps vkDestroyAccelerationStructureNV. Provided by VK_NV_ray_tracing.

Safety

• device (self) must be valid and not destroyed.
• accelerationStructure must be externally synchronized.

Panics

Panics if vkDestroyAccelerationStructureNV was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Destroys an NV acceleration structure created with create_acceleration_structure_nv. The structure must not be referenced by any pending command.

Requires VK_NV_ray_tracing.

pub unsafe fn get_acceleration_structure_memory_requirements_nv( &self, p_info: &AccelerationStructureMemoryRequirementsInfoNV, ) -> MemoryRequirements2KHR

Wraps vkGetAccelerationStructureMemoryRequirementsNV. Provided by VK_NV_ray_tracing.

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetAccelerationStructureMemoryRequirementsNV was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Queries the memory requirements for an NV acceleration structure, including the scratch memory needed for builds and updates. Use the result to allocate and bind memory before building.

Requires VK_NV_ray_tracing.

pub unsafe fn bind_acceleration_structure_memory_nv( &self, p_bind_infos: &[BindAccelerationStructureMemoryInfoNV], ) -> VkResult<()>

Wraps vkBindAccelerationStructureMemoryNV. Provided by VK_NV_ray_tracing.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkBindAccelerationStructureMemoryNV was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Binds device memory to one or more NV acceleration structures. Must be called before the structure can be used in a build or trace command.

Requires VK_NV_ray_tracing.

pub unsafe fn cmd_copy_acceleration_structure_nv( &self, command_buffer: CommandBuffer, dst: AccelerationStructureNV, src: AccelerationStructureNV, mode: CopyAccelerationStructureModeKHR, )

Wraps vkCmdCopyAccelerationStructureNV. Provided by VK_NV_ray_tracing.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdCopyAccelerationStructureNV was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Copies or compacts an NV acceleration structure. Use CLONE mode for a full copy, or COMPACT to produce a smaller copy after querying the compacted size.

Requires VK_NV_ray_tracing.

pub unsafe fn cmd_copy_acceleration_structure_khr( &self, command_buffer: CommandBuffer, p_info: &CopyAccelerationStructureInfoKHR, )

Wraps vkCmdCopyAccelerationStructureKHR. Provided by VK_KHR_acceleration_structure.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdCopyAccelerationStructureKHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Copies an acceleration structure. Modes:

• COPY_ACCELERATION_STRUCTURE_MODE_CLONE: full copy. The destination must be at least as large as the source.
• COPY_ACCELERATION_STRUCTURE_MODE_COMPACT: copies into a smaller buffer after a compaction query. Use this to reclaim memory after building.

Compaction workflow:

1. Build with BUILD_ACCELERATION_STRUCTURE_ALLOW_COMPACTION.
2. Query compacted size with cmd_write_acceleration_structures_properties_khr.
3. Create a new, smaller backing buffer.
4. Copy with MODE_COMPACT.

Compaction typically saves 50–70% of BLAS memory.

pub unsafe fn copy_acceleration_structure_khr( &self, deferred_operation: DeferredOperationKHR, p_info: &CopyAccelerationStructureInfoKHR, ) -> VkResult<()>

Wraps vkCopyAccelerationStructureKHR. Provided by VK_KHR_acceleration_structure.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkCopyAccelerationStructureKHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Host-side acceleration structure copy. The CPU counterpart to cmd_copy_acceleration_structure_khr. Supports the same clone and compact modes.

Requires the acceleration_structure_host_commands feature.

pub unsafe fn cmd_copy_acceleration_structure_to_memory_khr( &self, command_buffer: CommandBuffer, p_info: &CopyAccelerationStructureToMemoryInfoKHR, )

Wraps vkCmdCopyAccelerationStructureToMemoryKHR. Provided by VK_KHR_acceleration_structure.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdCopyAccelerationStructureToMemoryKHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Serializes an acceleration structure into a buffer for storage or transfer. The serialized format is opaque and driver-specific, it can only be deserialized on the same driver and hardware.

Use this for:

• Saving acceleration structures to disk for faster subsequent loads.
• Transferring structures between devices in a device group.

Deserialize with cmd_copy_memory_to_acceleration_structure_khr.

pub unsafe fn copy_acceleration_structure_to_memory_khr( &self, deferred_operation: DeferredOperationKHR, p_info: &CopyAccelerationStructureToMemoryInfoKHR, ) -> VkResult<()>

Wraps vkCopyAccelerationStructureToMemoryKHR. Provided by VK_KHR_acceleration_structure.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkCopyAccelerationStructureToMemoryKHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Host-side acceleration structure serialization. The CPU counterpart to cmd_copy_acceleration_structure_to_memory_khr.

Requires the acceleration_structure_host_commands feature.

pub unsafe fn cmd_copy_memory_to_acceleration_structure_khr( &self, command_buffer: CommandBuffer, p_info: &CopyMemoryToAccelerationStructureInfoKHR, )

Wraps vkCmdCopyMemoryToAccelerationStructureKHR. Provided by VK_KHR_acceleration_structure.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdCopyMemoryToAccelerationStructureKHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Deserializes an acceleration structure from a buffer that was previously written by cmd_copy_acceleration_structure_to_memory_khr.

The data must have been serialized on the same driver and hardware (check acceleration_structure_uuid compatibility before loading).

After deserialization the acceleration structure is ready for use in ray tracing commands.

pub unsafe fn copy_memory_to_acceleration_structure_khr( &self, deferred_operation: DeferredOperationKHR, p_info: &CopyMemoryToAccelerationStructureInfoKHR, ) -> VkResult<()>

Wraps vkCopyMemoryToAccelerationStructureKHR. Provided by VK_KHR_acceleration_structure.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkCopyMemoryToAccelerationStructureKHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Host-side acceleration structure deserialization. The CPU counterpart to cmd_copy_memory_to_acceleration_structure_khr.

Requires the acceleration_structure_host_commands feature.

pub unsafe fn cmd_write_acceleration_structures_properties_khr( &self, command_buffer: CommandBuffer, p_acceleration_structures: &[AccelerationStructureKHR], query_type: QueryType, query_pool: QueryPool, first_query: u32, )

Wraps vkCmdWriteAccelerationStructuresPropertiesKHR. Provided by VK_KHR_acceleration_structure.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdWriteAccelerationStructuresPropertiesKHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Writes acceleration structure properties into a query pool. The primary use is querying QUERY_TYPE_ACCELERATION_STRUCTURE_COMPACTED_SIZE after a build to determine the compacted size before copying.

Compaction workflow:

1. Build with ALLOW_COMPACTION.
2. cmd_write_acceleration_structures_properties_khr with COMPACTED_SIZE query type.
3. Read the result from the query pool.
4. Create a smaller buffer and copy with MODE_COMPACT.

Also supports QUERY_TYPE_ACCELERATION_STRUCTURE_SERIALIZATION_SIZE for estimating serialization buffer requirements.

pub unsafe fn cmd_write_acceleration_structures_properties_nv( &self, command_buffer: CommandBuffer, p_acceleration_structures: &[AccelerationStructureNV], query_type: QueryType, query_pool: QueryPool, first_query: u32, )

Wraps vkCmdWriteAccelerationStructuresPropertiesNV. Provided by VK_NV_ray_tracing.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdWriteAccelerationStructuresPropertiesNV was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Writes acceleration structure properties (such as compacted size) into a query pool. Use this after a build to determine the compacted size before calling cmd_copy_acceleration_structure_nv with COMPACT mode.

Requires VK_NV_ray_tracing.

pub unsafe fn cmd_build_acceleration_structure_nv( &self, command_buffer: CommandBuffer, p_info: &AccelerationStructureInfoNV, instance_data: Buffer, instance_offset: u64, update: bool, dst: AccelerationStructureNV, src: AccelerationStructureNV, scratch: Buffer, scratch_offset: u64, )

Wraps vkCmdBuildAccelerationStructureNV. Provided by VK_NV_ray_tracing.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdBuildAccelerationStructureNV was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Builds or updates an NV acceleration structure from geometry data. Set update to non-zero to refit an existing structure in place (faster but lower quality than a full rebuild).

A scratch buffer is required; query its size with get_acceleration_structure_memory_requirements_nv.

Requires VK_NV_ray_tracing.

pub unsafe fn write_acceleration_structures_properties_khr( &self, p_acceleration_structures: &[AccelerationStructureKHR], query_type: QueryType, data_size: usize, p_data: *mut c_void, stride: usize, ) -> VkResult<()>

Wraps vkWriteAccelerationStructuresPropertiesKHR. Provided by VK_KHR_acceleration_structure.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkWriteAccelerationStructuresPropertiesKHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Host-side query of acceleration structure properties. The CPU counterpart to cmd_write_acceleration_structures_properties_khr.

Writes results directly to a host buffer rather than a query pool. Supports the same query types: compacted size and serialization size.

Requires the acceleration_structure_host_commands feature.

pub unsafe fn cmd_trace_rays_khr( &self, command_buffer: CommandBuffer, p_raygen_shader_binding_table: &StridedDeviceAddressRegionKHR, p_miss_shader_binding_table: &StridedDeviceAddressRegionKHR, p_hit_shader_binding_table: &StridedDeviceAddressRegionKHR, p_callable_shader_binding_table: &StridedDeviceAddressRegionKHR, width: u32, height: u32, depth: u32, )

Wraps vkCmdTraceRaysKHR. Provided by VK_KHR_ray_tracing_pipeline.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdTraceRaysKHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Dispatches rays into the scene. This is the primary ray tracing dispatch command, the ray tracing equivalent of cmd_draw or cmd_dispatch.

Each of the four shader binding table (SBT) regions points to a device memory region containing shader group handles:

• Raygen: exactly one entry, the ray generation shader.
• Miss: shaders invoked when a ray hits nothing.
• Hit: shader groups invoked on ray-geometry intersection.
• Callable: shaders invoked explicitly from other stages.

The width, height, and depth parameters define the 3D launch dimensions. Each invocation gets a unique gl_LaunchIDEXT. For a fullscreen ray trace, use the render target resolution with depth = 1.

The SBT entries must be built from handles retrieved with get_ray_tracing_shader_group_handles_khr, stored in a buffer with BUFFER_USAGE_SHADER_BINDING_TABLE. Each region’s stride must be a multiple of shaderGroupHandleAlignment and the base address must be aligned to shaderGroupBaseAlignment.

pub unsafe fn cmd_trace_rays_nv( &self, command_buffer: CommandBuffer, raygen_shader_binding_table_buffer: Buffer, raygen_shader_binding_offset: u64, miss_shader_binding_table_buffer: Buffer, miss_shader_binding_offset: u64, miss_shader_binding_stride: u64, hit_shader_binding_table_buffer: Buffer, hit_shader_binding_offset: u64, hit_shader_binding_stride: u64, callable_shader_binding_table_buffer: Buffer, callable_shader_binding_offset: u64, callable_shader_binding_stride: u64, width: u32, height: u32, depth: u32, )

Wraps vkCmdTraceRaysNV. Provided by VK_NV_ray_tracing.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdTraceRaysNV was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Dispatches a ray tracing workload using the NV ray tracing pipeline. Takes shader binding table buffer/offset/stride for each shader stage (raygen, miss, closest hit, callable) and the dispatch dimensions.

This is the legacy NV path; prefer cmd_trace_rays_khr for new code.

Requires VK_NV_ray_tracing.

pub unsafe fn get_ray_tracing_shader_group_handles_khr( &self, pipeline: Pipeline, first_group: u32, group_count: u32, data_size: usize, p_data: *mut c_void, ) -> VkResult<()>

Wraps vkGetRayTracingShaderGroupHandlesKHR. Provided by VK_KHR_ray_tracing_pipeline.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetRayTracingShaderGroupHandlesKHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Retrieves opaque shader group handles from a ray tracing pipeline. These handles are copied into GPU-visible buffers to build the shader binding table (SBT) that cmd_trace_rays_khr indexes into.

Each handle is shaderGroupHandleSize bytes (query from PhysicalDeviceRayTracingPipelinePropertiesKHR, typically 32 bytes). The p_data buffer must be at least group_count * shaderGroupHandleSize bytes.

first_group and group_count index into the groups array from RayTracingPipelineCreateInfoKHR. Handles are written sequentially, group first_group first, then first_group + 1, and so on.

After retrieving handles, copy them into a buffer with BUFFER_USAGE_SHADER_BINDING_TABLE at the correct stride and alignment for each SBT region.

pub unsafe fn get_ray_tracing_capture_replay_shader_group_handles_khr( &self, pipeline: Pipeline, first_group: u32, group_count: u32, data_size: usize, p_data: *mut c_void, ) -> VkResult<()>

Wraps vkGetRayTracingCaptureReplayShaderGroupHandlesKHR. Provided by VK_KHR_ray_tracing_pipeline.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetRayTracingCaptureReplayShaderGroupHandlesKHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Retrieves opaque capture/replay handles for shader groups. These handles allow recreating a ray tracing pipeline with identical shader group assignments on a subsequent run, enabling deterministic replay of GPU traces.

Use this for tools, profilers, and capture-replay frameworks. The handles are passed back via RayTracingShaderGroupCreateInfoKHR::shader_group_capture_replay_handle when recreating the pipeline.

The pipeline must have been created with PIPELINE_CREATE_RAY_TRACING_SHADER_GROUP_HANDLE_CAPTURE_REPLAY. Requires the rayTracingPipelineShaderGroupHandleCaptureReplay device feature.

The buffer layout is the same as get_ray_tracing_shader_group_handles_khr, sequential handles of shaderGroupHandleSize bytes each.

pub unsafe fn get_acceleration_structure_handle_nv( &self, acceleration_structure: AccelerationStructureNV, data_size: usize, p_data: *mut c_void, ) -> VkResult<()>

Wraps vkGetAccelerationStructureHandleNV. Provided by VK_NV_ray_tracing.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetAccelerationStructureHandleNV was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Retrieves an opaque handle for an NV acceleration structure. This handle is used when building a top-level acceleration structure that references bottom-level structures.

Requires VK_NV_ray_tracing.

pub unsafe fn create_ray_tracing_pipelines_nv( &self, pipeline_cache: PipelineCache, p_create_infos: &[RayTracingPipelineCreateInfoNV], allocator: Option<&AllocationCallbacks>, ) -> VkResult<Vec<Pipeline>>

Wraps vkCreateRayTracingPipelinesNV. Provided by VK_NV_ray_tracing.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_INVALID_SHADER_NV
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.
• pipelineCache must be externally synchronized.

Panics

Panics if vkCreateRayTracingPipelinesNV was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Creates one or more ray tracing pipelines using the NV model. This is the legacy NV path; prefer create_ray_tracing_pipelines_khr for new code.

Supports a pipeline cache for faster subsequent creation.

Requires VK_NV_ray_tracing.

pub unsafe fn create_ray_tracing_pipelines_khr( &self, deferred_operation: DeferredOperationKHR, pipeline_cache: PipelineCache, p_create_infos: &[RayTracingPipelineCreateInfoKHR], allocator: Option<&AllocationCallbacks>, ) -> VkResult<Vec<Pipeline>>

Wraps vkCreateRayTracingPipelinesKHR. Provided by VK_KHR_ray_tracing_pipeline.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_INVALID_OPAQUE_CAPTURE_ADDRESS
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.
• pipelineCache must be externally synchronized.

Panics

Panics if vkCreateRayTracingPipelinesKHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Creates one or more ray tracing pipelines. A ray tracing pipeline contains the shader stages (ray generation, miss, closest hit, any hit, intersection, callable) and shader groups that define how rays interact with geometry.

Unlike graphics pipelines, ray tracing pipelines organize shaders into groups:

• General: ray generation, miss, or callable shaders.
• Triangles hit: closest hit + optional any hit for triangles.
• Procedural hit: intersection + closest hit + optional any hit for custom geometry (AABBs).

Pass a DeferredOperationKHR handle to compile asynchronously, the call returns OPERATION_DEFERRED_KHR and the pipeline handles are not valid until the deferred operation completes. Pass a null handle for synchronous creation.

Supports pipeline_cache for faster creation on subsequent runs and base_pipeline_handle / base_pipeline_index for derivative pipelines when PIPELINE_CREATE_DERIVATIVE is set.

After creation, retrieve shader group handles with get_ray_tracing_shader_group_handles_khr to build the shader binding table.

pub unsafe fn cmd_trace_rays_indirect_khr( &self, command_buffer: CommandBuffer, p_raygen_shader_binding_table: &StridedDeviceAddressRegionKHR, p_miss_shader_binding_table: &StridedDeviceAddressRegionKHR, p_hit_shader_binding_table: &StridedDeviceAddressRegionKHR, p_callable_shader_binding_table: &StridedDeviceAddressRegionKHR, indirect_device_address: u64, )

Wraps vkCmdTraceRaysIndirectKHR. Provided by VK_KHR_ray_tracing_pipeline.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdTraceRaysIndirectKHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Dispatches rays with launch dimensions read from a GPU buffer. Identical to cmd_trace_rays_khr except the width, height, and depth are sourced from a TraceRaysIndirectCommandKHR struct at indirect_device_address.

This enables the GPU to determine ray dispatch dimensions without a CPU round-trip, useful when the dispatch size depends on prior GPU work such as culling, tile classification, or adaptive sampling.

The indirect buffer must have been created with BUFFER_USAGE_INDIRECT_BUFFER and the address must be aligned to 4 bytes. The SBT parameters are still provided directly on the CPU side.

Requires the rayTracingPipelineTraceRaysIndirect feature.

pub unsafe fn cmd_trace_rays_indirect2_khr( &self, command_buffer: CommandBuffer, indirect_device_address: u64, )

Wraps vkCmdTraceRaysIndirect2KHR. Provided by VK_KHR_ray_tracing_maintenance1.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdTraceRaysIndirect2KHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Fully indirect ray dispatch, both the shader binding table addresses and the launch dimensions are read from a GPU buffer. This is the most flexible ray tracing dispatch.

The indirect_device_address points to a TraceRaysIndirectCommand2KHR struct on the device, which contains all four SBT regions (raygen, miss, hit, callable) plus the width, height, and depth.

This allows the GPU to dynamically select which shaders to use and how many rays to launch, enabling advanced techniques like GPU-driven material sorting or multi-pass ray tracing without CPU synchronization.

Provided by VK_KHR_ray_tracing_maintenance1, not the base ray tracing pipeline extension. Requires the rayTracingPipelineTraceRaysIndirect2 feature.

pub unsafe fn get_cluster_acceleration_structure_build_sizes_nv( &self, p_info: &ClusterAccelerationStructureInputInfoNV, p_size_info: &mut AccelerationStructureBuildSizesInfoKHR, )

Wraps vkGetClusterAccelerationStructureBuildSizesNV. Provided by VK_NV_cluster_acceleration_structure.

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetClusterAccelerationStructureBuildSizesNV was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Queries the buffer sizes needed to build a cluster acceleration structure. Use the returned sizes to allocate the destination and scratch buffers before building.

Requires VK_NV_cluster_acceleration_structure.

pub unsafe fn cmd_build_cluster_acceleration_structure_indirect_nv( &self, command_buffer: CommandBuffer, p_command_infos: &ClusterAccelerationStructureCommandsInfoNV, )

Wraps vkCmdBuildClusterAccelerationStructureIndirectNV. Provided by VK_NV_cluster_acceleration_structure.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdBuildClusterAccelerationStructureIndirectNV was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Builds a cluster acceleration structure using indirect parameters. Cluster acceleration structures organize geometry into spatial clusters for more efficient ray traversal on NVIDIA hardware.

Requires VK_NV_cluster_acceleration_structure.

pub unsafe fn get_device_acceleration_structure_compatibility_khr( &self, p_version_info: &AccelerationStructureVersionInfoKHR, ) -> AccelerationStructureCompatibilityKHR

Wraps vkGetDeviceAccelerationStructureCompatibilityKHR. Provided by VK_KHR_acceleration_structure.

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetDeviceAccelerationStructureCompatibilityKHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Checks whether a serialized acceleration structure (from copy_acceleration_structure_to_memory_khr) is compatible with this device and can be deserialized.

Returns ACCELERATION_STRUCTURE_COMPATIBILITY_COMPATIBLE if the data can be loaded, or INCOMPATIBLE if not. Incompatibility typically means the data was serialized on different hardware or a different driver version.

Check compatibility before attempting deserialization to avoid errors.

pub unsafe fn get_ray_tracing_shader_group_stack_size_khr( &self, pipeline: Pipeline, group: u32, group_shader: ShaderGroupShaderKHR, )

Wraps vkGetRayTracingShaderGroupStackSizeKHR. Provided by VK_KHR_ray_tracing_pipeline.

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetRayTracingShaderGroupStackSizeKHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Queries the stack size contribution of a single shader within a shader group. The result is used to compute the total pipeline stack size for cmd_set_ray_tracing_pipeline_stack_size_khr.

group indexes into the shader groups array from pipeline creation. group_shader selects which shader within the group: GENERAL, CLOSEST_HIT, ANY_HIT, or INTERSECTION.

The default pipeline stack size is computed automatically at creation time, but it assumes worst-case recursion. If you know your actual maxPipelineRayRecursionDepth is lower, query individual stack sizes and compute a tighter total to reduce scratch memory usage.

Stack size computation formula (from spec):

raygen + max(closesthit + intersection, miss, callable) * maxRecursionDepth

Call this per-shader, aggregate across all groups, then set the result with cmd_set_ray_tracing_pipeline_stack_size_khr.

pub unsafe fn cmd_set_ray_tracing_pipeline_stack_size_khr( &self, command_buffer: CommandBuffer, pipeline_stack_size: u32, )

Wraps vkCmdSetRayTracingPipelineStackSizeKHR. Provided by VK_KHR_ray_tracing_pipeline.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdSetRayTracingPipelineStackSizeKHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Overrides the default ray tracing pipeline stack size for the bound pipeline. The stack is scratch memory used during shader execution and recursion.

The default stack size (set at pipeline creation) assumes worst-case recursion depth across all shader groups. If your application uses a lower effective recursion depth or only a subset of shader groups, setting a smaller stack size reduces per-invocation memory usage and may improve occupancy.

Compute the required size by querying individual shader contributions with get_ray_tracing_shader_group_stack_size_khr and applying the recursion formula from the spec.

This is a dynamic state command, it takes effect for subsequent cmd_trace_rays_khr calls within the same command buffer. Binding a new pipeline resets the stack size to the pipeline’s default.

pub unsafe fn get_image_view_handle_nvx(&self, p_info: &ImageViewHandleInfoNVX)

Wraps vkGetImageViewHandleNVX. Provided by VK_NVX_image_view_handle.

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetImageViewHandleNVX was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Returns a 32-bit handle for an image view that can be used as a bindless descriptor index. Use with get_image_view_address_nvx for fully bindless texture access.

Requires VK_NVX_image_view_handle.

pub unsafe fn get_image_view_handle64_nvx( &self, p_info: &ImageViewHandleInfoNVX, )

Wraps vkGetImageViewHandle64NVX. Provided by VK_NVX_image_view_handle.

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetImageViewHandle64NVX was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Returns a 64-bit handle for an image view. The 64-bit variant accommodates larger descriptor heaps than the 32-bit get_image_view_handle_nvx.

Requires VK_NVX_image_view_handle.

pub unsafe fn get_image_view_address_nvx( &self, image_view: ImageView, p_properties: &mut ImageViewAddressPropertiesNVX, ) -> VkResult<()>

Wraps vkGetImageViewAddressNVX. Provided by VK_NVX_image_view_handle.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetImageViewAddressNVX was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Queries the device address and size of an image view’s descriptor data. The address can be used for raw pointer-based descriptor access in shaders.

Requires VK_NVX_image_view_handle.

pub unsafe fn get_device_combined_image_sampler_index_nvx( &self, image_view_index: u64, sampler_index: u64, )

Wraps vkGetDeviceCombinedImageSamplerIndexNVX. Provided by VK_NVX_image_view_handle.

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetDeviceCombinedImageSamplerIndexNVX was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Returns the combined image-sampler descriptor index for a given image view and sampler pair. Used for bindless combined image- sampler access.

Requires VK_NVX_image_view_handle.

pub unsafe fn get_device_group_surface_present_modes2_ext( &self, p_surface_info: &PhysicalDeviceSurfaceInfo2KHR, ) -> VkResult<DeviceGroupPresentModeFlagsKHR>

Wraps vkGetDeviceGroupSurfacePresentModes2EXT. Provided by VK_EXT_full_screen_exclusive.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_SURFACE_LOST_KHR
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetDeviceGroupSurfacePresentModes2EXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Queries the supported present modes for a device group and surface, using the extended surface info structure. This is the VK_EXT_full_screen_exclusive variant of get_device_group_surface_present_modes_khr, allowing full-screen exclusive configuration to be factored into the query.

Requires VK_EXT_full_screen_exclusive. Windows only.

pub unsafe fn acquire_full_screen_exclusive_mode_ext( &self, swapchain: SwapchainKHR, ) -> VkResult<()>

Wraps vkAcquireFullScreenExclusiveModeEXT. Provided by VK_EXT_full_screen_exclusive.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_INITIALIZATION_FAILED
• VK_ERROR_SURFACE_LOST_KHR
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkAcquireFullScreenExclusiveModeEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Acquires full-screen exclusive mode for a swapchain, giving the application direct control over the display output. This can reduce latency and enable adaptive sync.

The swapchain must have been created with SurfaceFullScreenExclusiveInfoEXT in its pNext chain. Release with release_full_screen_exclusive_mode_ext.

Requires VK_EXT_full_screen_exclusive. Windows only.

pub unsafe fn release_full_screen_exclusive_mode_ext( &self, swapchain: SwapchainKHR, ) -> VkResult<()>

Wraps vkReleaseFullScreenExclusiveModeEXT. Provided by VK_EXT_full_screen_exclusive.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_SURFACE_LOST_KHR
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkReleaseFullScreenExclusiveModeEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Releases full-screen exclusive mode previously acquired with acquire_full_screen_exclusive_mode_ext. The swapchain returns to shared/composed presentation.

Requires VK_EXT_full_screen_exclusive. Windows only.

pub unsafe fn acquire_profiling_lock_khr( &self, p_info: &AcquireProfilingLockInfoKHR, ) -> VkResult<()>

Wraps vkAcquireProfilingLockKHR. Provided by VK_KHR_performance_query.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_TIMEOUT
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkAcquireProfilingLockKHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Acquires the device profiling lock, which must be held while submitting command buffers that contain performance queries. Only one thread can hold the lock at a time.

The AcquireProfilingLockInfoKHR specifies a timeout in nanoseconds. Returns TIMEOUT if the lock cannot be acquired within that period.

Release with release_profiling_lock_khr when profiling submission is complete.

Requires VK_KHR_performance_query.

pub unsafe fn release_profiling_lock_khr(&self)

Wraps vkReleaseProfilingLockKHR. Provided by VK_KHR_performance_query.

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkReleaseProfilingLockKHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Releases the device profiling lock previously acquired with acquire_profiling_lock_khr. Must be called after all command buffers containing performance queries have been submitted.

Requires VK_KHR_performance_query.

pub unsafe fn get_image_drm_format_modifier_properties_ext( &self, image: Image, p_properties: &mut ImageDrmFormatModifierPropertiesEXT, ) -> VkResult<()>

Wraps vkGetImageDrmFormatModifierPropertiesEXT. Provided by VK_EXT_image_drm_format_modifier.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetImageDrmFormatModifierPropertiesEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Queries which DRM format modifier was selected for an image created with VK_IMAGE_TILING_DRM_FORMAT_MODIFIER_EXT. The chosen modifier determines the memory layout and is needed when sharing the image with other DRM/KMS clients.

Requires VK_EXT_image_drm_format_modifier.

pub unsafe fn get_buffer_opaque_capture_address( &self, p_info: &BufferDeviceAddressInfo, )

Wraps vkGetBufferOpaqueCaptureAddress. Provided by VK_BASE_VERSION_1_2.

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetBufferOpaqueCaptureAddress was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Returns an opaque capture address for a buffer that was created with BUFFER_CREATE_DEVICE_ADDRESS_CAPTURE_REPLAY. This address is used to recreate the buffer at the same virtual address in a replay session (e.g. for GPU crash dump replay or deterministic replay tools).

Most applications do not need this, it is primarily for debugging and profiling tools. Use get_buffer_device_address for runtime buffer address access.

pub unsafe fn get_buffer_device_address(&self, p_info: &BufferDeviceAddressInfo)

Wraps vkGetBufferDeviceAddress. Provided by VK_BASE_VERSION_1_2.

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetBufferDeviceAddress was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Returns a 64-bit GPU virtual address for a buffer. This address can be passed to shaders via push constants or descriptors, enabling direct pointer-style access to buffer data from GPU code.

The buffer must have been created with BUFFER_USAGE_SHADER_DEVICE_ADDRESS and the buffer_device_address feature must be enabled.

Use cases:

• Bindless rendering: pass buffer addresses in a storage buffer or push constant instead of binding individual descriptors.
• Acceleration structures: ray tracing BLASes and TLASes reference geometry buffers by device address.
• GPU-driven pipelines: indirect command generators read vertex and index data by address.

The address remains valid for the lifetime of the buffer. If the buffer was created with BUFFER_CREATE_DEVICE_ADDRESS_CAPTURE_REPLAY, the address can be captured and replayed across sessions.

pub unsafe fn initialize_performance_api_intel( &self, p_initialize_info: &InitializePerformanceApiInfoINTEL, ) -> VkResult<()>

Wraps vkInitializePerformanceApiINTEL. Provided by VK_INTEL_performance_query.

Errors

• VK_ERROR_TOO_MANY_OBJECTS
• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkInitializePerformanceApiINTEL was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Initializes the Intel performance query API on a device. Must be called before using any other Intel performance query commands. Uninitialize with uninitialize_performance_api_intel.

Requires VK_INTEL_performance_query.

pub unsafe fn uninitialize_performance_api_intel(&self)

Wraps vkUninitializePerformanceApiINTEL. Provided by VK_INTEL_performance_query.

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkUninitializePerformanceApiINTEL was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Shuts down the Intel performance query API on a device, releasing any internal resources. Call when performance profiling is complete.

Requires VK_INTEL_performance_query.

pub unsafe fn cmd_set_performance_marker_intel( &self, command_buffer: CommandBuffer, p_marker_info: &PerformanceMarkerInfoINTEL, ) -> VkResult<()>

Wraps vkCmdSetPerformanceMarkerINTEL. Provided by VK_INTEL_performance_query.

Errors

• VK_ERROR_TOO_MANY_OBJECTS
• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdSetPerformanceMarkerINTEL was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Sets a performance marker in the command buffer to delimit a region of interest for Intel GPU profiling. Counters sampled between markers are attributed to the marked region.

Requires VK_INTEL_performance_query.

pub unsafe fn cmd_set_performance_stream_marker_intel( &self, command_buffer: CommandBuffer, p_marker_info: &PerformanceStreamMarkerInfoINTEL, ) -> VkResult<()>

Wraps vkCmdSetPerformanceStreamMarkerINTEL. Provided by VK_INTEL_performance_query.

Errors

• VK_ERROR_TOO_MANY_OBJECTS
• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdSetPerformanceStreamMarkerINTEL was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Sets a performance stream marker in the command buffer. Stream markers identify points in the command stream for correlating performance counter data with specific GPU work.

Requires VK_INTEL_performance_query.

pub unsafe fn cmd_set_performance_override_intel( &self, command_buffer: CommandBuffer, p_override_info: &PerformanceOverrideInfoINTEL, ) -> VkResult<()>

Wraps vkCmdSetPerformanceOverrideINTEL. Provided by VK_INTEL_performance_query.

Errors

• VK_ERROR_TOO_MANY_OBJECTS
• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdSetPerformanceOverrideINTEL was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Overrides hardware performance settings for the remainder of the command buffer (e.g., forcing specific EU thread counts). Used for controlled profiling experiments.

Requires VK_INTEL_performance_query.

pub unsafe fn acquire_performance_configuration_intel( &self, p_acquire_info: &PerformanceConfigurationAcquireInfoINTEL, ) -> VkResult<PerformanceConfigurationINTEL>

Wraps vkAcquirePerformanceConfigurationINTEL. Provided by VK_INTEL_performance_query.

Errors

• VK_ERROR_TOO_MANY_OBJECTS
• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkAcquirePerformanceConfigurationINTEL was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Acquires a performance configuration that enables specific hardware counters for profiling. Apply to a queue with queue_set_performance_configuration_intel. Release with release_performance_configuration_intel.

Requires VK_INTEL_performance_query.

pub unsafe fn release_performance_configuration_intel( &self, configuration: PerformanceConfigurationINTEL, ) -> VkResult<()>

Wraps vkReleasePerformanceConfigurationINTEL. Provided by VK_INTEL_performance_query.

Errors

• VK_ERROR_TOO_MANY_OBJECTS
• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.
• configuration must be externally synchronized.

Panics

Panics if vkReleasePerformanceConfigurationINTEL was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Releases a performance configuration acquired with acquire_performance_configuration_intel.

Requires VK_INTEL_performance_query.

pub unsafe fn queue_set_performance_configuration_intel( &self, queue: Queue, configuration: PerformanceConfigurationINTEL, ) -> VkResult<()>

Wraps vkQueueSetPerformanceConfigurationINTEL. Provided by VK_INTEL_performance_query.

Errors

• VK_ERROR_TOO_MANY_OBJECTS
• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• queue (self) must be valid and not destroyed.
• queue must be externally synchronized.

Panics

Panics if vkQueueSetPerformanceConfigurationINTEL was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Applies a performance configuration to a queue, enabling the selected hardware counters for all subsequent submissions to that queue.

Requires VK_INTEL_performance_query.

pub unsafe fn get_performance_parameter_intel( &self, parameter: PerformanceParameterTypeINTEL, ) -> VkResult<PerformanceValueINTEL>

Wraps vkGetPerformanceParameterINTEL. Provided by VK_INTEL_performance_query.

Errors

• VK_ERROR_TOO_MANY_OBJECTS
• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetPerformanceParameterINTEL was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Queries a performance parameter value from the Intel driver, such as GPU clock frequency or EU count. Useful for normalizing profiling results.

Requires VK_INTEL_performance_query.

pub unsafe fn get_device_memory_opaque_capture_address( &self, p_info: &DeviceMemoryOpaqueCaptureAddressInfo, )

Wraps vkGetDeviceMemoryOpaqueCaptureAddress. Provided by VK_BASE_VERSION_1_2.

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetDeviceMemoryOpaqueCaptureAddress was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Returns an opaque capture address for a device memory allocation that was created with MEMORY_ALLOCATE_DEVICE_ADDRESS_CAPTURE_REPLAY. Used in conjunction with get_buffer_opaque_capture_address to replay buffer address assignments.

This is a debugging/replay tool feature. Most applications do not need this.

pub unsafe fn get_pipeline_executable_properties_khr( &self, p_pipeline_info: &PipelineInfoKHR, ) -> VkResult<Vec<PipelineExecutablePropertiesKHR>>

Wraps vkGetPipelineExecutablePropertiesKHR. Provided by VK_KHR_pipeline_executable_properties.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetPipelineExecutablePropertiesKHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Lists the executable components within a pipeline. A single pipeline may contain multiple executables, for example, a graphics pipeline typically has separate vertex and fragment shader executables.

Each returned PipelineExecutablePropertiesKHR contains a name, description, and shader stage flags identifying the executable. Use these to index into get_pipeline_executable_statistics_khr and get_pipeline_executable_internal_representations_khr.

The pipeline must have been created with PIPELINE_CREATE_CAPTURE_STATISTICS or PIPELINE_CREATE_CAPTURE_INTERNAL_REPRESENTATIONS. This is a debugging and profiling tool, not intended for shipping builds.

pub unsafe fn get_pipeline_executable_statistics_khr( &self, p_executable_info: &PipelineExecutableInfoKHR, ) -> VkResult<Vec<PipelineExecutableStatisticKHR>>

Wraps vkGetPipelineExecutableStatisticsKHR. Provided by VK_KHR_pipeline_executable_properties.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetPipelineExecutableStatisticsKHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Retrieves compiler statistics for a specific pipeline executable. Statistics include metrics like register usage, instruction count, scratch memory, and other driver-specific values.

Identify the executable by index from get_pipeline_executable_properties_khr via PipelineExecutableInfoKHR.

Each statistic has a name, description, format (bool, int, float, or string), and value. The available statistics are driver-specific, different vendors report different metrics.

The pipeline must have been created with PIPELINE_CREATE_CAPTURE_STATISTICS. This is a profiling tool for shader optimization, use it to compare register pressure or instruction counts across shader variants.

pub unsafe fn get_pipeline_executable_internal_representations_khr( &self, p_executable_info: &PipelineExecutableInfoKHR, ) -> VkResult<Vec<PipelineExecutableInternalRepresentationKHR>>

Wraps vkGetPipelineExecutableInternalRepresentationsKHR. Provided by VK_KHR_pipeline_executable_properties.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetPipelineExecutableInternalRepresentationsKHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Retrieves internal representations (IR) of a pipeline executable. These are driver-specific intermediate or final shader representations, for example, SPIR-V, vendor IR, or GPU ISA disassembly.

Each representation has a name, description, and opaque data blob. Whether the data is human-readable text or binary depends on is_text in the returned structure.

The pipeline must have been created with PIPELINE_CREATE_CAPTURE_INTERNAL_REPRESENTATIONS. Enabling this flag may disable optimizations, so only use it for debugging and shader analysis, not in production.

Identify the executable by index from get_pipeline_executable_properties_khr.

pub unsafe fn cmd_set_line_stipple( &self, command_buffer: CommandBuffer, line_stipple_factor: u32, line_stipple_pattern: u16, )

Wraps vkCmdSetLineStipple. Provided by VK_GRAPHICS_VERSION_1_4.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdSetLineStipple was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Dynamically sets the line stipple pattern and repeat factor. Only takes effect if the pipeline was created with DYNAMIC_STATE_LINE_STIPPLE.

The stipple_factor (1–256) controls how many pixels each bit of the pattern spans. The stipple_pattern is a 16-bit bitmask where each bit represents a pixel, 1 is drawn, 0 is discarded.

Line stippling requires VK_EXT_line_rasterization and the stippled_*_lines device features, depending on which line rasterisation mode you use.

Core dynamic state in Vulkan 1.4.

pub unsafe fn get_fault_data( &self, fault_query_behavior: FaultQueryBehavior, p_unrecorded_faults: *mut u32, ) -> VkResult<Vec<FaultData>>

Wraps vkGetFaultData. Provided by VKSC_VERSION_1_0.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetFaultData was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Queries recorded fault data from the device. Part of Vulkan SC (Safety Critical) for fault reporting in safety-certified environments. Uses the two-call idiom. The fault_query_behavior controls whether queried faults are cleared. Also reports the count of unrecorded faults that overflowed the internal buffer.

Requires Vulkan SC.

pub unsafe fn create_acceleration_structure_khr( &self, p_create_info: &AccelerationStructureCreateInfoKHR, allocator: Option<&AllocationCallbacks>, ) -> VkResult<AccelerationStructureKHR>

Wraps vkCreateAccelerationStructureKHR. Provided by VK_KHR_acceleration_structure.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_INVALID_OPAQUE_CAPTURE_ADDRESS_KHR
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkCreateAccelerationStructureKHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Creates an acceleration structure for hardware ray tracing. An acceleration structure is a spatial data structure (typically a BVH) that the GPU traverses during ray intersection tests.

Two levels:

• Bottom-level (BLAS): contains geometry (triangles or AABBs). Create one per mesh or mesh group.
• Top-level (TLAS): contains instances that reference BLASes with per-instance transforms. Create one per scene.

The acceleration structure needs a backing buffer created with BUFFER_USAGE_ACCELERATION_STRUCTURE_STORAGE. Query the required size with get_acceleration_structure_build_sizes_khr first.

After creation, build the structure with cmd_build_acceleration_structures_khr. The structure is not usable for tracing until built.

pub unsafe fn cmd_build_acceleration_structures_khr( &self, command_buffer: CommandBuffer, p_infos: &[AccelerationStructureBuildGeometryInfoKHR], pp_build_range_infos: *const *const AccelerationStructureBuildRangeInfoKHR, )

Wraps vkCmdBuildAccelerationStructuresKHR. Provided by VK_KHR_acceleration_structure.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdBuildAccelerationStructuresKHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Records a GPU-side acceleration structure build or update. This is the primary way to build BLASes and TLASes for ray tracing.

Build vs update: an initial build creates the structure from scratch. An update (BUILD_ACCELERATION_STRUCTURE_MODE_UPDATE) modifies an existing structure in-place, which is faster but produces lower traversal quality. Use updates for dynamic geometry (e.g. animated characters) and full rebuilds when geometry changes significantly.

Scratch buffer: builds require a temporary scratch buffer. Query the required size with get_acceleration_structure_build_sizes_khr and create a buffer with BUFFER_USAGE_STORAGE_BUFFER.

Multiple builds can be batched in a single call. The driver may execute them in parallel.

Must be recorded outside a render pass.

pub unsafe fn cmd_build_acceleration_structures_indirect_khr( &self, command_buffer: CommandBuffer, p_infos: &[AccelerationStructureBuildGeometryInfoKHR], p_indirect_device_addresses: &u64, p_indirect_strides: *const u32, pp_max_primitive_counts: *const *const u32, )

Wraps vkCmdBuildAccelerationStructuresIndirectKHR. Provided by VK_KHR_acceleration_structure.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdBuildAccelerationStructuresIndirectKHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

GPU-side acceleration structure build with indirect parameters. The primitive counts and build ranges are read from GPU buffers rather than specified on the CPU.

This enables fully GPU-driven scene management where a compute shader determines which geometry to include and writes the build parameters.

Requires the acceleration_structure_indirect_build feature.

pub unsafe fn build_acceleration_structures_khr( &self, deferred_operation: DeferredOperationKHR, p_infos: &[AccelerationStructureBuildGeometryInfoKHR], pp_build_range_infos: *const *const AccelerationStructureBuildRangeInfoKHR, ) -> VkResult<()>

Wraps vkBuildAccelerationStructuresKHR. Provided by VK_KHR_acceleration_structure.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkBuildAccelerationStructuresKHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Builds or updates acceleration structures on the host (CPU). This is the CPU-side alternative to cmd_build_acceleration_structures_khr.

Host builds are useful for offline processing, tools, or when GPU build capacity is limited. However, GPU builds are significantly faster for real-time applications.

Requires the acceleration_structure_host_commands feature.

pub unsafe fn get_acceleration_structure_device_address_khr( &self, p_info: &AccelerationStructureDeviceAddressInfoKHR, )

Wraps vkGetAccelerationStructureDeviceAddressKHR. Provided by VK_KHR_acceleration_structure.

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetAccelerationStructureDeviceAddressKHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Returns the GPU device address of an acceleration structure. This address is used when building a TLAS, each instance in the TLAS references a BLAS by its device address.

The address remains valid for the lifetime of the acceleration structure.

pub unsafe fn create_deferred_operation_khr( &self, allocator: Option<&AllocationCallbacks>, ) -> VkResult<DeferredOperationKHR>

Wraps vkCreateDeferredOperationKHR. Provided by VK_KHR_deferred_host_operations.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkCreateDeferredOperationKHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Creates a deferred operation handle. Deferred operations allow expensive host-side work (such as ray tracing pipeline compilation) to be split across multiple CPU threads.

The typical workflow:

1. Create a deferred operation with this command.
2. Pass the handle to a deferrable command (e.g., create_ray_tracing_pipelines_khr). If deferred, it returns OPERATION_DEFERRED_KHR.
3. Query get_deferred_operation_max_concurrency_khr to learn how many threads can contribute.
4. Call deferred_operation_join_khr from each worker thread.
5. Once all joins return SUCCESS, retrieve the result with get_deferred_operation_result_khr.
6. Destroy the handle with destroy_deferred_operation_khr.

The handle itself is lightweight, it is just a token for tracking the deferred work.

pub unsafe fn destroy_deferred_operation_khr( &self, operation: DeferredOperationKHR, allocator: Option<&AllocationCallbacks>, )

Wraps vkDestroyDeferredOperationKHR. Provided by VK_KHR_deferred_host_operations.

Safety

• device (self) must be valid and not destroyed.
• operation must be externally synchronized.

Panics

Panics if vkDestroyDeferredOperationKHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Destroys a deferred operation handle. The operation must have completed before destruction, either all joining threads returned SUCCESS or THREAD_DONE_KHR, or the operation was never deferred.

Do not destroy while threads are still joined to the operation.

pub unsafe fn get_deferred_operation_max_concurrency_khr( &self, operation: DeferredOperationKHR, )

Wraps vkGetDeferredOperationMaxConcurrencyKHR. Provided by VK_KHR_deferred_host_operations.

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetDeferredOperationMaxConcurrencyKHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Returns the maximum number of threads that can usefully join this deferred operation. Spawning more threads than this value wastes resources, additional joins will return THREAD_IDLE_KHR.

The returned value may decrease over time as work completes, so query it just before spawning worker threads.

A return value of zero means the operation is already complete or requires no additional threads.

pub unsafe fn get_deferred_operation_result_khr( &self, operation: DeferredOperationKHR, ) -> VkResult<()>

Wraps vkGetDeferredOperationResultKHR. Provided by VK_KHR_deferred_host_operations.

Errors

• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetDeferredOperationResultKHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Returns the result of a completed deferred operation. The returned VkResult is the same value that the original deferrable command would have returned if it had executed synchronously.

Only call this after the operation has fully completed (all joins returned SUCCESS or THREAD_DONE_KHR). Calling on an in-progress operation returns NOT_READY.

For example, if create_ray_tracing_pipelines_khr was deferred, this returns whether pipeline creation succeeded or failed.

pub unsafe fn deferred_operation_join_khr( &self, operation: DeferredOperationKHR, ) -> VkResult<()>

Wraps vkDeferredOperationJoinKHR. Provided by VK_KHR_deferred_host_operations.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkDeferredOperationJoinKHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Joins the calling thread to a deferred operation, contributing CPU time to its completion. Multiple threads can join the same operation concurrently.

Return values:

• SUCCESS, the operation completed. The calling thread was the last one needed.
• THREAD_DONE_KHR, this thread’s contribution is finished, but the operation may still be in progress on other threads.
• THREAD_IDLE_KHR, the operation has enough threads; this one was not needed. Retry later or move on.

Call this in a loop per thread until it returns SUCCESS or THREAD_DONE_KHR. After all threads finish, check the final result with get_deferred_operation_result_khr.

The number of useful threads is bounded by get_deferred_operation_max_concurrency_khr.

pub unsafe fn get_pipeline_indirect_memory_requirements_nv( &self, p_create_info: &ComputePipelineCreateInfo, p_memory_requirements: &mut MemoryRequirements2, )

Wraps vkGetPipelineIndirectMemoryRequirementsNV. Provided by VK_NV_device_generated_commands_compute.

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetPipelineIndirectMemoryRequirementsNV was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Queries the memory requirements for storing a compute pipeline’s indirect dispatch metadata. Allocate a buffer of this size and pass it when creating the pipeline for device-generated compute dispatch.

Requires VK_NV_device_generated_commands_compute.

pub unsafe fn get_pipeline_indirect_device_address_nv( &self, p_info: &PipelineIndirectDeviceAddressInfoNV, )

Wraps vkGetPipelineIndirectDeviceAddressNV. Provided by VK_NV_device_generated_commands_compute.

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetPipelineIndirectDeviceAddressNV was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Returns the device address of a compute pipeline’s indirect dispatch metadata. The GPU writes to this address to select which pipeline to dispatch in device-generated compute workflows.

Requires VK_NV_device_generated_commands_compute.

pub unsafe fn anti_lag_update_amd(&self, p_data: &AntiLagDataAMD)

Wraps vkAntiLagUpdateAMD. Provided by VK_AMD_anti_lag.

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkAntiLagUpdateAMD was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Submits anti-lag timing data to reduce input-to-display latency. Called once per frame with presentation timing hints so the driver can pace GPU work to minimise latency.

Requires VK_AMD_anti_lag.

pub unsafe fn cmd_set_cull_mode( &self, command_buffer: CommandBuffer, cull_mode: CullModeFlags, )

Wraps vkCmdSetCullMode. Provided by VK_GRAPHICS_VERSION_1_3.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdSetCullMode was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Dynamically sets the triangle culling mode. Only takes effect if the pipeline was created with DYNAMIC_STATE_CULL_MODE.

Values: CULL_MODE_NONE, CULL_MODE_FRONT, CULL_MODE_BACK, CULL_MODE_FRONT_AND_BACK.

Common pattern: set CULL_MODE_BACK for opaque geometry and CULL_MODE_NONE for double-sided or transparent materials, without needing separate pipelines.

Core dynamic state in Vulkan 1.3.

pub unsafe fn cmd_set_front_face( &self, command_buffer: CommandBuffer, front_face: FrontFace, )

Wraps vkCmdSetFrontFace. Provided by VK_GRAPHICS_VERSION_1_3.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdSetFrontFace was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Dynamically sets which triangle winding order is considered front-facing. Only takes effect if the pipeline was created with DYNAMIC_STATE_FRONT_FACE.

Values: FRONT_FACE_COUNTER_CLOCKWISE (the Vulkan default) or FRONT_FACE_CLOCKWISE.

Useful when rendering mirrored or reflected geometry where the winding order is flipped, without needing a separate pipeline.

Core dynamic state in Vulkan 1.3.

pub unsafe fn cmd_set_primitive_topology( &self, command_buffer: CommandBuffer, primitive_topology: PrimitiveTopology, )

Wraps vkCmdSetPrimitiveTopology. Provided by VK_GRAPHICS_VERSION_1_3.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdSetPrimitiveTopology was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Dynamically sets the primitive topology. Only takes effect if the pipeline was created with DYNAMIC_STATE_PRIMITIVE_TOPOLOGY.

Values include POINT_LIST, LINE_LIST, LINE_STRIP, TRIANGLE_LIST, TRIANGLE_STRIP, TRIANGLE_FAN, LINE_LIST_WITH_ADJACENCY, PATCH_LIST, etc.

The dynamic topology must be in the same topology class as the pipeline’s static topology (e.g. you can switch between TRIANGLE_LIST and TRIANGLE_STRIP since both are triangle topologies, but not between TRIANGLE_LIST and LINE_LIST).

Core dynamic state in Vulkan 1.3.

pub unsafe fn cmd_set_viewport_with_count( &self, command_buffer: CommandBuffer, p_viewports: &[Viewport], )

Wraps vkCmdSetViewportWithCount. Provided by VK_GRAPHICS_VERSION_1_3.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdSetViewportWithCount was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Dynamically sets both the viewports and the viewport count. Only takes effect if the pipeline was created with DYNAMIC_STATE_VIEWPORT_WITH_COUNT.

Unlike cmd_set_viewport (which requires the count to match the pipeline’s static viewport_count), this command also sets the count dynamically. The viewport count must match the scissor count set by cmd_set_scissor_with_count.

Core dynamic state in Vulkan 1.3.

pub unsafe fn cmd_set_scissor_with_count( &self, command_buffer: CommandBuffer, p_scissors: &[Rect2D], )

Wraps vkCmdSetScissorWithCount. Provided by VK_GRAPHICS_VERSION_1_3.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdSetScissorWithCount was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Dynamically sets both the scissor rectangles and the scissor count. Only takes effect if the pipeline was created with DYNAMIC_STATE_SCISSOR_WITH_COUNT.

Unlike cmd_set_scissor (which requires the count to match the pipeline’s static viewport_count), this command also sets the count dynamically. The scissor count must match the viewport count set by cmd_set_viewport_with_count.

Core dynamic state in Vulkan 1.3.

pub unsafe fn cmd_bind_index_buffer2( &self, command_buffer: CommandBuffer, buffer: Buffer, offset: u64, size: u64, index_type: IndexType, )

Wraps vkCmdBindIndexBuffer2. Provided by VK_GRAPHICS_VERSION_1_4.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdBindIndexBuffer2 was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Vulkan 1.4 version of cmd_bind_index_buffer that additionally accepts a size parameter specifying the valid range of the index buffer. This enables the driver to perform bounds checking.

Pass VK_WHOLE_SIZE for the size to use the remainder of the buffer from the offset.

Prefer this over cmd_bind_index_buffer when targeting Vulkan 1.4+.

pub unsafe fn cmd_bind_vertex_buffers2( &self, command_buffer: CommandBuffer, first_binding: u32, p_buffers: &[Buffer], p_offsets: &u64, p_sizes: Option<&u64>, p_strides: Option<&u64>, )

Wraps vkCmdBindVertexBuffers2. Provided by VK_GRAPHICS_VERSION_1_3.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdBindVertexBuffers2 was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Vulkan 1.3 version of cmd_bind_vertex_buffers that additionally accepts optional buffer sizes and strides.

Sizes: if provided, the driver knows where each buffer ends and can perform bounds checking. Pass null to use the full buffer size.

Strides: if provided, overrides the stride specified in the pipeline’s vertex input state. This enables dynamic vertex stride without creating separate pipeline permutations. Pass null to use the pipeline’s static stride.

Prefer this over cmd_bind_vertex_buffers when targeting Vulkan 1.3+.

pub unsafe fn cmd_set_depth_test_enable( &self, command_buffer: CommandBuffer, depth_test_enable: bool, )

Wraps vkCmdSetDepthTestEnable. Provided by VK_GRAPHICS_VERSION_1_3.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdSetDepthTestEnable was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Dynamically enables or disables depth testing. Only takes effect if the pipeline was created with DYNAMIC_STATE_DEPTH_TEST_ENABLE.

When disabled, all fragments pass the depth test regardless of the depth buffer contents. Useful for UI overlays, skyboxes, or full-screen post-processing passes.

Core dynamic state in Vulkan 1.3.

pub unsafe fn cmd_set_depth_write_enable( &self, command_buffer: CommandBuffer, depth_write_enable: bool, )

Wraps vkCmdSetDepthWriteEnable. Provided by VK_GRAPHICS_VERSION_1_3.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdSetDepthWriteEnable was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Dynamically enables or disables writes to the depth buffer. Only takes effect if the pipeline was created with DYNAMIC_STATE_DEPTH_WRITE_ENABLE.

Disable depth writes when:

• Drawing transparent objects (they should test against depth but not write to it).
• Drawing skyboxes after the opaque pass.
• Performing post-processing with depth reads but no depth output.

Depth testing and depth writing are independent controls, you can test without writing, or write without testing.

Core dynamic state in Vulkan 1.3.

pub unsafe fn cmd_set_depth_compare_op( &self, command_buffer: CommandBuffer, depth_compare_op: CompareOp, )

Wraps vkCmdSetDepthCompareOp. Provided by VK_GRAPHICS_VERSION_1_3.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdSetDepthCompareOp was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Dynamically sets the depth comparison operator. Only takes effect if the pipeline was created with DYNAMIC_STATE_DEPTH_COMPARE_OP.

Values: COMPARE_OP_LESS (the common default for forward rendering), COMPARE_OP_GREATER (for reverse-Z), COMPARE_OP_LESS_OR_EQUAL, COMPARE_OP_ALWAYS, etc.

Reverse-Z: using a reversed depth buffer (near=1.0, far=0.0) with COMPARE_OP_GREATER provides better floating-point precision distribution across the depth range. This is the recommended setup for modern rendering.

Core dynamic state in Vulkan 1.3.

pub unsafe fn cmd_set_depth_bounds_test_enable( &self, command_buffer: CommandBuffer, depth_bounds_test_enable: bool, )

Wraps vkCmdSetDepthBoundsTestEnable. Provided by VK_GRAPHICS_VERSION_1_3.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdSetDepthBoundsTestEnable was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Dynamically enables or disables the depth bounds test. Only takes effect if the pipeline was created with DYNAMIC_STATE_DEPTH_BOUNDS_TEST_ENABLE.

When enabled, fragments are discarded if the existing depth buffer value falls outside the range set by cmd_set_depth_bounds. When disabled, the test is skipped.

Requires the depth_bounds device feature.

Core dynamic state in Vulkan 1.3.

pub unsafe fn cmd_set_stencil_test_enable( &self, command_buffer: CommandBuffer, stencil_test_enable: bool, )

Wraps vkCmdSetStencilTestEnable. Provided by VK_GRAPHICS_VERSION_1_3.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdSetStencilTestEnable was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Dynamically enables or disables stencil testing. Only takes effect if the pipeline was created with DYNAMIC_STATE_STENCIL_TEST_ENABLE.

When disabled, fragments pass the stencil test unconditionally and no stencil writes occur.

Core dynamic state in Vulkan 1.3.

pub unsafe fn cmd_set_stencil_op( &self, command_buffer: CommandBuffer, face_mask: StencilFaceFlags, fail_op: StencilOp, pass_op: StencilOp, depth_fail_op: StencilOp, compare_op: CompareOp, )

Wraps vkCmdSetStencilOp. Provided by VK_GRAPHICS_VERSION_1_3.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdSetStencilOp was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Dynamically sets the stencil operations for front-facing, back-facing, or both face sets. Only takes effect if the pipeline was created with DYNAMIC_STATE_STENCIL_OP.

Sets four values per face:

• fail_op: action when the stencil test fails.
• pass_op: action when both stencil and depth tests pass.
• depth_fail_op: action when stencil passes but depth fails.
• compare_op: the stencil comparison function.

Common operations: KEEP, REPLACE, INCREMENT_AND_CLAMP, DECREMENT_AND_CLAMP, INVERT, ZERO.

Core dynamic state in Vulkan 1.3.

pub unsafe fn cmd_set_patch_control_points_ext( &self, command_buffer: CommandBuffer, patch_control_points: u32, )

Wraps vkCmdSetPatchControlPointsEXT. Provided by VK_EXT_extended_dynamic_state2.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdSetPatchControlPointsEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Dynamically sets the number of control points per patch for tessellation. Overrides the value specified at pipeline creation.

Typical values: 3 (triangles), 4 (quads), 16 (bicubic patches). The maximum is maxTessellationPatchSize (at least 32).

Requires the extendedDynamicState2PatchControlPoints feature.

Provided by VK_EXT_extended_dynamic_state2.

pub unsafe fn cmd_set_rasterizer_discard_enable( &self, command_buffer: CommandBuffer, rasterizer_discard_enable: bool, )

Wraps vkCmdSetRasterizerDiscardEnable. Provided by VK_GRAPHICS_VERSION_1_3.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdSetRasterizerDiscardEnable was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Dynamically enables or disables rasterizer discard. Only takes effect if the pipeline was created with DYNAMIC_STATE_RASTERIZER_DISCARD_ENABLE.

When enabled, primitives are discarded before rasterisation, no fragments are generated and no colour/depth output is produced. The vertex and geometry shader stages still execute.

Use cases:

• Transform feedback only: capture transformed vertices without rendering.
• Occlusion pre-pass: skip fragment shading when only the depth or stencil output matters (though depth writes still require rasterisation).

Core dynamic state in Vulkan 1.3.

pub unsafe fn cmd_set_depth_bias_enable( &self, command_buffer: CommandBuffer, depth_bias_enable: bool, )

Wraps vkCmdSetDepthBiasEnable. Provided by VK_GRAPHICS_VERSION_1_3.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdSetDepthBiasEnable was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Dynamically enables or disables depth bias. Only takes effect if the pipeline was created with DYNAMIC_STATE_DEPTH_BIAS_ENABLE.

When enabled, the depth bias values set by cmd_set_depth_bias are applied to fragment depth values. When disabled, no bias is applied regardless of the bias values.

Useful for toggling shadow map bias without separate pipelines.

Core dynamic state in Vulkan 1.3.

pub unsafe fn cmd_set_logic_op_ext( &self, command_buffer: CommandBuffer, logic_op: LogicOp, )

Wraps vkCmdSetLogicOpEXT. Provided by VK_EXT_extended_dynamic_state2.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdSetLogicOpEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Dynamically sets the logic operation used for color blending. Logic ops (AND, OR, XOR, etc.) operate on the raw integer bit patterns of the fragment and framebuffer values.

Only effective when logic op is enabled in the pipeline (logicOpEnable). Requires the extendedDynamicState2LogicOp feature.

Provided by VK_EXT_extended_dynamic_state2.

pub unsafe fn cmd_set_primitive_restart_enable( &self, command_buffer: CommandBuffer, primitive_restart_enable: bool, )

Wraps vkCmdSetPrimitiveRestartEnable. Provided by VK_GRAPHICS_VERSION_1_3.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdSetPrimitiveRestartEnable was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Dynamically enables or disables primitive restart. Only takes effect if the pipeline was created with DYNAMIC_STATE_PRIMITIVE_RESTART_ENABLE.

When enabled, a special index value (0xFFFF for UINT16, 0xFFFFFFFF for UINT32) in the index buffer ends the current primitive and starts a new one. This lets you draw multiple triangle strips or line strips from a single draw call without degenerate triangles.

Only meaningful for strip topologies (TRIANGLE_STRIP, LINE_STRIP, TRIANGLE_FAN, etc.).

Core dynamic state in Vulkan 1.3.

pub unsafe fn cmd_set_tessellation_domain_origin_ext( &self, command_buffer: CommandBuffer, domain_origin: TessellationDomainOrigin, )

Wraps vkCmdSetTessellationDomainOriginEXT. Provided by VK_EXT_extended_dynamic_state3.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdSetTessellationDomainOriginEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Dynamically sets the tessellation domain origin:

• UPPER_LEFT: Vulkan default, (0,0) is the upper-left corner.
• LOWER_LEFT: OpenGL convention, (0,0) is the lower-left.

Affects how tessellation coordinates are interpreted. Useful when porting OpenGL tessellation shaders.

Provided by VK_EXT_extended_dynamic_state3.

pub unsafe fn cmd_set_depth_clamp_enable_ext( &self, command_buffer: CommandBuffer, depth_clamp_enable: bool, )

Wraps vkCmdSetDepthClampEnableEXT. Provided by VK_EXT_extended_dynamic_state3.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdSetDepthClampEnableEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Dynamically enables or disables depth clamping. When enabled, fragments outside the near/far depth range are clamped instead of being clipped.

Useful for shadow mapping and other techniques where clipping at the near/far plane is undesirable.

Requires the depthClamp device feature.

Provided by VK_EXT_extended_dynamic_state3.

pub unsafe fn cmd_set_polygon_mode_ext( &self, command_buffer: CommandBuffer, polygon_mode: PolygonMode, )

Wraps vkCmdSetPolygonModeEXT. Provided by VK_EXT_extended_dynamic_state3.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdSetPolygonModeEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Dynamically sets the polygon rasterization mode:

• FILL: normal filled triangles (default).
• LINE: wireframe rendering.
• POINT: vertices only.

LINE and POINT require the fillModeNonSolid device feature.

Provided by VK_EXT_extended_dynamic_state3.

pub unsafe fn cmd_set_rasterization_samples_ext( &self, command_buffer: CommandBuffer, rasterization_samples: SampleCountFlagBits, )

Wraps vkCmdSetRasterizationSamplesEXT. Provided by VK_EXT_extended_dynamic_state3.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdSetRasterizationSamplesEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Dynamically sets the number of rasterization samples (multisampling level). Overrides the sample count specified at pipeline creation.

The sample count must be compatible with the render pass attachments.

Provided by VK_EXT_extended_dynamic_state3.

pub unsafe fn cmd_set_sample_mask_ext( &self, command_buffer: CommandBuffer, samples: SampleCountFlagBits, p_sample_mask: Option<&u32>, )

Wraps vkCmdSetSampleMaskEXT. Provided by VK_EXT_extended_dynamic_state3.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdSetSampleMaskEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Dynamically sets the sample mask. The sample mask is ANDed with the coverage mask to determine which samples are written. Bits that are zero disable the corresponding sample.

The slice length must match ceil(rasterizationSamples / 32).

Provided by VK_EXT_extended_dynamic_state3.

pub unsafe fn cmd_set_alpha_to_coverage_enable_ext( &self, command_buffer: CommandBuffer, alpha_to_coverage_enable: bool, )

Wraps vkCmdSetAlphaToCoverageEnableEXT. Provided by VK_EXT_extended_dynamic_state3.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdSetAlphaToCoverageEnableEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Dynamically enables or disables alpha-to-coverage multisample mode. When enabled, the fragment’s alpha value determines which samples are covered, providing order-independent transparency for foliage, fences, etc.

Provided by VK_EXT_extended_dynamic_state3.

pub unsafe fn cmd_set_alpha_to_one_enable_ext( &self, command_buffer: CommandBuffer, alpha_to_one_enable: bool, )

Wraps vkCmdSetAlphaToOneEnableEXT. Provided by VK_EXT_extended_dynamic_state3.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdSetAlphaToOneEnableEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Dynamically enables or disables alpha-to-one mode. When enabled, the fragment’s alpha value is replaced with 1.0 after alpha-to-coverage processing.

Requires the alphaToOne device feature.

Provided by VK_EXT_extended_dynamic_state3.

pub unsafe fn cmd_set_logic_op_enable_ext( &self, command_buffer: CommandBuffer, logic_op_enable: bool, )

Wraps vkCmdSetLogicOpEnableEXT. Provided by VK_EXT_extended_dynamic_state3.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdSetLogicOpEnableEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Dynamically enables or disables logic operations for color blending. When enabled, fragments are combined with framebuffer values using the logic op set by cmd_set_logic_op_ext instead of standard blend equations.

Logic ops operate on integer bit patterns. They have no effect on floating-point color attachments.

Provided by VK_EXT_extended_dynamic_state3.

pub unsafe fn cmd_set_color_blend_enable_ext( &self, command_buffer: CommandBuffer, first_attachment: u32, p_color_blend_enables: &[u32], )

Wraps vkCmdSetColorBlendEnableEXT. Provided by VK_EXT_extended_dynamic_state3.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdSetColorBlendEnableEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Dynamically enables or disables color blending for each color attachment. Pass a slice of Bool32 values, one per attachment starting at first_attachment.

When blending is disabled for an attachment, the fragment color is written directly (no src/dst blending).

Provided by VK_EXT_extended_dynamic_state3.

pub unsafe fn cmd_set_color_blend_equation_ext( &self, command_buffer: CommandBuffer, first_attachment: u32, p_color_blend_equations: &[ColorBlendEquationEXT], )

Wraps vkCmdSetColorBlendEquationEXT. Provided by VK_EXT_extended_dynamic_state3.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdSetColorBlendEquationEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Dynamically sets the blend equation (src factor, dst factor, blend op) for each color attachment. Each ColorBlendEquationEXT specifies both the color and alpha blend parameters.

Overrides the values set at pipeline creation. Only effective for attachments where blending is enabled.

Provided by VK_EXT_extended_dynamic_state3.

pub unsafe fn cmd_set_color_write_mask_ext( &self, command_buffer: CommandBuffer, first_attachment: u32, p_color_write_masks: &[ColorComponentFlags], )

Wraps vkCmdSetColorWriteMaskEXT. Provided by VK_EXT_extended_dynamic_state3.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdSetColorWriteMaskEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Dynamically sets the color write mask for each color attachment. Each ColorComponentFlags value controls which channels (R, G, B, A) are written for the corresponding attachment.

Provided by VK_EXT_extended_dynamic_state3.

pub unsafe fn cmd_set_rasterization_stream_ext( &self, command_buffer: CommandBuffer, rasterization_stream: u32, )

Wraps vkCmdSetRasterizationStreamEXT. Provided by VK_EXT_extended_dynamic_state3.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdSetRasterizationStreamEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Dynamically sets which vertex stream is used for rasterization when transform feedback is active. Stream 0 is always rasterized by default; other streams can be selected with this command.

Requires VK_EXT_transform_feedback and the geometryStreams feature.

Provided by VK_EXT_extended_dynamic_state3.

pub unsafe fn cmd_set_conservative_rasterization_mode_ext( &self, command_buffer: CommandBuffer, conservative_rasterization_mode: ConservativeRasterizationModeEXT, )

Wraps vkCmdSetConservativeRasterizationModeEXT. Provided by VK_EXT_extended_dynamic_state3.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdSetConservativeRasterizationModeEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Dynamically sets the conservative rasterization mode:

• DISABLED: normal rasterization.
• OVERESTIMATE: a fragment is generated if the primitive overlaps any part of the pixel.
• UNDERESTIMATE: a fragment is generated only if the pixel is fully covered.

Requires VK_EXT_conservative_rasterization.

Provided by VK_EXT_extended_dynamic_state3.

pub unsafe fn cmd_set_extra_primitive_overestimation_size_ext( &self, command_buffer: CommandBuffer, extra_primitive_overestimation_size: f32, )

Wraps vkCmdSetExtraPrimitiveOverestimationSizeEXT. Provided by VK_EXT_extended_dynamic_state3.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdSetExtraPrimitiveOverestimationSizeEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Dynamically sets the extra overestimation size for conservative rasterization. This expands the primitive by additional pixels beyond the minimum overestimation guaranteed by the implementation.

The value is in units of pixels. 0.0 means no extra overestimation beyond the implementation minimum.

Requires VK_EXT_conservative_rasterization.

Provided by VK_EXT_extended_dynamic_state3.

pub unsafe fn cmd_set_depth_clip_enable_ext( &self, command_buffer: CommandBuffer, depth_clip_enable: bool, )

Wraps vkCmdSetDepthClipEnableEXT. Provided by VK_EXT_extended_dynamic_state3.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdSetDepthClipEnableEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Dynamically enables or disables depth clipping. When disabled, primitives are not clipped against the near and far planes (equivalent to depthClampEnable, but controlled separately).

Requires VK_EXT_depth_clip_enable and the depthClipEnable feature.

Provided by VK_EXT_extended_dynamic_state3.

pub unsafe fn cmd_set_sample_locations_enable_ext( &self, command_buffer: CommandBuffer, sample_locations_enable: bool, )

Wraps vkCmdSetSampleLocationsEnableEXT. Provided by VK_EXT_extended_dynamic_state3.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdSetSampleLocationsEnableEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Dynamically enables or disables custom sample locations. When enabled, the sample positions used for multisampling are taken from the locations set by cmd_set_sample_locations_ext instead of the implementation defaults.

Requires VK_EXT_sample_locations.

Provided by VK_EXT_extended_dynamic_state3.

pub unsafe fn cmd_set_color_blend_advanced_ext( &self, command_buffer: CommandBuffer, first_attachment: u32, p_color_blend_advanced: &[ColorBlendAdvancedEXT], )

Wraps vkCmdSetColorBlendAdvancedEXT. Provided by VK_EXT_extended_dynamic_state3.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdSetColorBlendAdvancedEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Dynamically sets advanced blend parameters per color attachment. Each ColorBlendAdvancedEXT specifies the advanced blend operation, whether src/dst are premultiplied, and the blend overlap mode (uncorrelated, disjoint, conjoint).

Only applies when using advanced blend operations (not the standard blend factors). Requires VK_EXT_blend_operation_advanced.

Provided by VK_EXT_extended_dynamic_state3.

pub unsafe fn cmd_set_provoking_vertex_mode_ext( &self, command_buffer: CommandBuffer, provoking_vertex_mode: ProvokingVertexModeEXT, )

Wraps vkCmdSetProvokingVertexModeEXT. Provided by VK_EXT_extended_dynamic_state3.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdSetProvokingVertexModeEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Dynamically sets which vertex in a primitive is the provoking vertex (the vertex whose flat-shaded attributes are used):

• FIRST_VERTEX: first vertex of the primitive (Vulkan default).
• LAST_VERTEX: last vertex (OpenGL convention).

Requires VK_EXT_provoking_vertex and the provokingVertexLast feature for LAST_VERTEX mode.

Provided by VK_EXT_extended_dynamic_state3.

pub unsafe fn cmd_set_line_rasterization_mode_ext( &self, command_buffer: CommandBuffer, line_rasterization_mode: LineRasterizationModeEXT, )

Wraps vkCmdSetLineRasterizationModeEXT. Provided by VK_EXT_extended_dynamic_state3.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdSetLineRasterizationModeEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Dynamically sets the line rasterization mode:

• DEFAULT: implementation default.
• RECTANGULAR: lines are rasterized as parallelograms (Vulkan default).
• BRESENHAM: pixel-exact Bresenham lines.
• RECTANGULAR_SMOOTH: antialiased rectangular lines.

Requires VK_EXT_line_rasterization and the corresponding device feature for the chosen mode.

Provided by VK_EXT_extended_dynamic_state3.

pub unsafe fn cmd_set_line_stipple_enable_ext( &self, command_buffer: CommandBuffer, stippled_line_enable: bool, )

Wraps vkCmdSetLineStippleEnableEXT. Provided by VK_EXT_extended_dynamic_state3.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdSetLineStippleEnableEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Dynamically enables or disables line stippling. When enabled, lines are drawn with a repeating pattern defined by cmd_set_line_stipple (core 1.4) or cmd_set_line_stipple_ext.

Requires VK_EXT_line_rasterization and the stippledLineEnable feature for the active line rasterization mode.

Provided by VK_EXT_extended_dynamic_state3.

pub unsafe fn cmd_set_depth_clip_negative_one_to_one_ext( &self, command_buffer: CommandBuffer, negative_one_to_one: bool, )

Wraps vkCmdSetDepthClipNegativeOneToOneEXT. Provided by VK_EXT_extended_dynamic_state3.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdSetDepthClipNegativeOneToOneEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Dynamically sets the depth clip range convention:

• true: NDC depth range is [-1, 1] (OpenGL convention).
• false: NDC depth range is [0, 1] (Vulkan default).

Useful for porting OpenGL content or using reversed-Z with OpenGL-style projections.

Requires VK_EXT_depth_clip_control and the depthClipControl feature.

Provided by VK_EXT_extended_dynamic_state3.

pub unsafe fn cmd_set_viewport_w_scaling_enable_nv( &self, command_buffer: CommandBuffer, viewport_w_scaling_enable: bool, )

Wraps vkCmdSetViewportWScalingEnableNV. Provided by VK_EXT_extended_dynamic_state3.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdSetViewportWScalingEnableNV was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Dynamically enables or disables viewport W scaling. When enabled, the x and y viewport coordinates are divided by an additional per-viewport W scaling factor, which can be used for lens-matched shading in VR.

Part of VK_NV_clip_space_w_scaling.

Provided by VK_EXT_extended_dynamic_state3.

pub unsafe fn cmd_set_viewport_swizzle_nv( &self, command_buffer: CommandBuffer, first_viewport: u32, p_viewport_swizzles: &[ViewportSwizzleNV], )

Wraps vkCmdSetViewportSwizzleNV. Provided by VK_EXT_extended_dynamic_state3.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdSetViewportSwizzleNV was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Dynamically sets the viewport swizzle for each viewport. A swizzle remaps each output component (x, y, z, w) to any input component, optionally negated.

Useful for single-pass cubemap rendering and other multi-view projection tricks.

Part of VK_NV_viewport_swizzle.

Provided by VK_EXT_extended_dynamic_state3.

pub unsafe fn cmd_set_coverage_to_color_enable_nv( &self, command_buffer: CommandBuffer, coverage_to_color_enable: bool, )

Wraps vkCmdSetCoverageToColorEnableNV. Provided by VK_EXT_extended_dynamic_state3.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdSetCoverageToColorEnableNV was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Dynamically enables or disables coverage-to-color mode. When enabled, the fragment coverage mask is written to a specified color attachment instead of (or in addition to) the normal color output.

Part of VK_NV_fragment_coverage_to_color.

Provided by VK_EXT_extended_dynamic_state3.

pub unsafe fn cmd_set_coverage_to_color_location_nv( &self, command_buffer: CommandBuffer, coverage_to_color_location: u32, )

Wraps vkCmdSetCoverageToColorLocationNV. Provided by VK_EXT_extended_dynamic_state3.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdSetCoverageToColorLocationNV was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Dynamically sets which color attachment receives the coverage mask when coverage-to-color is enabled.

Only effective when cmd_set_coverage_to_color_enable_nv is true.

Part of VK_NV_fragment_coverage_to_color.

Provided by VK_EXT_extended_dynamic_state3.

pub unsafe fn cmd_set_coverage_modulation_mode_nv( &self, command_buffer: CommandBuffer, coverage_modulation_mode: CoverageModulationModeNV, )

Wraps vkCmdSetCoverageModulationModeNV. Provided by VK_EXT_extended_dynamic_state3.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdSetCoverageModulationModeNV was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Dynamically sets how the coverage mask is combined with the fragment’s color. Modes include NONE, RGB, ALPHA, and RGBA.

Part of the NV coverage modulation feature used with VK_NV_framebuffer_mixed_samples.

Provided by VK_EXT_extended_dynamic_state3.

pub unsafe fn cmd_set_coverage_modulation_table_enable_nv( &self, command_buffer: CommandBuffer, coverage_modulation_table_enable: bool, )

Wraps vkCmdSetCoverageModulationTableEnableNV. Provided by VK_EXT_extended_dynamic_state3.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdSetCoverageModulationTableEnableNV was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Dynamically enables or disables the coverage modulation lookup table. When enabled, the coverage value indexes into a table set by cmd_set_coverage_modulation_table_nv instead of using the default linear modulation.

Part of VK_NV_framebuffer_mixed_samples.

Provided by VK_EXT_extended_dynamic_state3.

pub unsafe fn cmd_set_coverage_modulation_table_nv( &self, command_buffer: CommandBuffer, p_coverage_modulation_table: &[f32], )

Wraps vkCmdSetCoverageModulationTableNV. Provided by VK_EXT_extended_dynamic_state3.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdSetCoverageModulationTableNV was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Dynamically sets the coverage modulation lookup table. Each entry maps a coverage sample count to a modulation factor (0.0–1.0).

Only used when coverage modulation table is enabled via cmd_set_coverage_modulation_table_enable_nv.

Part of VK_NV_framebuffer_mixed_samples.

Provided by VK_EXT_extended_dynamic_state3.

pub unsafe fn cmd_set_shading_rate_image_enable_nv( &self, command_buffer: CommandBuffer, shading_rate_image_enable: bool, )

Wraps vkCmdSetShadingRateImageEnableNV. Provided by VK_EXT_extended_dynamic_state3.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdSetShadingRateImageEnableNV was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Dynamically enables or disables the shading rate image. When enabled, fragment shading rate is read from a shading rate image attachment instead of using a uniform rate.

Part of VK_NV_shading_rate_image.

Provided by VK_EXT_extended_dynamic_state3.

pub unsafe fn cmd_set_coverage_reduction_mode_nv( &self, command_buffer: CommandBuffer, coverage_reduction_mode: CoverageReductionModeNV, )

Wraps vkCmdSetCoverageReductionModeNV. Provided by VK_EXT_extended_dynamic_state3.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdSetCoverageReductionModeNV was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Dynamically sets the coverage reduction mode, which controls how the multisampled coverage mask is reduced to a color sample mask when the number of color samples is less than the rasterization samples.

Part of VK_NV_coverage_reduction_mode.

Provided by VK_EXT_extended_dynamic_state3.

pub unsafe fn cmd_set_representative_fragment_test_enable_nv( &self, command_buffer: CommandBuffer, representative_fragment_test_enable: bool, )

Wraps vkCmdSetRepresentativeFragmentTestEnableNV. Provided by VK_EXT_extended_dynamic_state3.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdSetRepresentativeFragmentTestEnableNV was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Dynamically enables or disables the representative fragment test. When enabled, only one fragment per pixel is shaded (the “representative” fragment), and the rest are discarded early.

Useful for visibility-only passes (e.g., occlusion culling) where full shading is unnecessary.

Part of VK_NV_representative_fragment_test.

Provided by VK_EXT_extended_dynamic_state3.

pub unsafe fn create_private_data_slot( &self, p_create_info: &PrivateDataSlotCreateInfo, allocator: Option<&AllocationCallbacks>, ) -> VkResult<PrivateDataSlot>

Wraps vkCreatePrivateDataSlot. Provided by VK_BASE_VERSION_1_3.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkCreatePrivateDataSlot was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Creates a private data slot that can be used to attach arbitrary application-defined u64 values to any Vulkan object. Private data is a lightweight alternative to external hash maps for per-object metadata.

Each slot represents one “key” that can be set on any object via set_private_data and read back via get_private_data.

Use cases:

• Tagging objects with debug IDs.
• Associating application-specific state without a separate lookup table.
• Layer and tool implementations that need per-object bookkeeping.

Private data slots are cheap. The slot itself is just an index; the per-object storage is allocated lazily by the driver.

pub unsafe fn destroy_private_data_slot( &self, private_data_slot: PrivateDataSlot, allocator: Option<&AllocationCallbacks>, )

Wraps vkDestroyPrivateDataSlot. Provided by VK_BASE_VERSION_1_3.

Safety

• device (self) must be valid and not destroyed.
• privateDataSlot must be externally synchronized.

Panics

Panics if vkDestroyPrivateDataSlot was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Destroys a private data slot. Any data previously stored via set_private_data with this slot is discarded. Objects that had data set are not affected, they continue to exist normally.

pub unsafe fn set_private_data( &self, object_type: ObjectType, object_handle: u64, private_data_slot: PrivateDataSlot, data: u64, ) -> VkResult<()>

Wraps vkSetPrivateData. Provided by VK_BASE_VERSION_1_3.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkSetPrivateData was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Stores a u64 value on any Vulkan object for the given private data slot. Overwrites any previously stored value for this object/slot pair.

Private data is per-device, the slot must have been created from the same device that owns the object. Setting data on an object from a different device is undefined behaviour.

To clear the association, set the value to 0 or destroy the slot.

pub unsafe fn get_private_data( &self, object_type: ObjectType, object_handle: u64, private_data_slot: PrivateDataSlot, ) -> u64

Wraps vkGetPrivateData. Provided by VK_BASE_VERSION_1_3.

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetPrivateData was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Retrieves the u64 value previously stored on a Vulkan object with set_private_data for the given private data slot. Returns 0 if no data was set for this object/slot combination.

This is a lightweight per-object metadata lookup with no hash table overhead. See create_private_data_slot for usage patterns.

pub unsafe fn cmd_copy_buffer2( &self, command_buffer: CommandBuffer, p_copy_buffer_info: &CopyBufferInfo2, )

Wraps vkCmdCopyBuffer2. Provided by VK_BASE_VERSION_1_3.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdCopyBuffer2 was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Vulkan 1.3 version of cmd_copy_buffer that uses an extensible CopyBufferInfo2 struct. Functionally identical to the 1.0 version but supports future extensions via pNext.

Prefer this over cmd_copy_buffer when targeting Vulkan 1.3+.

pub unsafe fn cmd_copy_image2( &self, command_buffer: CommandBuffer, p_copy_image_info: &CopyImageInfo2, )

Wraps vkCmdCopyImage2. Provided by VK_BASE_VERSION_1_3.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdCopyImage2 was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Vulkan 1.3 version of cmd_copy_image that uses an extensible CopyImageInfo2 struct.

Functionally identical to the 1.0 version. Prefer this when targeting Vulkan 1.3+.

pub unsafe fn cmd_blit_image2( &self, command_buffer: CommandBuffer, p_blit_image_info: &BlitImageInfo2, )

Wraps vkCmdBlitImage2. Provided by VK_GRAPHICS_VERSION_1_3.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdBlitImage2 was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Vulkan 1.3 version of cmd_blit_image that uses an extensible BlitImageInfo2 struct.

Chain BlitImageCubicWeightsInfoQCOM for cubic filtering on Qualcomm hardware. Otherwise functionally identical to cmd_blit_image.

Prefer this when targeting Vulkan 1.3+.

pub unsafe fn cmd_copy_buffer_to_image2( &self, command_buffer: CommandBuffer, p_copy_buffer_to_image_info: &CopyBufferToImageInfo2, )

Wraps vkCmdCopyBufferToImage2. Provided by VK_BASE_VERSION_1_3.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdCopyBufferToImage2 was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Vulkan 1.3 version of cmd_copy_buffer_to_image that uses an extensible CopyBufferToImageInfo2 struct.

Functionally identical to the 1.0 version. Prefer this when targeting Vulkan 1.3+.

pub unsafe fn cmd_copy_image_to_buffer2( &self, command_buffer: CommandBuffer, p_copy_image_to_buffer_info: &CopyImageToBufferInfo2, )

Wraps vkCmdCopyImageToBuffer2. Provided by VK_BASE_VERSION_1_3.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdCopyImageToBuffer2 was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Vulkan 1.3 version of cmd_copy_image_to_buffer that uses an extensible CopyImageToBufferInfo2 struct.

Functionally identical to the 1.0 version. Prefer this when targeting Vulkan 1.3+.

pub unsafe fn cmd_resolve_image2( &self, command_buffer: CommandBuffer, p_resolve_image_info: &ResolveImageInfo2, )

Wraps vkCmdResolveImage2. Provided by VK_GRAPHICS_VERSION_1_3.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdResolveImage2 was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Vulkan 1.3 version of cmd_resolve_image that uses an extensible ResolveImageInfo2 struct.

Functionally identical to cmd_resolve_image. Prefer this when targeting Vulkan 1.3+.

pub unsafe fn cmd_refresh_objects_khr( &self, command_buffer: CommandBuffer, p_refresh_objects: &RefreshObjectListKHR, )

Wraps vkCmdRefreshObjectsKHR.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdRefreshObjectsKHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Refreshes a set of Vulkan objects managed by a refreshable object type, resetting their internal state. Used in safety-critical Vulkan SC environments to periodically refresh objects and detect hardware faults.

Not relevant for desktop Vulkan. This is part of the VK_KHR_object_refresh extension used in automotive and embedded safety-critical environments.

pub unsafe fn cmd_set_fragment_shading_rate_khr( &self, command_buffer: CommandBuffer, p_fragment_size: &Extent2D, combiner_ops: FragmentShadingRateCombinerOpKHR, )

Wraps vkCmdSetFragmentShadingRateKHR. Provided by VK_KHR_fragment_shading_rate.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdSetFragmentShadingRateKHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Sets the pipeline fragment shading rate and combiner operations as dynamic state. This controls how many pixels each fragment shader invocation covers, larger fragment sizes reduce shading cost at the expense of detail.

p_fragment_size specifies the fragment size (e.g., 1x1, 1x2, 2x2, 2x4, 4x4). Not all sizes are supported, query get_physical_device_fragment_shading_rates_khr for the list.

combiner_ops defines how the pipeline rate, primitive rate (from the vertex shader), and attachment rate (from a shading rate image) are combined to produce the final rate.

Requires VK_KHR_fragment_shading_rate and the pipelineFragmentShadingRate device feature.

pub unsafe fn cmd_set_fragment_shading_rate_enum_nv( &self, command_buffer: CommandBuffer, shading_rate: FragmentShadingRateNV, combiner_ops: FragmentShadingRateCombinerOpKHR, )

Wraps vkCmdSetFragmentShadingRateEnumNV. Provided by VK_NV_fragment_shading_rate_enums.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdSetFragmentShadingRateEnumNV was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Dynamically sets the fragment shading rate using the NV-specific enum values. Provides finer control than the KHR variant, including support for supersample and no-invocations rates.

Requires VK_NV_fragment_shading_rate_enums.

pub unsafe fn get_acceleration_structure_build_sizes_khr( &self, build_type: AccelerationStructureBuildTypeKHR, p_build_info: &AccelerationStructureBuildGeometryInfoKHR, p_max_primitive_counts: *const u32, p_size_info: &mut AccelerationStructureBuildSizesInfoKHR, )

Wraps vkGetAccelerationStructureBuildSizesKHR. Provided by VK_KHR_acceleration_structure.

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetAccelerationStructureBuildSizesKHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Queries the buffer sizes needed to build an acceleration structure: the final structure size, the scratch buffer size for builds, and the scratch buffer size for updates.

Call this before creating the acceleration structure and scratch buffers. The max_primitive_counts parameter specifies the maximum number of primitives per geometry, the returned sizes are worst-case guarantees for those counts.

The actual built size may be smaller. For BLASes, build with ALLOW_COMPACTION and query the compacted size afterwards to reclaim excess memory.

pub unsafe fn cmd_set_vertex_input_ext( &self, command_buffer: CommandBuffer, p_vertex_binding_descriptions: &[VertexInputBindingDescription2EXT], p_vertex_attribute_descriptions: &[VertexInputAttributeDescription2EXT], )

Wraps vkCmdSetVertexInputEXT. Provided by VK_EXT_vertex_input_dynamic_state.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdSetVertexInputEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Dynamically sets the complete vertex input state: bindings (stride, input rate) and attributes (location, format, offset).

Replaces VertexInputBindingDescription and VertexInputAttributeDescription from pipeline creation. Pass arrays of VertexInputBindingDescription2EXT and VertexInputAttributeDescription2EXT.

Provided by VK_EXT_vertex_input_dynamic_state.

pub unsafe fn cmd_set_color_write_enable_ext( &self, command_buffer: CommandBuffer, p_color_write_enables: &[u32], )

Wraps vkCmdSetColorWriteEnableEXT. Provided by VK_EXT_color_write_enable.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdSetColorWriteEnableEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Dynamically enables or disables color writing for each color attachment. Pass a slice of Bool32 values, one per attachment.

When color write is disabled for an attachment, no color output is written regardless of blend state.

Unlike cmd_set_color_write_mask_ext (which controls per-channel masking), this is a simple on/off toggle per attachment.

Provided by VK_EXT_color_write_enable.

pub unsafe fn cmd_set_event2( &self, command_buffer: CommandBuffer, event: Event, p_dependency_info: &DependencyInfo, )

Wraps vkCmdSetEvent2. Provided by VK_COMPUTE_VERSION_1_3.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdSetEvent2 was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Vulkan 1.3 version of cmd_set_event that takes a DependencyInfo describing the memory dependencies, rather than just a stage mask.

This provides more precise dependency information to the driver and supports 64-bit stage and access flags. The dependency info specifies exactly what memory accesses and pipeline stages are involved, which can reduce unnecessary stalls.

Prefer this over cmd_set_event when targeting Vulkan 1.3+.

pub unsafe fn cmd_reset_event2( &self, command_buffer: CommandBuffer, event: Event, stage_mask: PipelineStageFlags2, )

Wraps vkCmdResetEvent2. Provided by VK_COMPUTE_VERSION_1_3.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdResetEvent2 was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Vulkan 1.3 version of cmd_reset_event that uses 64-bit pipeline stage flags. Supports newer stages not available in the original 32-bit field.

Prefer this over cmd_reset_event when targeting Vulkan 1.3+.

pub unsafe fn cmd_wait_events2( &self, command_buffer: CommandBuffer, p_events: &[Event], p_dependency_infos: &DependencyInfo, )

Wraps vkCmdWaitEvents2. Provided by VK_COMPUTE_VERSION_1_3.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdWaitEvents2 was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Vulkan 1.3 version of cmd_wait_events that takes per-event DependencyInfo structs instead of global stage masks and barriers.

Each event can have its own set of memory barriers and stage masks, giving the driver more precise information about what each event protects.

Prefer this over cmd_wait_events when targeting Vulkan 1.3+.

pub unsafe fn cmd_pipeline_barrier2( &self, command_buffer: CommandBuffer, p_dependency_info: &DependencyInfo, )

Wraps vkCmdPipelineBarrier2. Provided by VK_BASE_VERSION_1_3.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdPipelineBarrier2 was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Vulkan 1.3 version of cmd_pipeline_barrier that uses DependencyInfo with MemoryBarrier2, BufferMemoryBarrier2, and ImageMemoryBarrier2 structs.

The key improvement over the 1.0 version is that stage and access masks are specified per barrier rather than globally for the entire call. This gives the driver more precise dependency information, which can reduce unnecessary stalls.

The 1.3 barrier structs also use 64-bit stage and access flags, supporting stages and access types that do not fit in the original 32-bit fields (e.g. ray tracing, mesh shading).

Prefer this over cmd_pipeline_barrier when targeting Vulkan 1.3+.

pub unsafe fn queue_submit2( &self, queue: Queue, p_submits: &[SubmitInfo2], fence: Fence, ) -> VkResult<()>

Wraps vkQueueSubmit2. Provided by VK_BASE_VERSION_1_3.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_DEVICE_LOST
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• queue (self) must be valid and not destroyed.
• queue must be externally synchronized.
• fence must be externally synchronized.

Panics

Panics if vkQueueSubmit2 was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Vulkan 1.3 version of queue_submit that uses SubmitInfo2 with per-semaphore stage masks and 64-bit pipeline stage flags.

Key improvements over queue_submit:

• Per-semaphore stage masks: each wait semaphore has its own stage mask in SemaphoreSubmitInfo, instead of a parallel array. Clearer and less error-prone.
• 64-bit stages: supports newer pipeline stages.
• Timeline semaphores: timeline values are embedded in SemaphoreSubmitInfo instead of requiring a separate pNext chain.

Prefer this over queue_submit when targeting Vulkan 1.3+. The fence parameter works identically.

pub unsafe fn cmd_write_timestamp2( &self, command_buffer: CommandBuffer, stage: PipelineStageFlags2, query_pool: QueryPool, query: u32, )

Wraps vkCmdWriteTimestamp2. Provided by VK_BASE_VERSION_1_3.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdWriteTimestamp2 was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Vulkan 1.3 version of cmd_write_timestamp that uses 64-bit pipeline stage flags (PipelineStageFlags2).

The wider stage flags support newer stages like PIPELINE_STAGE_2_RAY_TRACING_SHADER_BIT_KHR and PIPELINE_STAGE_2_MESH_SHADER_BIT_EXT that do not fit in the original 32-bit field.

Prefer this over cmd_write_timestamp when targeting Vulkan 1.3+.

pub unsafe fn cmd_write_buffer_marker2_amd( &self, command_buffer: CommandBuffer, stage: PipelineStageFlags2, dst_buffer: Buffer, dst_offset: u64, marker: u32, )

Wraps vkCmdWriteBufferMarker2AMD. Provided by VK_AMD_buffer_marker.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdWriteBufferMarker2AMD was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Extended version of cmd_write_buffer_marker_amd that uses PipelineStageFlags2 for the stage mask. Supports the full synchronization2 stage flags.

Requires VK_AMD_buffer_marker + VK_KHR_synchronization2.

pub unsafe fn get_queue_checkpoint_data2_nv( &self, queue: Queue, ) -> Vec<CheckpointData2NV>

Wraps vkGetQueueCheckpointData2NV. Provided by VK_NV_device_diagnostic_checkpoints.

Safety

• queue (self) must be valid and not destroyed.

Panics

Panics if vkGetQueueCheckpointData2NV was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Extended version of get_queue_checkpoint_data_nv that returns pipeline stage information alongside the checkpoint markers. Use for finer-grained post-mortem debugging after device loss.

Requires VK_NV_device_diagnostic_checkpoints + VK_KHR_synchronization2.

pub unsafe fn copy_memory_to_image( &self, p_copy_memory_to_image_info: &CopyMemoryToImageInfo, ) -> VkResult<()>

Wraps vkCopyMemoryToImage. Provided by VK_BASE_VERSION_1_4.

Errors

• VK_ERROR_INITIALIZATION_FAILED
• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_MEMORY_MAP_FAILED
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkCopyMemoryToImage was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Vulkan 1.4 host-side image upload. Copies texel data from a host memory pointer directly into an image without a staging buffer or command buffer.

The image must be in GENERAL layout and must have been created with HOST_TRANSFER usage. The copy happens synchronously.

This simplifies upload workflows that previously required a staging buffer + cmd_copy_buffer_to_image + layout transitions. However, the image must support host transfer and be in GENERAL layout, which may not be optimal for subsequent GPU reads.

pub unsafe fn copy_image_to_memory( &self, p_copy_image_to_memory_info: &CopyImageToMemoryInfo, ) -> VkResult<()>

Wraps vkCopyImageToMemory. Provided by VK_BASE_VERSION_1_4.

Errors

• VK_ERROR_INITIALIZATION_FAILED
• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_MEMORY_MAP_FAILED
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkCopyImageToMemory was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Vulkan 1.4 host-side image readback. Copies texel data from an image directly to a host memory pointer without a staging buffer or command buffer.

The image must be in GENERAL layout and must have been created with HOST_TRANSFER usage. The copy happens synchronously.

This simplifies CPU readback workflows that previously required a staging buffer + cmd_copy_image_to_buffer + fence wait + map. However, it requires the image to support host transfer, which not all implementations or formats support.

pub unsafe fn copy_image_to_image( &self, p_copy_image_to_image_info: &CopyImageToImageInfo, ) -> VkResult<()>

Wraps vkCopyImageToImage. Provided by VK_BASE_VERSION_1_4.

Errors

• VK_ERROR_INITIALIZATION_FAILED
• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_MEMORY_MAP_FAILED
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkCopyImageToImage was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Vulkan 1.4 host-side image-to-image copy. Copies texel data between two images from the CPU without recording a command buffer.

Both images must be in GENERAL layout and must have been created with HOST_TRANSFER usage. The copy happens synchronously on the calling thread.

Use cases are limited to CPU-side image manipulation (e.g. test utilities, offline processing). For GPU-side copies in a render loop, cmd_copy_image2 is the standard path.

pub unsafe fn transition_image_layout( &self, p_transitions: &[HostImageLayoutTransitionInfo], ) -> VkResult<()>

Wraps vkTransitionImageLayout. Provided by VK_BASE_VERSION_1_4.

Errors

• VK_ERROR_INITIALIZATION_FAILED
• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_MEMORY_MAP_FAILED
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkTransitionImageLayout was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Vulkan 1.4 host-side image layout transition. Transitions an image between layouts from the CPU without recording a command buffer.

The image must have been created with HOST_TRANSFER usage. The transition happens synchronously on the calling thread.

This simplifies workflows where you need to transition an image layout outside of a command buffer, for example, transitioning a newly created host-transfer image from UNDEFINED to GENERAL before performing host-side copies.

For GPU-side layout transitions (the common case), use cmd_pipeline_barrier2 with an image memory barrier.

pub unsafe fn get_command_pool_memory_consumption( &self, command_pool: CommandPool, command_buffer: CommandBuffer, p_consumption: &mut CommandPoolMemoryConsumption, )

Wraps vkGetCommandPoolMemoryConsumption. Provided by VKSC_VERSION_1_0.

Safety

• device (self) must be valid and not destroyed.
• commandPool must be externally synchronized.
• commandBuffer must be externally synchronized.

Panics

Panics if vkGetCommandPoolMemoryConsumption was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Queries the memory consumption of a command pool or a specific command buffer within it. Part of Vulkan SC (Safety Critical) for tracking resource budgets in safety-certified environments. Pass a null command buffer to query the entire pool.

Requires Vulkan SC.

pub unsafe fn create_video_session_khr( &self, p_create_info: &VideoSessionCreateInfoKHR, allocator: Option<&AllocationCallbacks>, ) -> VkResult<VideoSessionKHR>

Wraps vkCreateVideoSessionKHR. Provided by VK_KHR_video_queue.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_INITIALIZATION_FAILED
• VK_ERROR_VIDEO_STD_VERSION_NOT_SUPPORTED_KHR
• VK_ERROR_INVALID_VIDEO_STD_PARAMETERS_KHR
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkCreateVideoSessionKHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Creates a video session for hardware-accelerated video decoding or encoding. The session defines the video codec profile, resolution range, format, and maximum reference picture count.

Key fields in VideoSessionCreateInfoKHR:

• video_profile: codec (H.264, H.265, AV1) and profile/level.
• max_coded_extent: maximum frame resolution.
• picture_format / reference_picture_format: image formats for decoded pictures and DPB (decoded picture buffer) slots.
• max_dpb_slots / max_active_reference_pictures: reference frame capacity.

After creation, query memory requirements with get_video_session_memory_requirements_khr, allocate and bind memory with bind_video_session_memory_khr, then create session parameters with create_video_session_parameters_khr.

pub unsafe fn destroy_video_session_khr( &self, video_session: VideoSessionKHR, allocator: Option<&AllocationCallbacks>, )

Wraps vkDestroyVideoSessionKHR. Provided by VK_KHR_video_queue.

Safety

• device (self) must be valid and not destroyed.
• videoSession must be externally synchronized.

Panics

Panics if vkDestroyVideoSessionKHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Destroys a video session and releases its internal resources. Any video session parameters created against this session become invalid, destroy them first.

All command buffers referencing this session must have completed execution before destruction.

pub unsafe fn create_video_session_parameters_khr( &self, p_create_info: &VideoSessionParametersCreateInfoKHR, allocator: Option<&AllocationCallbacks>, ) -> VkResult<VideoSessionParametersKHR>

Wraps vkCreateVideoSessionParametersKHR. Provided by VK_KHR_video_queue.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_INITIALIZATION_FAILED
• VK_ERROR_INVALID_VIDEO_STD_PARAMETERS_KHR
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkCreateVideoSessionParametersKHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Creates a video session parameters object that holds codec-specific parameter sets (SPS, PPS for H.264/H.265, sequence headers for AV1). These are referenced during decode/encode operations.

Chain the appropriate codec-specific struct into pNext:

• VideoDecodeH264SessionParametersCreateInfoKHR for H.264 decode.
• VideoDecodeH265SessionParametersCreateInfoKHR for H.265 decode.
• VideoEncodeH264SessionParametersCreateInfoKHR for H.264 encode.

Parameters can be added incrementally with update_video_session_parameters_khr. A template parameter object can be specified to inherit existing parameters.

pub unsafe fn update_video_session_parameters_khr( &self, video_session_parameters: VideoSessionParametersKHR, p_update_info: &VideoSessionParametersUpdateInfoKHR, ) -> VkResult<()>

Wraps vkUpdateVideoSessionParametersKHR. Provided by VK_KHR_video_queue.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_INVALID_VIDEO_STD_PARAMETERS_KHR
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkUpdateVideoSessionParametersKHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Adds new codec-specific parameter sets to an existing video session parameters object. For example, adding new SPS/PPS entries for H.264 as they are encountered in the bitstream.

Chain the codec-specific update struct into the pNext of VideoSessionParametersUpdateInfoKHR. The update_sequence_count must increment monotonically with each update.

Parameters cannot be removed or modified, only new entries can be added. If a parameter set with the same ID already exists, the update fails.

pub unsafe fn get_encoded_video_session_parameters_khr( &self, p_video_session_parameters_info: &VideoEncodeSessionParametersGetInfoKHR, p_feedback_info: &mut VideoEncodeSessionParametersFeedbackInfoKHR, p_data: *mut c_void, ) -> VkResult<usize>

Wraps vkGetEncodedVideoSessionParametersKHR. Provided by VK_KHR_video_encode_queue.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetEncodedVideoSessionParametersKHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Retrieves the encoded (serialized) form of video session parameters, typically codec headers (SPS/PPS for H.264/H.265, sequence header for AV1) that must be prepended to the encoded bitstream.

Uses the two-call pattern: call with null p_data to query the size, allocate, then call again to fill the buffer.

The p_feedback_info output indicates whether the driver modified or overrode any parameters relative to what was requested (check has_overrides).

This data is the codec parameter payload that decoders need to initialize before processing encoded frames.

pub unsafe fn destroy_video_session_parameters_khr( &self, video_session_parameters: VideoSessionParametersKHR, allocator: Option<&AllocationCallbacks>, )

Wraps vkDestroyVideoSessionParametersKHR. Provided by VK_KHR_video_queue.

Safety

• device (self) must be valid and not destroyed.
• videoSessionParameters must be externally synchronized.

Panics

Panics if vkDestroyVideoSessionParametersKHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Destroys a video session parameters object. All command buffers referencing these parameters must have completed execution before destruction.

The video session itself is not affected, other parameter objects associated with the same session remain valid.

pub unsafe fn get_video_session_memory_requirements_khr( &self, video_session: VideoSessionKHR, ) -> VkResult<Vec<VideoSessionMemoryRequirementsKHR>>

Wraps vkGetVideoSessionMemoryRequirementsKHR. Provided by VK_KHR_video_queue.

Errors

• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetVideoSessionMemoryRequirementsKHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Queries the memory requirements for a video session. A video session may require multiple memory bindings (each with different memory type requirements) for internal buffers used during decode/encode.

Each returned VideoSessionMemoryRequirementsKHR has a memory_bind_index and a MemoryRequirements describing the size, alignment, and compatible memory types.

Allocate a DeviceMemory for each requirement and bind them all with bind_video_session_memory_khr before using the session.

pub unsafe fn bind_video_session_memory_khr( &self, video_session: VideoSessionKHR, p_bind_session_memory_infos: &[BindVideoSessionMemoryInfoKHR], ) -> VkResult<()>

Wraps vkBindVideoSessionMemoryKHR. Provided by VK_KHR_video_queue.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.
• videoSession must be externally synchronized.

Panics

Panics if vkBindVideoSessionMemoryKHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Binds device memory to a video session. Each binding corresponds to a memory_bind_index from get_video_session_memory_requirements_khr.

All required memory bindings must be satisfied before the session can be used in video coding operations. Each BindVideoSessionMemoryInfoKHR specifies the bind index, memory object, offset, and size.

Memory can only be bound once per index, rebinding is not allowed.

pub unsafe fn cmd_decode_video_khr( &self, command_buffer: CommandBuffer, p_decode_info: &VideoDecodeInfoKHR, )

Wraps vkCmdDecodeVideoKHR. Provided by VK_KHR_video_decode_queue.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdDecodeVideoKHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Decodes a single video frame from a compressed bitstream. Must be recorded within a video coding scope (cmd_begin_video_coding_khr / cmd_end_video_coding_khr).

VideoDecodeInfoKHR specifies:

• src_buffer / src_buffer_offset / src_buffer_range: the bitstream data containing the compressed frame.
• dst_picture_resource: the output image view for the decoded frame.
• setup_reference_slot: DPB slot to store this frame for use as a reference by future frames.
• reference_slots: previously decoded reference frames needed to decode this frame.

Chain codec-specific decode info (e.g., VideoDecodeH264PictureInfoKHR) into pNext.

pub unsafe fn cmd_begin_video_coding_khr( &self, command_buffer: CommandBuffer, p_begin_info: &VideoBeginCodingInfoKHR, )

Wraps vkCmdBeginVideoCodingKHR. Provided by VK_KHR_video_queue.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdBeginVideoCodingKHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Begins a video coding scope within a command buffer. All video decode and encode commands must be recorded between cmd_begin_video_coding_khr and cmd_end_video_coding_khr.

VideoBeginCodingInfoKHR specifies:

• video_session: the session to use.
• video_session_parameters: codec parameters (SPS/PPS, etc.).
• reference_slots: DPB (decoded picture buffer) slots and their associated image views for reference pictures.

The command buffer must be allocated from a queue family that supports the appropriate video operations (decode or encode), as reported by QueueFamilyVideoPropertiesKHR.

pub unsafe fn cmd_control_video_coding_khr( &self, command_buffer: CommandBuffer, p_coding_control_info: &VideoCodingControlInfoKHR, )

Wraps vkCmdControlVideoCodingKHR. Provided by VK_KHR_video_queue.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdControlVideoCodingKHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Issues control commands within a video coding scope. Used to reset the video session state or set encode quality/rate control parameters.

VideoCodingControlInfoKHR flags include:

• RESET: resets the video session to a clean state, clearing all DPB slots and internal codec state.
• ENCODE_RATE_CONTROL: applies rate control settings (chain VideoEncodeRateControlInfoKHR into pNext).
• ENCODE_QUALITY_LEVEL: sets the encode quality level.

Must be recorded between cmd_begin_video_coding_khr and cmd_end_video_coding_khr.

pub unsafe fn cmd_end_video_coding_khr( &self, command_buffer: CommandBuffer, p_end_coding_info: &VideoEndCodingInfoKHR, )

Wraps vkCmdEndVideoCodingKHR. Provided by VK_KHR_video_queue.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdEndVideoCodingKHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Ends a video coding scope previously started with cmd_begin_video_coding_khr. After this call, video decode and encode commands can no longer be recorded until a new scope is started.

The VideoEndCodingInfoKHR struct is currently reserved for future use (no flags defined).

pub unsafe fn cmd_encode_video_khr( &self, command_buffer: CommandBuffer, p_encode_info: &VideoEncodeInfoKHR, )

Wraps vkCmdEncodeVideoKHR. Provided by VK_KHR_video_encode_queue.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdEncodeVideoKHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Encodes a single video frame into a compressed bitstream. Must be recorded within a video coding scope (cmd_begin_video_coding_khr / cmd_end_video_coding_khr).

VideoEncodeInfoKHR specifies:

• dst_buffer / dst_buffer_offset / dst_buffer_range: where to write the compressed output.
• src_picture_resource: the input image view to encode.
• setup_reference_slot: DPB slot to store the reconstructed frame for future reference.
• reference_slots: reference frames for inter-prediction.

Chain codec-specific encode info (e.g., VideoEncodeH264PictureInfoKHR) into pNext. Configure rate control beforehand with cmd_control_video_coding_khr.

pub unsafe fn cmd_decompress_memory_nv( &self, command_buffer: CommandBuffer, p_decompress_memory_regions: &[DecompressMemoryRegionNV], )

Wraps vkCmdDecompressMemoryNV. Provided by VK_NV_memory_decompression.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdDecompressMemoryNV was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Decompresses one or more memory regions on the GPU. Each region specifies source, destination, size, and decompression method.

Requires VK_NV_memory_decompression.

pub unsafe fn cmd_decompress_memory_indirect_count_nv( &self, command_buffer: CommandBuffer, indirect_commands_address: u64, indirect_commands_count_address: u64, stride: u32, )

Wraps vkCmdDecompressMemoryIndirectCountNV. Provided by VK_NV_memory_decompression.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdDecompressMemoryIndirectCountNV was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Indirect-count variant of cmd_decompress_memory_nv. Reads the decompression region descriptors and count from GPU buffer addresses, enabling fully GPU-driven decompression.

Requires VK_NV_memory_decompression.

pub unsafe fn get_partitioned_acceleration_structures_build_sizes_nv( &self, p_info: &PartitionedAccelerationStructureInstancesInputNV, p_size_info: &mut AccelerationStructureBuildSizesInfoKHR, )

Wraps vkGetPartitionedAccelerationStructuresBuildSizesNV. Provided by VK_NV_partitioned_acceleration_structure.

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetPartitionedAccelerationStructuresBuildSizesNV was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Queries the buffer sizes needed to build a partitioned acceleration structure. Use the returned sizes to allocate destination and scratch buffers.

Requires VK_NV_partitioned_acceleration_structure.

pub unsafe fn cmd_build_partitioned_acceleration_structures_nv( &self, command_buffer: CommandBuffer, p_build_info: &BuildPartitionedAccelerationStructureInfoNV, )

Wraps vkCmdBuildPartitionedAccelerationStructuresNV. Provided by VK_NV_partitioned_acceleration_structure.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdBuildPartitionedAccelerationStructuresNV was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Builds a partitioned acceleration structure where instances are grouped into independently updatable partitions. This allows updating subsets of the TLAS without rebuilding the entire structure.

Requires VK_NV_partitioned_acceleration_structure.

pub unsafe fn cmd_decompress_memory_ext( &self, command_buffer: CommandBuffer, p_decompress_memory_info_ext: &DecompressMemoryInfoEXT, )

Wraps vkCmdDecompressMemoryEXT. Provided by VK_EXT_memory_decompression.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdDecompressMemoryEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Decompresses data from one memory region into another on the GPU. The decompression algorithm is specified in the info structure.

Useful for loading compressed assets directly on the GPU without a CPU round-trip.

Requires VK_EXT_memory_decompression.

pub unsafe fn cmd_decompress_memory_indirect_count_ext( &self, command_buffer: CommandBuffer, decompression_method: MemoryDecompressionMethodFlagsEXT, indirect_commands_address: u64, indirect_commands_count_address: u64, max_decompression_count: u32, stride: u32, )

Wraps vkCmdDecompressMemoryIndirectCountEXT. Provided by VK_EXT_memory_decompression.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdDecompressMemoryIndirectCountEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Indirect variant of cmd_decompress_memory_ext. Reads the decompression parameters and count from GPU-visible buffer addresses, enabling fully GPU-driven decompression workflows.

Requires VK_EXT_memory_decompression.

pub unsafe fn create_cu_module_nvx( &self, p_create_info: &CuModuleCreateInfoNVX, allocator: Option<&AllocationCallbacks>, ) -> VkResult<CuModuleNVX>

Wraps vkCreateCuModuleNVX. Provided by VK_NVX_binary_import.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_INITIALIZATION_FAILED
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkCreateCuModuleNVX was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Creates a CUDA module from binary data using the legacy NVX path. Prefer create_cuda_module_nv for new code.

Destroy with destroy_cu_module_nvx.

Requires VK_NVX_binary_import.

pub unsafe fn create_cu_function_nvx( &self, p_create_info: &CuFunctionCreateInfoNVX, allocator: Option<&AllocationCallbacks>, ) -> VkResult<CuFunctionNVX>

Wraps vkCreateCuFunctionNVX. Provided by VK_NVX_binary_import.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_INITIALIZATION_FAILED
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkCreateCuFunctionNVX was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Creates a CUDA function handle from an NVX binary module. This is the legacy NVX path; prefer create_cuda_function_nv for new code.

Destroy with destroy_cu_function_nvx.

Requires VK_NVX_binary_import.

pub unsafe fn destroy_cu_module_nvx( &self, module: CuModuleNVX, allocator: Option<&AllocationCallbacks>, )

Wraps vkDestroyCuModuleNVX. Provided by VK_NVX_binary_import.

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkDestroyCuModuleNVX was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Destroys a CUDA module created with create_cu_module_nvx.

Requires VK_NVX_binary_import.

pub unsafe fn destroy_cu_function_nvx( &self, function: CuFunctionNVX, allocator: Option<&AllocationCallbacks>, )

Wraps vkDestroyCuFunctionNVX. Provided by VK_NVX_binary_import.

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkDestroyCuFunctionNVX was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Destroys a CUDA function handle created with create_cu_function_nvx.

Requires VK_NVX_binary_import.

pub unsafe fn cmd_cu_launch_kernel_nvx( &self, command_buffer: CommandBuffer, p_launch_info: &CuLaunchInfoNVX, )

Wraps vkCmdCuLaunchKernelNVX. Provided by VK_NVX_binary_import.

Safety

• commandBuffer (self) must be valid and not destroyed.

Panics

Panics if vkCmdCuLaunchKernelNVX was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Launches a CUDA kernel from a Vulkan command buffer using the legacy NVX binary import path. Prefer cmd_cuda_launch_kernel_nv for new code.

Requires VK_NVX_binary_import.

pub unsafe fn get_descriptor_set_layout_size_ext( &self, layout: DescriptorSetLayout, ) -> u64

Wraps vkGetDescriptorSetLayoutSizeEXT. Provided by VK_EXT_descriptor_buffer.

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetDescriptorSetLayoutSizeEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Returns the total byte size required to store all descriptors for the given descriptor set layout in a descriptor buffer.

Use this to allocate the correct amount of buffer memory for each descriptor set, then write individual descriptors at offsets obtained from get_descriptor_set_layout_binding_offset_ext.

Requires VK_EXT_descriptor_buffer.

pub unsafe fn get_descriptor_set_layout_binding_offset_ext( &self, layout: DescriptorSetLayout, binding: u32, ) -> u64

Wraps vkGetDescriptorSetLayoutBindingOffsetEXT. Provided by VK_EXT_descriptor_buffer.

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetDescriptorSetLayoutBindingOffsetEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Returns the byte offset of a specific binding within the descriptor buffer layout for the given descriptor set layout.

Use this to compute where to write a descriptor with get_descriptor_ext within the buffer region for a set.

Requires VK_EXT_descriptor_buffer.

pub unsafe fn get_descriptor_ext( &self, p_descriptor_info: &DescriptorGetInfoEXT, data_size: usize, p_descriptor: *mut c_void, )

Wraps vkGetDescriptorEXT. Provided by VK_EXT_descriptor_buffer.

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetDescriptorEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Writes a descriptor directly into caller-provided memory. DescriptorGetInfoEXT specifies the descriptor type and resource (buffer, image, sampler, etc.). The descriptor is written to p_descriptor and must be data_size bytes.

Query the required size per descriptor type with PhysicalDeviceDescriptorBufferPropertiesEXT.

This is the core operation of descriptor buffers, instead of allocating descriptor sets, you write descriptors directly into mapped buffer memory.

Requires VK_EXT_descriptor_buffer.

pub unsafe fn cmd_bind_descriptor_buffers_ext( &self, command_buffer: CommandBuffer, p_binding_infos: &[DescriptorBufferBindingInfoEXT], )

Wraps vkCmdBindDescriptorBuffersEXT. Provided by VK_EXT_descriptor_buffer.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdBindDescriptorBuffersEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Binds one or more descriptor buffers to a command buffer. Each DescriptorBufferBindingInfoEXT specifies a buffer address and usage (resource descriptors, sampler descriptors, or push descriptors).

After binding, use cmd_set_descriptor_buffer_offsets_ext to point specific descriptor sets at offsets within the bound buffers.

Descriptor buffers are an alternative to descriptor sets/pools that stores descriptors inline in buffer memory.

Requires VK_EXT_descriptor_buffer.

pub unsafe fn cmd_set_descriptor_buffer_offsets_ext( &self, command_buffer: CommandBuffer, pipeline_bind_point: PipelineBindPoint, layout: PipelineLayout, first_set: u32, p_buffer_indices: &[u32], p_offsets: &u64, )

Wraps vkCmdSetDescriptorBufferOffsetsEXT. Provided by VK_EXT_descriptor_buffer.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdSetDescriptorBufferOffsetsEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Sets the offsets into bound descriptor buffers for one or more descriptor set slots. Each pair of (buffer_index, offset) maps a descriptor set to a region of a previously bound descriptor buffer.

Must be called after cmd_bind_descriptor_buffers_ext.

For the pNext-extensible variant, see cmd_set_descriptor_buffer_offsets2_ext.

Requires VK_EXT_descriptor_buffer.

pub unsafe fn cmd_bind_descriptor_buffer_embedded_samplers_ext( &self, command_buffer: CommandBuffer, pipeline_bind_point: PipelineBindPoint, layout: PipelineLayout, set: u32, )

Wraps vkCmdBindDescriptorBufferEmbeddedSamplersEXT. Provided by VK_EXT_descriptor_buffer.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdBindDescriptorBufferEmbeddedSamplersEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Binds embedded immutable samplers from a descriptor set layout that was created with CREATE_EMBEDDED_IMMUTABLE_SAMPLERS_BIT. These samplers are baked into the layout and do not need buffer memory.

Specify the pipeline_bind_point, layout, and set index.

For the pNext-extensible variant, see cmd_bind_descriptor_buffer_embedded_samplers2_ext.

Requires VK_EXT_descriptor_buffer.

pub unsafe fn get_buffer_opaque_capture_descriptor_data_ext( &self, p_info: &BufferCaptureDescriptorDataInfoEXT, ) -> VkResult<c_void>

Wraps vkGetBufferOpaqueCaptureDescriptorDataEXT. Provided by VK_EXT_descriptor_buffer.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetBufferOpaqueCaptureDescriptorDataEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Retrieves opaque capture data for a buffer descriptor. The returned data can be used to reconstruct the descriptor in a replay or capture/replay scenario.

The buffer must have been created with CREATE_DESCRIPTOR_BUFFER_CAPTURE_REPLAY_BIT.

Requires VK_EXT_descriptor_buffer and descriptorBufferCaptureReplay.

pub unsafe fn get_image_opaque_capture_descriptor_data_ext( &self, p_info: &ImageCaptureDescriptorDataInfoEXT, ) -> VkResult<c_void>

Wraps vkGetImageOpaqueCaptureDescriptorDataEXT. Provided by VK_EXT_descriptor_buffer.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetImageOpaqueCaptureDescriptorDataEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Retrieves opaque capture data for an image descriptor. The returned data can be used to reconstruct the descriptor in a capture/replay scenario.

The image must have been created with CREATE_DESCRIPTOR_BUFFER_CAPTURE_REPLAY_BIT.

Requires VK_EXT_descriptor_buffer and descriptorBufferCaptureReplay.

pub unsafe fn get_image_view_opaque_capture_descriptor_data_ext( &self, p_info: &ImageViewCaptureDescriptorDataInfoEXT, ) -> VkResult<c_void>

Wraps vkGetImageViewOpaqueCaptureDescriptorDataEXT. Provided by VK_EXT_descriptor_buffer.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetImageViewOpaqueCaptureDescriptorDataEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Retrieves opaque capture data for an image view descriptor. The returned data can be used to reconstruct the descriptor in a capture/replay scenario.

The image view must have been created with CREATE_DESCRIPTOR_BUFFER_CAPTURE_REPLAY_BIT.

Requires VK_EXT_descriptor_buffer and descriptorBufferCaptureReplay.

pub unsafe fn get_sampler_opaque_capture_descriptor_data_ext( &self, p_info: &SamplerCaptureDescriptorDataInfoEXT, ) -> VkResult<c_void>

Wraps vkGetSamplerOpaqueCaptureDescriptorDataEXT. Provided by VK_EXT_descriptor_buffer.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetSamplerOpaqueCaptureDescriptorDataEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Retrieves opaque capture data for a sampler descriptor. The returned data can be used to reconstruct the descriptor in a capture/replay scenario.

The sampler must have been created with CREATE_DESCRIPTOR_BUFFER_CAPTURE_REPLAY_BIT.

Requires VK_EXT_descriptor_buffer and descriptorBufferCaptureReplay.

pub unsafe fn get_acceleration_structure_opaque_capture_descriptor_data_ext( &self, p_info: &AccelerationStructureCaptureDescriptorDataInfoEXT, ) -> VkResult<c_void>

Wraps vkGetAccelerationStructureOpaqueCaptureDescriptorDataEXT. Provided by VK_EXT_descriptor_buffer.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetAccelerationStructureOpaqueCaptureDescriptorDataEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Retrieves opaque capture data for an acceleration structure descriptor. The returned data can be used to reconstruct the descriptor in a replay or capture/replay scenario.

The acceleration structure must have been created with CREATE_DESCRIPTOR_BUFFER_CAPTURE_REPLAY_BIT.

Requires VK_EXT_descriptor_buffer and descriptorBufferCaptureReplay.

pub unsafe fn set_device_memory_priority_ext( &self, memory: DeviceMemory, priority: f32, )

Wraps vkSetDeviceMemoryPriorityEXT. Provided by VK_EXT_pageable_device_local_memory.

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkSetDeviceMemoryPriorityEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Dynamically updates the priority of a device memory allocation. Higher-priority allocations are less likely to be evicted under memory pressure. Use this to promote frequently accessed resources or demote resources that are no longer critical.

Requires VK_EXT_pageable_device_local_memory.

pub unsafe fn wait_for_present2_khr( &self, swapchain: SwapchainKHR, p_present_wait2_info: &PresentWait2InfoKHR, ) -> VkResult<()>

Wraps vkWaitForPresent2KHR. Provided by VK_KHR_present_wait2.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_DEVICE_LOST
• VK_ERROR_OUT_OF_DATE_KHR
• VK_ERROR_SURFACE_LOST_KHR
• VK_ERROR_FULL_SCREEN_EXCLUSIVE_MODE_LOST_EXT
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.
• swapchain must be externally synchronized.

Panics

Panics if vkWaitForPresent2KHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Extensible version of wait_for_present_khr. Takes a PresentWait2InfoKHR struct (with pNext support) instead of separate present_id and timeout parameters.

Provided by VK_KHR_present_wait2. Otherwise identical in behavior, blocks until the specified present ID completes or the timeout expires.

pub unsafe fn wait_for_present_khr( &self, swapchain: SwapchainKHR, present_id: u64, timeout: u64, ) -> VkResult<()>

Wraps vkWaitForPresentKHR. Provided by VK_KHR_present_wait.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_DEVICE_LOST
• VK_ERROR_OUT_OF_DATE_KHR
• VK_ERROR_SURFACE_LOST_KHR
• VK_ERROR_FULL_SCREEN_EXCLUSIVE_MODE_LOST_EXT
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.
• swapchain must be externally synchronized.

Panics

Panics if vkWaitForPresentKHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Blocks the calling thread until a specific present operation completes on the display. present_id identifies which present to wait for, set it via PresentIdKHR chained into PresentInfoKHR during queue_present_khr.

timeout is in nanoseconds. Returns TIMEOUT if the deadline expires before the present completes, SUCCESS if the present finished. Use u64::MAX for an indefinite wait.

Requires VK_KHR_present_wait and VK_KHR_present_id.

This is useful for frame pacing, wait for the previous frame’s present to complete before starting the next frame’s work to avoid queuing excessive frames.

pub unsafe fn create_buffer_collection_fuchsia( &self, p_create_info: &BufferCollectionCreateInfoFUCHSIA, allocator: Option<&AllocationCallbacks>, ) -> VkResult<BufferCollectionFUCHSIA>

Wraps vkCreateBufferCollectionFUCHSIA. Provided by VK_FUCHSIA_buffer_collection.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_INVALID_EXTERNAL_HANDLE
• VK_ERROR_INITIALIZATION_FAILED
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkCreateBufferCollectionFUCHSIA was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Creates a Fuchsia buffer collection that negotiates memory constraints between Vulkan and other Fuchsia services (e.g. scenic, camera). Fuchsia OS only. After creation, set buffer or image constraints before allocating.

Requires VK_FUCHSIA_buffer_collection.

pub unsafe fn set_buffer_collection_buffer_constraints_fuchsia( &self, collection: BufferCollectionFUCHSIA, p_buffer_constraints_info: &BufferConstraintsInfoFUCHSIA, ) -> VkResult<()>

Wraps vkSetBufferCollectionBufferConstraintsFUCHSIA. Provided by VK_FUCHSIA_buffer_collection.

Errors

• VK_ERROR_INITIALIZATION_FAILED
• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_FORMAT_NOT_SUPPORTED
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkSetBufferCollectionBufferConstraintsFUCHSIA was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Sets buffer constraints on a Fuchsia buffer collection. The constraints describe the Vulkan buffer usage requirements that must be negotiated with other collection participants. Fuchsia OS only.

Requires VK_FUCHSIA_buffer_collection.

pub unsafe fn set_buffer_collection_image_constraints_fuchsia( &self, collection: BufferCollectionFUCHSIA, p_image_constraints_info: &ImageConstraintsInfoFUCHSIA, ) -> VkResult<()>

Wraps vkSetBufferCollectionImageConstraintsFUCHSIA. Provided by VK_FUCHSIA_buffer_collection.

Errors

• VK_ERROR_INITIALIZATION_FAILED
• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_FORMAT_NOT_SUPPORTED
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkSetBufferCollectionImageConstraintsFUCHSIA was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Sets image constraints on a Fuchsia buffer collection. The constraints describe the Vulkan image format and usage requirements that must be negotiated with other collection participants. Fuchsia OS only.

Requires VK_FUCHSIA_buffer_collection.

pub unsafe fn destroy_buffer_collection_fuchsia( &self, collection: BufferCollectionFUCHSIA, allocator: Option<&AllocationCallbacks>, )

Wraps vkDestroyBufferCollectionFUCHSIA. Provided by VK_FUCHSIA_buffer_collection.

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkDestroyBufferCollectionFUCHSIA was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Destroys a Fuchsia buffer collection. The collection must not be in use by any pending operations. Fuchsia OS only.

Requires VK_FUCHSIA_buffer_collection.

pub unsafe fn get_buffer_collection_properties_fuchsia( &self, collection: BufferCollectionFUCHSIA, p_properties: &mut BufferCollectionPropertiesFUCHSIA, ) -> VkResult<()>

Wraps vkGetBufferCollectionPropertiesFUCHSIA. Provided by VK_FUCHSIA_buffer_collection.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_INITIALIZATION_FAILED
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetBufferCollectionPropertiesFUCHSIA was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Queries the negotiated properties of a Fuchsia buffer collection after constraints have been set. Returns memory type index, format, and other details needed for allocation. Fuchsia OS only.

Requires VK_FUCHSIA_buffer_collection.

pub unsafe fn create_cuda_module_nv( &self, p_create_info: &CudaModuleCreateInfoNV, allocator: Option<&AllocationCallbacks>, ) -> VkResult<CudaModuleNV>

Wraps vkCreateCudaModuleNV. Provided by VK_NV_cuda_kernel_launch.

Errors

• VK_ERROR_INITIALIZATION_FAILED
• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkCreateCudaModuleNV was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Creates a CUDA module from PTX or cubin data. The module contains one or more kernel entry points that can be extracted with create_cuda_function_nv.

Destroy with destroy_cuda_module_nv.

Requires VK_NV_cuda_kernel_launch.

pub unsafe fn get_cuda_module_cache_nv( &self, module: CudaModuleNV, p_cache_data: *mut c_void, ) -> VkResult<usize>

Wraps vkGetCudaModuleCacheNV. Provided by VK_NV_cuda_kernel_launch.

Errors

• VK_ERROR_INITIALIZATION_FAILED
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetCudaModuleCacheNV was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Retrieves the compiled cache data from a CUDA module for serialization. Call once with a null buffer to query the size, then again with an appropriately sized buffer. Feed the data back into create_cuda_module_nv on the next run to skip compilation.

Requires VK_NV_cuda_kernel_launch.

pub unsafe fn create_cuda_function_nv( &self, p_create_info: &CudaFunctionCreateInfoNV, allocator: Option<&AllocationCallbacks>, ) -> VkResult<CudaFunctionNV>

Wraps vkCreateCudaFunctionNV. Provided by VK_NV_cuda_kernel_launch.

Errors

• VK_ERROR_INITIALIZATION_FAILED
• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkCreateCudaFunctionNV was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Creates a CUDA function handle from a CUDA module, identifying a specific kernel entry point. Use with cmd_cuda_launch_kernel_nv to dispatch the kernel.

Destroy with destroy_cuda_function_nv.

Requires VK_NV_cuda_kernel_launch.

pub unsafe fn destroy_cuda_module_nv( &self, module: CudaModuleNV, allocator: Option<&AllocationCallbacks>, )

Wraps vkDestroyCudaModuleNV. Provided by VK_NV_cuda_kernel_launch.

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkDestroyCudaModuleNV was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Destroys a CUDA module created with create_cuda_module_nv. All functions extracted from this module must be destroyed first.

Requires VK_NV_cuda_kernel_launch.

pub unsafe fn destroy_cuda_function_nv( &self, function: CudaFunctionNV, allocator: Option<&AllocationCallbacks>, )

Wraps vkDestroyCudaFunctionNV. Provided by VK_NV_cuda_kernel_launch.

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkDestroyCudaFunctionNV was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Destroys a CUDA function handle created with create_cuda_function_nv.

Requires VK_NV_cuda_kernel_launch.

pub unsafe fn cmd_cuda_launch_kernel_nv( &self, command_buffer: CommandBuffer, p_launch_info: &CudaLaunchInfoNV, )

Wraps vkCmdCudaLaunchKernelNV. Provided by VK_NV_cuda_kernel_launch.

Safety

• commandBuffer (self) must be valid and not destroyed.

Panics

Panics if vkCmdCudaLaunchKernelNV was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Launches a CUDA kernel from within a Vulkan command buffer. The kernel is specified by a CUDA function handle created with create_cuda_function_nv. Grid dimensions and parameters are provided in the launch info.

Requires VK_NV_cuda_kernel_launch.

pub unsafe fn cmd_begin_rendering( &self, command_buffer: CommandBuffer, p_rendering_info: &RenderingInfo, )

Wraps vkCmdBeginRendering. Provided by VK_GRAPHICS_VERSION_1_3.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdBeginRendering was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Begins dynamic rendering, a Vulkan 1.3 alternative to render pass objects that specifies attachments inline at command recording time.

Advantages over render passes:

• No RenderPass or Framebuffer objects to create and manage.
• Attachments are specified directly as image views in RenderingInfo.
• Simpler code for applications that do not benefit from tile-based subpass optimisations.

RenderingInfo specifies:

• Colour attachments: image views, load/store ops, clear values.
• Depth attachment: optional, with its own load/store ops.
• Stencil attachment: optional, can share the same image view as depth.
• Render area and layer count.

Flags:

• RENDERING_CONTENTS_SECONDARY_COMMAND_BUFFERS: draw commands come from secondary command buffers.
• RENDERING_SUSPENDING / RENDERING_RESUMING: split rendering across multiple command buffers.

Graphics pipelines used with dynamic rendering must be created with PipelineRenderingCreateInfo instead of a render pass handle.

pub unsafe fn cmd_end_rendering(&self, command_buffer: CommandBuffer)

Wraps vkCmdEndRendering. Provided by VK_GRAPHICS_VERSION_1_3.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdEndRendering was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Ends a dynamic rendering instance started by cmd_begin_rendering. Store operations and any resolve operations specified in the RenderingInfo are executed at this point.

After this call, no draw commands may be recorded until a new rendering or render pass instance is begun.

pub unsafe fn cmd_end_rendering2_khr( &self, command_buffer: CommandBuffer, p_rendering_end_info: Option<&RenderingEndInfoKHR>, )

Wraps vkCmdEndRendering2KHR. Provided by VK_KHR_maintenance10.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdEndRendering2KHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Extended version of cmd_end_rendering (core 1.3) that accepts an optional RenderingEndInfoKHR with pNext extensibility.

Ends the current dynamic rendering pass. If p_rendering_end_info is None, behaves identically to cmd_end_rendering.

Provided by VK_KHR_maintenance7.

pub unsafe fn get_descriptor_set_layout_host_mapping_info_valve( &self, p_binding_reference: &DescriptorSetBindingReferenceVALVE, p_host_mapping: &mut DescriptorSetLayoutHostMappingInfoVALVE, )

Wraps vkGetDescriptorSetLayoutHostMappingInfoVALVE. Provided by VK_VALVE_descriptor_set_host_mapping.

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetDescriptorSetLayoutHostMappingInfoVALVE was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Queries the host memory layout for a specific binding within a descriptor set layout. Returns the stride and offset needed to write descriptors directly via the host pointer obtained from get_descriptor_set_host_mapping_valve.

Requires VK_VALVE_descriptor_set_host_mapping.

pub unsafe fn get_descriptor_set_host_mapping_valve( &self, descriptor_set: DescriptorSet, pp_data: *mut *mut c_void, )

Wraps vkGetDescriptorSetHostMappingVALVE. Provided by VK_VALVE_descriptor_set_host_mapping.

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetDescriptorSetHostMappingVALVE was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Retrieves a host pointer to the internal memory backing a descriptor set. Allows direct CPU writes to descriptor data, bypassing the normal update_descriptor_sets path for lower overhead. The layout must match what was queried with get_descriptor_set_layout_host_mapping_info_valve.

Requires VK_VALVE_descriptor_set_host_mapping.

pub unsafe fn create_micromap_ext( &self, p_create_info: &MicromapCreateInfoEXT, allocator: Option<&AllocationCallbacks>, ) -> VkResult<MicromapEXT>

Wraps vkCreateMicromapEXT. Provided by VK_EXT_opacity_micromap.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_INVALID_OPAQUE_CAPTURE_ADDRESS_KHR
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkCreateMicromapEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Creates an opacity micromap object. Micromaps store per-triangle opacity or hit data at sub-triangle granularity, enabling the ray tracing implementation to skip fully transparent micro- triangles without invoking any-hit shaders.

The MicromapCreateInfoEXT specifies the backing buffer, size, and type (OPACITY_MICROMAP).

Build with cmd_build_micromaps_ext or build_micromaps_ext. Destroy with destroy_micromap_ext.

Requires VK_EXT_opacity_micromap.

pub unsafe fn cmd_build_micromaps_ext( &self, command_buffer: CommandBuffer, p_infos: &[MicromapBuildInfoEXT], )

Wraps vkCmdBuildMicromapsEXT. Provided by VK_EXT_opacity_micromap.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdBuildMicromapsEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Records a GPU-side micromap build into a command buffer. Each MicromapBuildInfoEXT specifies the source triangle opacity data, destination micromap, and scratch memory.

For the CPU-side equivalent, see build_micromaps_ext.

Requires VK_EXT_opacity_micromap.

pub unsafe fn build_micromaps_ext( &self, deferred_operation: DeferredOperationKHR, p_infos: &[MicromapBuildInfoEXT], ) -> VkResult<()>

Wraps vkBuildMicromapsEXT. Provided by VK_EXT_opacity_micromap.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkBuildMicromapsEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Builds one or more micromaps on the host. This is the CPU-side equivalent of cmd_build_micromaps_ext.

Each MicromapBuildInfoEXT specifies the source triangle data, destination micromap, and scratch memory.

Requires VK_EXT_opacity_micromap and the micromapHostCommands feature.

pub unsafe fn destroy_micromap_ext( &self, micromap: MicromapEXT, allocator: Option<&AllocationCallbacks>, )

Wraps vkDestroyMicromapEXT. Provided by VK_EXT_opacity_micromap.

Safety

• device (self) must be valid and not destroyed.
• micromap must be externally synchronized.

Panics

Panics if vkDestroyMicromapEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Destroys a micromap created with create_micromap_ext. The backing buffer is not freed, the application must manage buffer lifetime separately.

Requires VK_EXT_opacity_micromap.

pub unsafe fn cmd_copy_micromap_ext( &self, command_buffer: CommandBuffer, p_info: &CopyMicromapInfoEXT, )

Wraps vkCmdCopyMicromapEXT. Provided by VK_EXT_opacity_micromap.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdCopyMicromapEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Copies or compacts a micromap. The CopyMicromapInfoEXT specifies source, destination, and mode (CLONE or COMPACT).

Use COMPACT after building with BUILD_ALLOW_COMPACTION to reduce memory usage. Query the compacted size with cmd_write_micromaps_properties_ext.

Requires VK_EXT_opacity_micromap.

pub unsafe fn copy_micromap_ext( &self, deferred_operation: DeferredOperationKHR, p_info: &CopyMicromapInfoEXT, ) -> VkResult<()>

Wraps vkCopyMicromapEXT. Provided by VK_EXT_opacity_micromap.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkCopyMicromapEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Host-side micromap copy or compaction. This is the CPU equivalent of cmd_copy_micromap_ext.

Requires VK_EXT_opacity_micromap and the micromapHostCommands feature.

pub unsafe fn cmd_copy_micromap_to_memory_ext( &self, command_buffer: CommandBuffer, p_info: &CopyMicromapToMemoryInfoEXT, )

Wraps vkCmdCopyMicromapToMemoryEXT. Provided by VK_EXT_opacity_micromap.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdCopyMicromapToMemoryEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Serializes a micromap into a buffer for storage or transfer. The CopyMicromapToMemoryInfoEXT specifies the source micromap and destination device address.

Deserialize with cmd_copy_memory_to_micromap_ext.

Requires VK_EXT_opacity_micromap.

pub unsafe fn copy_micromap_to_memory_ext( &self, deferred_operation: DeferredOperationKHR, p_info: &CopyMicromapToMemoryInfoEXT, ) -> VkResult<()>

Wraps vkCopyMicromapToMemoryEXT. Provided by VK_EXT_opacity_micromap.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkCopyMicromapToMemoryEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Host-side serialization of a micromap into a buffer. This is the CPU equivalent of cmd_copy_micromap_to_memory_ext.

Requires VK_EXT_opacity_micromap and the micromapHostCommands feature.

pub unsafe fn cmd_copy_memory_to_micromap_ext( &self, command_buffer: CommandBuffer, p_info: &CopyMemoryToMicromapInfoEXT, )

Wraps vkCmdCopyMemoryToMicromapEXT. Provided by VK_EXT_opacity_micromap.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdCopyMemoryToMicromapEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Deserializes a micromap from a buffer into a micromap object. This is the reverse of cmd_copy_micromap_to_memory_ext.

Used for loading previously serialized micromaps from disk or transferring between devices.

Requires VK_EXT_opacity_micromap.

pub unsafe fn copy_memory_to_micromap_ext( &self, deferred_operation: DeferredOperationKHR, p_info: &CopyMemoryToMicromapInfoEXT, ) -> VkResult<()>

Wraps vkCopyMemoryToMicromapEXT. Provided by VK_EXT_opacity_micromap.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkCopyMemoryToMicromapEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Host-side deserialization of a micromap from a buffer. This is the CPU equivalent of cmd_copy_memory_to_micromap_ext.

Requires VK_EXT_opacity_micromap and the micromapHostCommands feature.

pub unsafe fn cmd_write_micromaps_properties_ext( &self, command_buffer: CommandBuffer, p_micromaps: &[MicromapEXT], query_type: QueryType, query_pool: QueryPool, first_query: u32, )

Wraps vkCmdWriteMicromapsPropertiesEXT. Provided by VK_EXT_opacity_micromap.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdWriteMicromapsPropertiesEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Writes micromap properties (e.g., compacted size) to a query pool. Use QUERY_TYPE_MICROMAP_COMPACTED_SIZE_EXT to query the size needed for a compacted copy.

Requires VK_EXT_opacity_micromap.

pub unsafe fn write_micromaps_properties_ext( &self, p_micromaps: &[MicromapEXT], query_type: QueryType, data_size: usize, p_data: *mut c_void, stride: usize, ) -> VkResult<()>

Wraps vkWriteMicromapsPropertiesEXT. Provided by VK_EXT_opacity_micromap.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkWriteMicromapsPropertiesEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Host-side query of micromap properties. This is the CPU equivalent of cmd_write_micromaps_properties_ext.

Typically used to query compacted size (QUERY_TYPE_MICROMAP_COMPACTED_SIZE_EXT) without going through a query pool.

Requires VK_EXT_opacity_micromap and the micromapHostCommands feature.

pub unsafe fn get_device_micromap_compatibility_ext( &self, p_version_info: &MicromapVersionInfoEXT, ) -> AccelerationStructureCompatibilityKHR

Wraps vkGetDeviceMicromapCompatibilityEXT. Provided by VK_EXT_opacity_micromap.

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetDeviceMicromapCompatibilityEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Checks whether a serialized micromap is compatible with the current device. Returns COMPATIBLE or INCOMPATIBLE.

Use this before deserializing a micromap (via cmd_copy_memory_to_micromap_ext or copy_memory_to_micromap_ext) to verify it will work on this device.

Requires VK_EXT_opacity_micromap.

pub unsafe fn get_micromap_build_sizes_ext( &self, build_type: AccelerationStructureBuildTypeKHR, p_build_info: &MicromapBuildInfoEXT, p_size_info: &mut MicromapBuildSizesInfoEXT, )

Wraps vkGetMicromapBuildSizesEXT. Provided by VK_EXT_opacity_micromap.

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetMicromapBuildSizesEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Queries the memory requirements for building a micromap. Returns the destination micromap size and scratch memory size needed.

Provide a MicromapBuildInfoEXT with the triangle counts and format. The addresses can be null, only the sizes and counts matter for this query.

Use the results to allocate the micromap buffer and scratch buffer before calling cmd_build_micromaps_ext.

Requires VK_EXT_opacity_micromap.

pub unsafe fn get_shader_module_identifier_ext( &self, shader_module: ShaderModule, p_identifier: &mut ShaderModuleIdentifierEXT, )

Wraps vkGetShaderModuleIdentifierEXT. Provided by VK_EXT_shader_module_identifier.

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetShaderModuleIdentifierEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Retrieves the identifier for an existing shader module. The identifier can be used for pipeline cache lookups via PipelineShaderStageModuleIdentifierCreateInfoEXT.

The identifier is deterministic for the same SPIR-V content on the same implementation.

Requires VK_EXT_shader_module_identifier.

pub unsafe fn get_shader_module_create_info_identifier_ext( &self, p_create_info: &ShaderModuleCreateInfo, p_identifier: &mut ShaderModuleIdentifierEXT, )

Wraps vkGetShaderModuleCreateInfoIdentifierEXT. Provided by VK_EXT_shader_module_identifier.

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetShaderModuleCreateInfoIdentifierEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Computes an identifier for a shader module from its create info (SPIR-V code) without creating the module. The identifier can be used in PipelineShaderStageModuleIdentifierCreateInfoEXT to create pipelines from cached shader data.

Useful for pipeline caching workflows where the SPIR-V is available but you want to avoid full module creation.

Requires VK_EXT_shader_module_identifier.

pub unsafe fn get_image_subresource_layout2( &self, image: Image, p_subresource: &ImageSubresource2, p_layout: &mut SubresourceLayout2, )

Wraps vkGetImageSubresourceLayout2. Provided by VK_BASE_VERSION_1_4.

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetImageSubresourceLayout2 was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Vulkan 1.4 version of get_image_subresource_layout that uses extensible structs via pNext.

Returns the layout (offset, size, row pitch, array pitch, depth pitch) for a given subresource of an existing image. Chain ImageCompressionPropertiesEXT to query fixed-rate compression state.

For linear-tiling images, this tells you how to access texels through a mapped pointer. For optimal-tiling images, the layout is opaque and implementation-defined.

pub unsafe fn get_pipeline_properties_ext( &self, p_pipeline_info: &PipelineInfoEXT, p_pipeline_properties: &mut BaseOutStructure, ) -> VkResult<()>

Wraps vkGetPipelinePropertiesEXT. Provided by VK_EXT_pipeline_properties.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetPipelinePropertiesEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Queries properties of a pipeline, such as its pipeline identifier. The identifier can be used to correlate pipelines across processes or with external tools.

Requires VK_EXT_pipeline_properties.

pub unsafe fn export_metal_objects_ext( &self, p_metal_objects_info: &mut ExportMetalObjectsInfoEXT, )

Wraps vkExportMetalObjectsEXT. Provided by VK_EXT_metal_objects.

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkExportMetalObjectsEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Exports the underlying Metal objects (device, command queue, textures, buffers, etc.) from Vulkan objects. Chain the appropriate export struct into the pNext of the info structure to select which Metal object to retrieve.

Useful for Metal interop on Apple platforms where both APIs share the same GPU resources.

Requires VK_EXT_metal_objects. macOS/iOS only.

pub unsafe fn cmd_bind_tile_memory_qcom( &self, command_buffer: CommandBuffer, p_tile_memory_bind_info: Option<&TileMemoryBindInfoQCOM>, )

Wraps vkCmdBindTileMemoryQCOM. Provided by VK_QCOM_tile_memory_heap.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdBindTileMemoryQCOM was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Binds tile memory for use within a per-tile execution region on Qualcomm tile-based GPUs. Pass None to unbind. Tile memory provides fast on-chip scratch storage scoped to each tile.

Requires VK_QCOM_tile_shading.

pub unsafe fn get_framebuffer_tile_properties_qcom( &self, framebuffer: Framebuffer, ) -> VkResult<Vec<TilePropertiesQCOM>>

Wraps vkGetFramebufferTilePropertiesQCOM. Provided by VK_QCOM_tile_properties.

Errors

• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetFramebufferTilePropertiesQCOM was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Queries tile properties (tile dimensions and extents) for a framebuffer on Qualcomm tile-based GPUs. Uses the two-call idiom. Useful for optimising rendering to match the hardware tile layout.

Requires VK_QCOM_tile_properties.

pub unsafe fn get_dynamic_rendering_tile_properties_qcom( &self, p_rendering_info: &RenderingInfo, p_properties: &mut TilePropertiesQCOM, ) -> VkResult<()>

Wraps vkGetDynamicRenderingTilePropertiesQCOM. Provided by VK_QCOM_tile_properties.

Errors

• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetDynamicRenderingTilePropertiesQCOM was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Queries tile properties for a dynamic rendering pass (as opposed to a framebuffer-based render pass). Returns tile dimensions that would be used for the given rendering info on Qualcomm tile-based GPUs.

Requires VK_QCOM_tile_properties.

pub unsafe fn create_optical_flow_session_nv( &self, p_create_info: &OpticalFlowSessionCreateInfoNV, allocator: Option<&AllocationCallbacks>, ) -> VkResult<OpticalFlowSessionNV>

Wraps vkCreateOpticalFlowSessionNV. Provided by VK_NV_optical_flow.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_INITIALIZATION_FAILED
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkCreateOpticalFlowSessionNV was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Creates an optical flow session for GPU-accelerated motion estimation between image pairs. Configure the session with image format, resolution, and flow precision in the create info.

Destroy with destroy_optical_flow_session_nv.

Requires VK_NV_optical_flow.

pub unsafe fn destroy_optical_flow_session_nv( &self, session: OpticalFlowSessionNV, allocator: Option<&AllocationCallbacks>, )

Wraps vkDestroyOpticalFlowSessionNV. Provided by VK_NV_optical_flow.

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkDestroyOpticalFlowSessionNV was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Destroys an optical flow session created with create_optical_flow_session_nv.

Requires VK_NV_optical_flow.

pub unsafe fn bind_optical_flow_session_image_nv( &self, session: OpticalFlowSessionNV, binding_point: OpticalFlowSessionBindingPointNV, view: ImageView, layout: ImageLayout, ) -> VkResult<()>

Wraps vkBindOpticalFlowSessionImageNV. Provided by VK_NV_optical_flow.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_INITIALIZATION_FAILED
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkBindOpticalFlowSessionImageNV was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Binds an image view to an optical flow session at a specific binding point (reference, target, flow vector output, etc.). All required binding points must be bound before executing the optical flow.

Requires VK_NV_optical_flow.

pub unsafe fn cmd_optical_flow_execute_nv( &self, command_buffer: CommandBuffer, session: OpticalFlowSessionNV, p_execute_info: &OpticalFlowExecuteInfoNV, )

Wraps vkCmdOpticalFlowExecuteNV. Provided by VK_NV_optical_flow.

Safety

• commandBuffer (self) must be valid and not destroyed.

Panics

Panics if vkCmdOpticalFlowExecuteNV was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Executes optical flow estimation on the GPU using the bound reference and target images. Results are written to the bound flow vector output image.

Requires VK_NV_optical_flow.

pub unsafe fn get_device_fault_info_ext( &self, p_fault_counts: &mut DeviceFaultCountsEXT, p_fault_info: &mut DeviceFaultInfoEXT, ) -> VkResult<()>

Wraps vkGetDeviceFaultInfoEXT. Provided by VK_EXT_device_fault.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetDeviceFaultInfoEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Retrieves diagnostic information after a device-lost error. Call once with a null info pointer to query the fault counts, then again with allocated structures to retrieve the fault details.

The returned data is vendor-specific and intended for crash reporting and post-mortem debugging.

Requires VK_EXT_device_fault.

pub unsafe fn get_device_fault_reports_khr( &self, timeout: u64, ) -> VkResult<Vec<DeviceFaultInfoKHR>>

Wraps vkGetDeviceFaultReportsKHR. Provided by VK_KHR_device_fault.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetDeviceFaultReportsKHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Retrieves fault reports after a device loss (ERROR_DEVICE_LOST). Returns a list of DeviceFaultInfoKHR structs describing what went wrong, address faults, vendor-specific fault codes, etc.

timeout specifies how long to wait (in nanoseconds) for the driver to collect fault data. Use UINT64_MAX to wait indefinitely.

For raw debug data suitable for vendor tools, follow up with get_device_fault_debug_info_khr.

Requires VK_KHR_device_fault.

pub unsafe fn get_device_fault_debug_info_khr( &self, p_debug_info: &mut DeviceFaultDebugInfoKHR, ) -> VkResult<()>

Wraps vkGetDeviceFaultDebugInfoKHR. Provided by VK_KHR_device_fault.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_NOT_ENOUGH_SPACE_KHR
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetDeviceFaultDebugInfoKHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Retrieves debug information after a device fault (GPU crash or hang). The returned DeviceFaultDebugInfoKHR contains vendor-specific binary data that can be passed to GPU vendor diagnostic tools.

Call get_device_fault_reports_khr first to get the high-level fault reports; this call provides the raw debug blob.

Requires VK_KHR_device_fault.

pub unsafe fn cmd_set_depth_bias2_ext( &self, command_buffer: CommandBuffer, p_depth_bias_info: &DepthBiasInfoEXT, )

Wraps vkCmdSetDepthBias2EXT. Provided by VK_EXT_depth_bias_control.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdSetDepthBias2EXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Extended version of cmd_set_depth_bias that takes a DepthBiasInfoEXT struct with pNext extensibility. This allows chaining DepthBiasRepresentationInfoEXT to control the depth bias representation (least-representable-value vs. float).

Requires VK_EXT_depth_bias_control.

pub unsafe fn release_swapchain_images_khr( &self, p_release_info: &ReleaseSwapchainImagesInfoKHR, ) -> VkResult<()>

Wraps vkReleaseSwapchainImagesKHR. Provided by VK_KHR_swapchain_maintenance1.

Errors

• VK_ERROR_SURFACE_LOST_KHR
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkReleaseSwapchainImagesKHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Releases acquired swapchain images back to the swapchain without presenting them. Provided by VK_KHR_swapchain_maintenance1.

Use this when you have acquired an image but decided not to render to it, for example, when aborting a frame due to a resize or error. Without this extension, the only way to return an acquired image is to present it.

The released images return to the pool and can be acquired again.

pub unsafe fn get_device_image_subresource_layout( &self, p_info: &DeviceImageSubresourceInfo, p_layout: &mut SubresourceLayout2, )

Wraps vkGetDeviceImageSubresourceLayout. Provided by VK_BASE_VERSION_1_4.

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetDeviceImageSubresourceLayout was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Vulkan 1.4 command that queries the subresource layout for an image without creating it first. Returns the offset, size, row pitch, array pitch, and depth pitch for a given subresource of a hypothetical image.

Useful for pre-planning host-side memory layouts when using HOST_TRANSFER images, or for calculating buffer sizes for staging uploads.

pub unsafe fn unmap_memory2( &self, p_memory_unmap_info: &MemoryUnmapInfo, ) -> VkResult<()>

Wraps vkUnmapMemory2. Provided by VK_BASE_VERSION_1_4.

Errors

• VK_ERROR_MEMORY_MAP_FAILED
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkUnmapMemory2 was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Vulkan 1.4 version of unmap_memory that uses an extensible MemoryUnmapInfo struct. Supports MEMORY_UNMAP_RESERVE (if available) to keep the virtual address range reserved after unmapping, useful for placed mappings.

For most applications, unmap_memory and unmap_memory2 are equivalent. Prefer this when targeting Vulkan 1.4+.

pub unsafe fn create_shaders_ext( &self, p_create_infos: &[ShaderCreateInfoEXT], allocator: Option<&AllocationCallbacks>, ) -> VkResult<Vec<ShaderEXT>>

Wraps vkCreateShadersEXT. Provided by VK_EXT_shader_object.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_INITIALIZATION_FAILED
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkCreateShadersEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Creates one or more shader objects from SPIR-V code. Shader objects are an alternative to pipelines, instead of baking shaders into monolithic pipeline objects, individual shader stages can be bound independently.

Each ShaderCreateInfoEXT specifies the stage, code, entry point, and optional specialization constants.

Bind with cmd_bind_shaders_ext. Destroy with destroy_shader_ext.

Requires VK_EXT_shader_object.

pub unsafe fn destroy_shader_ext( &self, shader: ShaderEXT, allocator: Option<&AllocationCallbacks>, )

Wraps vkDestroyShaderEXT. Provided by VK_EXT_shader_object.

Safety

• device (self) must be valid and not destroyed.
• shader must be externally synchronized.

Panics

Panics if vkDestroyShaderEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Destroys a shader object created with create_shaders_ext.

Requires VK_EXT_shader_object.

pub unsafe fn get_shader_binary_data_ext( &self, shader: ShaderEXT, p_data: *mut c_void, ) -> VkResult<usize>

Wraps vkGetShaderBinaryDataEXT. Provided by VK_EXT_shader_object.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetShaderBinaryDataEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Retrieves the binary representation of a compiled shader object. The binary data can be cached to disk and used to create shader objects without recompiling from SPIR-V on subsequent runs.

Call once with a null buffer to query the size, then again with an appropriately sized buffer. Shader binaries are implementation-specific and not portable between devices or driver versions.

Requires VK_EXT_shader_object.

pub unsafe fn cmd_bind_shaders_ext( &self, command_buffer: CommandBuffer, p_stages: &[ShaderStageFlagBits], p_shaders: Option<&ShaderEXT>, )

Wraps vkCmdBindShadersEXT. Provided by VK_EXT_shader_object.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdBindShadersEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Binds shader objects to specified shader stages for subsequent draw or dispatch commands. Pass arrays of stages and corresponding shader handles.

To unbind a stage, pass a null handle for that stage. All required stages for the draw/dispatch type must be bound.

When using shader objects, you must also set all relevant dynamic state, there is no pipeline to provide defaults.

Requires VK_EXT_shader_object.

pub unsafe fn set_swapchain_present_timing_queue_size_ext( &self, swapchain: SwapchainKHR, size: u32, ) -> VkResult<()>

Wraps vkSetSwapchainPresentTimingQueueSizeEXT. Provided by VK_EXT_present_timing.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.
• swapchain must be externally synchronized.

Panics

Panics if vkSetSwapchainPresentTimingQueueSizeEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Sets the maximum number of present timing results the driver will queue for later retrieval via get_past_presentation_timing_ext. A larger queue prevents timing data from being lost when the application cannot poll frequently.

Requires VK_EXT_present_timing.

pub unsafe fn get_swapchain_timing_properties_ext( &self, swapchain: SwapchainKHR, p_swapchain_timing_properties: &mut SwapchainTimingPropertiesEXT, ) -> VkResult<u64>

Wraps vkGetSwapchainTimingPropertiesEXT. Provided by VK_EXT_present_timing.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_SURFACE_LOST_KHR
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.
• swapchain must be externally synchronized.

Panics

Panics if vkGetSwapchainTimingPropertiesEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Queries the timing properties of a swapchain, such as the refresh duration and present margin. Use this to calibrate frame pacing and target the display’s native refresh interval.

Requires VK_EXT_present_timing.

pub unsafe fn get_swapchain_time_domain_properties_ext( &self, swapchain: SwapchainKHR, p_swapchain_time_domain_properties: &mut SwapchainTimeDomainPropertiesEXT, ) -> VkResult<u64>

Wraps vkGetSwapchainTimeDomainPropertiesEXT. Provided by VK_EXT_present_timing.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_SURFACE_LOST_KHR
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.
• swapchain must be externally synchronized.

Panics

Panics if vkGetSwapchainTimeDomainPropertiesEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Queries which time domains the swapchain supports for present timing. Use this to determine whether you can correlate present timestamps with the system clock or device-specific counters.

Requires VK_EXT_present_timing.

pub unsafe fn get_past_presentation_timing_ext( &self, p_past_presentation_timing_info: &PastPresentationTimingInfoEXT, p_past_presentation_timing_properties: &mut PastPresentationTimingPropertiesEXT, ) -> VkResult<()>

Wraps vkGetPastPresentationTimingEXT. Provided by VK_EXT_present_timing.

Errors

• VK_ERROR_DEVICE_LOST
• VK_ERROR_OUT_OF_DATE_KHR
• VK_ERROR_SURFACE_LOST_KHR
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetPastPresentationTimingEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Retrieves presentation timing data for previously presented images. Use this to measure actual vs. requested present times and detect missed frames or compositor latency.

Requires VK_EXT_present_timing.

pub unsafe fn get_screen_buffer_properties_qnx( &self, buffer: *const c_void, p_properties: &mut ScreenBufferPropertiesQNX, ) -> VkResult<()>

Wraps vkGetScreenBufferPropertiesQNX. Provided by VK_QNX_external_memory_screen_buffer.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_INVALID_EXTERNAL_HANDLE_KHR
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetScreenBufferPropertiesQNX was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Queries Vulkan memory properties for a QNX Screen buffer. Use before importing the buffer as Vulkan memory to determine compatible memory types and size. QNX only.

Requires VK_QNX_external_memory_screen_buffer.

pub unsafe fn get_execution_graph_pipeline_scratch_size_amdx( &self, execution_graph: Pipeline, p_size_info: &mut ExecutionGraphPipelineScratchSizeAMDX, ) -> VkResult<()>

Wraps vkGetExecutionGraphPipelineScratchSizeAMDX. Provided by VK_AMDX_shader_enqueue.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetExecutionGraphPipelineScratchSizeAMDX was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Queries the scratch memory size required to execute an execution graph pipeline. Allocate a buffer of at least this size and initialize it with cmd_initialize_graph_scratch_memory_amdx before dispatching the graph.

Requires VK_AMDX_shader_enqueue.

pub unsafe fn get_execution_graph_pipeline_node_index_amdx( &self, execution_graph: Pipeline, p_node_info: &PipelineShaderStageNodeCreateInfoAMDX, ) -> VkResult<u32>

Wraps vkGetExecutionGraphPipelineNodeIndexAMDX. Provided by VK_AMDX_shader_enqueue.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetExecutionGraphPipelineNodeIndexAMDX was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Queries the node index for a named shader node in an execution graph pipeline. The index is used when dispatching or enqueuing work to a specific node in the graph.

Requires VK_AMDX_shader_enqueue.

pub unsafe fn create_execution_graph_pipelines_amdx( &self, pipeline_cache: PipelineCache, p_create_infos: &[ExecutionGraphPipelineCreateInfoAMDX], allocator: Option<&AllocationCallbacks>, ) -> VkResult<Vec<Pipeline>>

Wraps vkCreateExecutionGraphPipelinesAMDX. Provided by VK_AMDX_shader_enqueue.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.
• pipelineCache must be externally synchronized.

Panics

Panics if vkCreateExecutionGraphPipelinesAMDX was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Creates one or more execution graph pipelines for GPU-driven shader dispatch. An execution graph is a DAG of shader nodes where each node can enqueue work to other nodes, enabling complex GPU-driven workflows without CPU round-trips.

Requires VK_AMDX_shader_enqueue.

pub unsafe fn cmd_initialize_graph_scratch_memory_amdx( &self, command_buffer: CommandBuffer, execution_graph: Pipeline, scratch: u64, scratch_size: u64, )

Wraps vkCmdInitializeGraphScratchMemoryAMDX. Provided by VK_AMDX_shader_enqueue.

Safety

• commandBuffer (self) must be valid and not destroyed.

Panics

Panics if vkCmdInitializeGraphScratchMemoryAMDX was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Initializes the scratch memory buffer for an execution graph pipeline. Must be called before any cmd_dispatch_graph_*_amdx command that uses this scratch buffer.

Requires VK_AMDX_shader_enqueue.

pub unsafe fn cmd_dispatch_graph_amdx( &self, command_buffer: CommandBuffer, scratch: u64, scratch_size: u64, p_count_info: &DispatchGraphCountInfoAMDX, )

Wraps vkCmdDispatchGraphAMDX. Provided by VK_AMDX_shader_enqueue.

Safety

• commandBuffer (self) must be valid and not destroyed.

Panics

Panics if vkCmdDispatchGraphAMDX was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Dispatches an execution graph starting from the specified root nodes. The scratch buffer must have been initialized with cmd_initialize_graph_scratch_memory_amdx.

Requires VK_AMDX_shader_enqueue.

pub unsafe fn cmd_dispatch_graph_indirect_amdx( &self, command_buffer: CommandBuffer, scratch: u64, scratch_size: u64, p_count_info: &DispatchGraphCountInfoAMDX, )

Wraps vkCmdDispatchGraphIndirectAMDX. Provided by VK_AMDX_shader_enqueue.

Safety

• commandBuffer (self) must be valid and not destroyed.

Panics

Panics if vkCmdDispatchGraphIndirectAMDX was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Indirect variant of cmd_dispatch_graph_amdx. Reads the graph dispatch payloads from a GPU buffer while the count info is provided on the CPU side.

Requires VK_AMDX_shader_enqueue.

pub unsafe fn cmd_dispatch_graph_indirect_count_amdx( &self, command_buffer: CommandBuffer, scratch: u64, scratch_size: u64, count_info: u64, )

Wraps vkCmdDispatchGraphIndirectCountAMDX. Provided by VK_AMDX_shader_enqueue.

Safety

• commandBuffer (self) must be valid and not destroyed.

Panics

Panics if vkCmdDispatchGraphIndirectCountAMDX was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Fully indirect variant of cmd_dispatch_graph_amdx. Both the dispatch payloads and the count are read from GPU buffers, enabling fully GPU-driven execution graph dispatch.

Requires VK_AMDX_shader_enqueue.

pub unsafe fn cmd_bind_descriptor_sets2( &self, command_buffer: CommandBuffer, p_bind_descriptor_sets_info: &BindDescriptorSetsInfo, )

Wraps vkCmdBindDescriptorSets2. Provided by VK_COMPUTE_VERSION_1_4.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdBindDescriptorSets2 was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Vulkan 1.4 version of cmd_bind_descriptor_sets that uses an extensible BindDescriptorSetsInfo struct.

Functionally equivalent to the 1.0 version. The extensible struct enables future extensions to modify binding behaviour via pNext.

Prefer this when targeting Vulkan 1.4+.

pub unsafe fn cmd_push_constants2( &self, command_buffer: CommandBuffer, p_push_constants_info: &PushConstantsInfo, )

Wraps vkCmdPushConstants2. Provided by VK_COMPUTE_VERSION_1_4.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdPushConstants2 was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Vulkan 1.4 version of cmd_push_constants that uses an extensible PushConstantsInfo struct.

Functionally equivalent to the 1.0 version. Prefer this when targeting Vulkan 1.4+.

pub unsafe fn cmd_push_descriptor_set2( &self, command_buffer: CommandBuffer, p_push_descriptor_set_info: &PushDescriptorSetInfo, )

Wraps vkCmdPushDescriptorSet2. Provided by VK_COMPUTE_VERSION_1_4.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdPushDescriptorSet2 was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Vulkan 1.4 version of cmd_push_descriptor_set that uses an extensible PushDescriptorSetInfo struct.

Functionally equivalent to the base version. Prefer this when targeting Vulkan 1.4+.

pub unsafe fn cmd_push_descriptor_set_with_template2( &self, command_buffer: CommandBuffer, p_push_descriptor_set_with_template_info: &PushDescriptorSetWithTemplateInfo, )

Wraps vkCmdPushDescriptorSetWithTemplate2. Provided by VK_COMPUTE_VERSION_1_4.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdPushDescriptorSetWithTemplate2 was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Vulkan 1.4 version of cmd_push_descriptor_set_with_template that uses an extensible PushDescriptorSetWithTemplateInfo struct.

Functionally equivalent to the base version. Prefer this when targeting Vulkan 1.4+.

pub unsafe fn cmd_set_descriptor_buffer_offsets2_ext( &self, command_buffer: CommandBuffer, p_set_descriptor_buffer_offsets_info: &SetDescriptorBufferOffsetsInfoEXT, )

Wraps vkCmdSetDescriptorBufferOffsets2EXT. Provided by VK_KHR_maintenance6.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdSetDescriptorBufferOffsets2EXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Extended version of cmd_set_descriptor_buffer_offsets_ext that takes a SetDescriptorBufferOffsetsInfoEXT struct with pNext extensibility.

Sets the offsets into bound descriptor buffers for the specified pipeline layout. Each offset points to the start of a descriptor set’s data within the bound descriptor buffer.

Provided by VK_KHR_maintenance6 (for the pNext-extensible variant of the VK_EXT_descriptor_buffer command).

pub unsafe fn cmd_bind_descriptor_buffer_embedded_samplers2_ext( &self, command_buffer: CommandBuffer, p_bind_descriptor_buffer_embedded_samplers_info: &BindDescriptorBufferEmbeddedSamplersInfoEXT, )

Wraps vkCmdBindDescriptorBufferEmbeddedSamplers2EXT. Provided by VK_KHR_maintenance6.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdBindDescriptorBufferEmbeddedSamplers2EXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Extended version of cmd_bind_descriptor_buffer_embedded_samplers_ext that takes a BindDescriptorBufferEmbeddedSamplersInfoEXT struct with pNext extensibility.

Binds embedded immutable samplers from a descriptor set layout that was created with CREATE_EMBEDDED_IMMUTABLE_SAMPLERS_BIT.

Provided by VK_KHR_maintenance6 (for the pNext-extensible variant of the VK_EXT_descriptor_buffer command).

pub unsafe fn set_latency_sleep_mode_nv( &self, swapchain: SwapchainKHR, p_sleep_mode_info: &LatencySleepModeInfoNV, ) -> VkResult<()>

Wraps vkSetLatencySleepModeNV. Provided by VK_NV_low_latency2.

Errors

• VK_ERROR_INITIALIZATION_FAILED
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkSetLatencySleepModeNV was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Configures the NVIDIA Reflex low-latency sleep mode for a swapchain. Enables or disables latency sleep and sets the target sleep duration. Call before latency_sleep_nv to activate the system.

Requires VK_NV_low_latency2.

pub unsafe fn latency_sleep_nv( &self, swapchain: SwapchainKHR, p_sleep_info: &LatencySleepInfoNV, ) -> VkResult<()>

Wraps vkLatencySleepNV. Provided by VK_NV_low_latency2.

Errors

• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkLatencySleepNV was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Sleeps the calling thread until the optimal time to begin the next frame, as determined by the NVIDIA Reflex low-latency system. Reduces input-to-display latency by preventing the CPU from running too far ahead of the GPU.

Requires VK_NV_low_latency2.

pub unsafe fn set_latency_marker_nv( &self, swapchain: SwapchainKHR, p_latency_marker_info: &SetLatencyMarkerInfoNV, )

Wraps vkSetLatencyMarkerNV. Provided by VK_NV_low_latency2.

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkSetLatencyMarkerNV was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Sets a latency marker at a specific point in the frame lifecycle (simulation start, render start, present, etc.). The markers are later retrieved with get_latency_timings_nv to measure per- stage latency.

Requires VK_NV_low_latency2.

pub unsafe fn get_latency_timings_nv( &self, swapchain: SwapchainKHR, p_latency_marker_info: &mut GetLatencyMarkerInfoNV, )

Wraps vkGetLatencyTimingsNV. Provided by VK_NV_low_latency2.

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetLatencyTimingsNV was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Retrieves per-frame latency timing data for a swapchain. Returns timestamps for each marker set with set_latency_marker_nv, enabling measurement of simulation, render, and present latency.

Requires VK_NV_low_latency2.

pub unsafe fn queue_notify_out_of_band_nv( &self, queue: Queue, p_queue_type_info: &OutOfBandQueueTypeInfoNV, )

Wraps vkQueueNotifyOutOfBandNV. Provided by VK_NV_low_latency2.

Safety

• queue (self) must be valid and not destroyed.

Panics

Panics if vkQueueNotifyOutOfBandNV was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Notifies the low-latency system that a queue submission is out-of-band (e.g., a loading or async compute submission that should not be counted toward frame latency tracking).

Requires VK_NV_low_latency2.

pub unsafe fn cmd_set_rendering_attachment_locations( &self, command_buffer: CommandBuffer, p_location_info: &RenderingAttachmentLocationInfo, )

Wraps vkCmdSetRenderingAttachmentLocations. Provided by VK_GRAPHICS_VERSION_1_4.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdSetRenderingAttachmentLocations was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Dynamically remaps colour attachment indices within a dynamic rendering instance. Allows fragment shader outputs to target different attachment slots without changing the pipeline.

This is useful when the same shader outputs different data to different attachments depending on the rendering pass (e.g. G-buffer vs forward).

Requires dynamic_rendering_local_read feature. Core in Vulkan 1.4.

pub unsafe fn cmd_set_rendering_input_attachment_indices( &self, command_buffer: CommandBuffer, p_input_attachment_index_info: &RenderingInputAttachmentIndexInfo, )

Wraps vkCmdSetRenderingInputAttachmentIndices. Provided by VK_GRAPHICS_VERSION_1_4.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdSetRenderingInputAttachmentIndices was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Dynamically remaps input attachment indices within a dynamic rendering instance. Allows fragment shaders to read from different colour or depth/stencil attachments without changing the pipeline.

Paired with cmd_set_rendering_attachment_locations to enable flexible attachment routing in multi-pass rendering with dynamic rendering.

Requires dynamic_rendering_local_read feature. Core in Vulkan 1.4.

pub unsafe fn cmd_set_depth_clamp_range_ext( &self, command_buffer: CommandBuffer, depth_clamp_mode: DepthClampModeEXT, p_depth_clamp_range: Option<&DepthClampRangeEXT>, )

Wraps vkCmdSetDepthClampRangeEXT. Provided by VK_EXT_shader_object.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdSetDepthClampRangeEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Dynamically sets the depth clamp range. When depth clamping is enabled, fragments are clamped to the specified min/max depth values instead of the viewport near/far range.

Pass null to use the default viewport depth range for clamping.

Requires VK_EXT_depth_clamp_control and the depthClampControl feature.

pub unsafe fn get_memory_metal_handle_ext( &self, p_get_metal_handle_info: &MemoryGetMetalHandleInfoEXT, p_handle: *mut *mut c_void, ) -> VkResult<()>

Wraps vkGetMemoryMetalHandleEXT. Provided by VK_EXT_external_memory_metal.

Errors

• VK_ERROR_TOO_MANY_OBJECTS
• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetMemoryMetalHandleEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Exports a Vulkan device memory allocation as a Metal resource handle for cross-API interop on Apple platforms.

Requires VK_EXT_external_memory_metal. macOS/iOS only.

pub unsafe fn get_memory_metal_handle_properties_ext( &self, handle_type: ExternalMemoryHandleTypeFlagBits, p_handle: *const c_void, p_memory_metal_handle_properties: &mut MemoryMetalHandlePropertiesEXT, ) -> VkResult<()>

Wraps vkGetMemoryMetalHandlePropertiesEXT. Provided by VK_EXT_external_memory_metal.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_INVALID_EXTERNAL_HANDLE
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetMemoryMetalHandlePropertiesEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Queries which Vulkan memory types are compatible with importing a Metal resource handle as external memory. Use before allocating device memory to determine valid memory type bits.

Requires VK_EXT_external_memory_metal. macOS/iOS only.

pub unsafe fn convert_cooperative_vector_matrix_nv( &self, p_info: &ConvertCooperativeVectorMatrixInfoNV, ) -> VkResult<()>

Wraps vkConvertCooperativeVectorMatrixNV. Provided by VK_NV_cooperative_vector.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkConvertCooperativeVectorMatrixNV was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Host-side conversion of cooperative vector matrix data between formats or layouts. Use to prepare weight matrices on the CPU before uploading to the GPU for cooperative vector operations.

Requires VK_NV_cooperative_vector.

pub unsafe fn cmd_convert_cooperative_vector_matrix_nv( &self, command_buffer: CommandBuffer, p_infos: &[ConvertCooperativeVectorMatrixInfoNV], )

Wraps vkCmdConvertCooperativeVectorMatrixNV. Provided by VK_NV_cooperative_vector.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdConvertCooperativeVectorMatrixNV was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

GPU-side conversion of cooperative vector matrix data between formats or layouts. Converts one or more matrices in a command buffer. Use for repacking weight matrices for neural network inference on the GPU.

Requires VK_NV_cooperative_vector.

pub unsafe fn cmd_dispatch_tile_qcom( &self, command_buffer: CommandBuffer, p_dispatch_tile_info: &DispatchTileInfoQCOM, )

Wraps vkCmdDispatchTileQCOM. Provided by VK_QCOM_tile_shading.

Safety

• commandBuffer (self) must be valid and not destroyed.

Panics

Panics if vkCmdDispatchTileQCOM was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Dispatches compute work within a per-tile execution region on Qualcomm tile-based GPUs. Must be called between cmd_begin_per_tile_execution_qcom and cmd_end_per_tile_execution_qcom.

Requires VK_QCOM_tile_shading.

pub unsafe fn cmd_begin_per_tile_execution_qcom( &self, command_buffer: CommandBuffer, p_per_tile_begin_info: &PerTileBeginInfoQCOM, )

Wraps vkCmdBeginPerTileExecutionQCOM. Provided by VK_QCOM_tile_shading.

Safety

• commandBuffer (self) must be valid and not destroyed.

Panics

Panics if vkCmdBeginPerTileExecutionQCOM was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Begins a per-tile execution region within a render pass on Qualcomm tile-based GPUs. Commands recorded between this and cmd_end_per_tile_execution_qcom are executed once per tile, enabling tile-local compute and shading optimisations.

Requires VK_QCOM_tile_shading.

pub unsafe fn cmd_end_per_tile_execution_qcom( &self, command_buffer: CommandBuffer, p_per_tile_end_info: &PerTileEndInfoQCOM, )

Wraps vkCmdEndPerTileExecutionQCOM. Provided by VK_QCOM_tile_shading.

Safety

• commandBuffer (self) must be valid and not destroyed.

Panics

Panics if vkCmdEndPerTileExecutionQCOM was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Ends a per-tile execution region started by cmd_begin_per_tile_execution_qcom. After this call, the command buffer returns to normal (non-tile-local) recording.

Requires VK_QCOM_tile_shading.

pub unsafe fn create_external_compute_queue_nv( &self, p_create_info: &ExternalComputeQueueCreateInfoNV, allocator: Option<&AllocationCallbacks>, ) -> VkResult<ExternalComputeQueueNV>

Wraps vkCreateExternalComputeQueueNV. Provided by VK_NV_external_compute_queue.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_TOO_MANY_OBJECTS
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkCreateExternalComputeQueueNV was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Creates an external compute queue that can be used to submit work from outside the Vulkan runtime (e.g., CUDA interop). Destroy with destroy_external_compute_queue_nv.

Requires VK_NV_external_compute_queue.

pub unsafe fn destroy_external_compute_queue_nv( &self, external_queue: ExternalComputeQueueNV, allocator: Option<&AllocationCallbacks>, )

Wraps vkDestroyExternalComputeQueueNV. Provided by VK_NV_external_compute_queue.

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkDestroyExternalComputeQueueNV was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Destroys an external compute queue created with create_external_compute_queue_nv.

Requires VK_NV_external_compute_queue.

pub unsafe fn create_shader_instrumentation_arm( &self, p_create_info: &ShaderInstrumentationCreateInfoARM, allocator: Option<&AllocationCallbacks>, ) -> VkResult<ShaderInstrumentationARM>

Wraps vkCreateShaderInstrumentationARM. Provided by VK_ARM_shader_instrumentation.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkCreateShaderInstrumentationARM was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Creates a shader instrumentation object that configures which metrics to collect during shader execution. The instrumentation is later bound to command buffers with cmd_begin_shader_instrumentation_arm.

Requires VK_ARM_shader_instrumentation.

pub unsafe fn destroy_shader_instrumentation_arm( &self, instrumentation: ShaderInstrumentationARM, allocator: Option<&AllocationCallbacks>, )

Wraps vkDestroyShaderInstrumentationARM. Provided by VK_ARM_shader_instrumentation.

Safety

• device (self) must be valid and not destroyed.
• instrumentation must be externally synchronized.

Panics

Panics if vkDestroyShaderInstrumentationARM was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Destroys a shader instrumentation object. The object must not be in use by any command buffer.

Requires VK_ARM_shader_instrumentation.

pub unsafe fn cmd_begin_shader_instrumentation_arm( &self, command_buffer: CommandBuffer, instrumentation: ShaderInstrumentationARM, )

Wraps vkCmdBeginShaderInstrumentationARM. Provided by VK_ARM_shader_instrumentation.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.
• instrumentation must be externally synchronized.

Panics

Panics if vkCmdBeginShaderInstrumentationARM was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Begins shader instrumentation collection in a command buffer. All subsequent draw and dispatch commands will collect the metrics configured in the instrumentation object until cmd_end_shader_instrumentation_arm is called.

Requires VK_ARM_shader_instrumentation.

pub unsafe fn cmd_end_shader_instrumentation_arm( &self, command_buffer: CommandBuffer, )

Wraps vkCmdEndShaderInstrumentationARM. Provided by VK_ARM_shader_instrumentation.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdEndShaderInstrumentationARM was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Ends shader instrumentation collection in a command buffer. Metrics collected since the matching cmd_begin_shader_instrumentation_arm can be retrieved with get_shader_instrumentation_values_arm after submission completes.

Requires VK_ARM_shader_instrumentation.

pub unsafe fn get_shader_instrumentation_values_arm( &self, instrumentation: ShaderInstrumentationARM, p_metric_block_count: *mut u32, flags: ShaderInstrumentationValuesFlagsARM, ) -> VkResult<c_void>

Wraps vkGetShaderInstrumentationValuesARM. Provided by VK_ARM_shader_instrumentation.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetShaderInstrumentationValuesARM was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Retrieves collected shader instrumentation metric values after instrumented commands have completed execution. Returns metric blocks whose count is written to p_metric_block_count. Ensure the instrumented submission has finished before querying.

Requires VK_ARM_shader_instrumentation.

pub unsafe fn clear_shader_instrumentation_metrics_arm( &self, instrumentation: ShaderInstrumentationARM, )

Wraps vkClearShaderInstrumentationMetricsARM. Provided by VK_ARM_shader_instrumentation.

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkClearShaderInstrumentationMetricsARM was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Resets accumulated metrics for a shader instrumentation object. Call between frames or profiling sessions to start collecting fresh data without destroying and recreating the object.

Requires VK_ARM_shader_instrumentation.

pub unsafe fn create_tensor_arm( &self, p_create_info: &TensorCreateInfoARM, allocator: Option<&AllocationCallbacks>, ) -> VkResult<TensorARM>

Wraps vkCreateTensorARM. Provided by VK_ARM_tensors.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkCreateTensorARM was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Creates a tensor object for ARM’s tensor extension. Tensors are multi-dimensional arrays with a defined format, dimensions, and usage flags. Must be bound to memory before use, similar to images and buffers.

Requires VK_ARM_tensors.

pub unsafe fn destroy_tensor_arm( &self, tensor: TensorARM, allocator: Option<&AllocationCallbacks>, )

Wraps vkDestroyTensorARM. Provided by VK_ARM_tensors.

Safety

• device (self) must be valid and not destroyed.
• tensor must be externally synchronized.

Panics

Panics if vkDestroyTensorARM was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Destroys a tensor object. The tensor must not be in use by any command buffer or referenced by any tensor view.

Requires VK_ARM_tensors.

pub unsafe fn create_tensor_view_arm( &self, p_create_info: &TensorViewCreateInfoARM, allocator: Option<&AllocationCallbacks>, ) -> VkResult<TensorViewARM>

Wraps vkCreateTensorViewARM. Provided by VK_ARM_tensors.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkCreateTensorViewARM was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Creates a view into a tensor, analogous to image views. A tensor view selects a subset of the tensor’s dimensions or reinterprets its format for use in descriptors and shaders.

Requires VK_ARM_tensors.

pub unsafe fn destroy_tensor_view_arm( &self, tensor_view: TensorViewARM, allocator: Option<&AllocationCallbacks>, )

Wraps vkDestroyTensorViewARM. Provided by VK_ARM_tensors.

Safety

• device (self) must be valid and not destroyed.
• tensorView must be externally synchronized.

Panics

Panics if vkDestroyTensorViewARM was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Destroys a tensor view. The view must not be in use by any command buffer or referenced by any descriptor set.

Requires VK_ARM_tensors.

pub unsafe fn get_tensor_memory_requirements_arm( &self, p_info: &TensorMemoryRequirementsInfoARM, p_memory_requirements: &mut MemoryRequirements2, )

Wraps vkGetTensorMemoryRequirementsARM. Provided by VK_ARM_tensors.

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetTensorMemoryRequirementsARM was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Queries the memory requirements (size, alignment, memory type bits) for an existing tensor object. Call before bind_tensor_memory_arm to determine the allocation needed.

Requires VK_ARM_tensors.

pub unsafe fn bind_tensor_memory_arm( &self, p_bind_infos: &[BindTensorMemoryInfoARM], ) -> VkResult<()>

Wraps vkBindTensorMemoryARM. Provided by VK_ARM_tensors.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkBindTensorMemoryARM was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Binds device memory to one or more tensors. Each bind info specifies the tensor, memory, and offset. Must be called before the tensor is used in any command. Similar to bind_buffer_memory2 / bind_image_memory2.

Requires VK_ARM_tensors.

pub unsafe fn get_device_tensor_memory_requirements_arm( &self, p_info: &DeviceTensorMemoryRequirementsARM, p_memory_requirements: &mut MemoryRequirements2, )

Wraps vkGetDeviceTensorMemoryRequirementsARM. Provided by VK_ARM_tensors.

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetDeviceTensorMemoryRequirementsARM was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Queries memory requirements for a tensor described by its create info, without creating the tensor first. Analogous to get_device_buffer_memory_requirements / get_device_image_memory_requirements.

Requires VK_ARM_tensors.

pub unsafe fn cmd_copy_tensor_arm( &self, command_buffer: CommandBuffer, p_copy_tensor_info: &CopyTensorInfoARM, )

Wraps vkCmdCopyTensorARM. Provided by VK_ARM_tensors.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdCopyTensorARM was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Copies data between tensors or between a tensor and a buffer. The copy info structure specifies the source and destination regions, similar to cmd_copy_buffer or cmd_copy_image.

Requires VK_ARM_tensors.

pub unsafe fn get_tensor_opaque_capture_descriptor_data_arm( &self, p_info: &TensorCaptureDescriptorDataInfoARM, ) -> VkResult<c_void>

Wraps vkGetTensorOpaqueCaptureDescriptorDataARM. Provided by VK_ARM_tensors.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetTensorOpaqueCaptureDescriptorDataARM was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Retrieves opaque capture data for a tensor’s descriptor, used for descriptor buffer capture and replay. The returned data can be stored and replayed to recreate an equivalent descriptor.

Requires VK_ARM_tensors + VK_EXT_descriptor_buffer.

pub unsafe fn get_tensor_view_opaque_capture_descriptor_data_arm( &self, p_info: &TensorViewCaptureDescriptorDataInfoARM, ) -> VkResult<c_void>

Wraps vkGetTensorViewOpaqueCaptureDescriptorDataARM. Provided by VK_ARM_tensors.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetTensorViewOpaqueCaptureDescriptorDataARM was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Retrieves opaque capture data for a tensor view’s descriptor, used for descriptor buffer capture and replay. Similar to get_tensor_opaque_capture_descriptor_data_arm but for tensor views.

Requires VK_ARM_tensors + VK_EXT_descriptor_buffer.

pub unsafe fn create_data_graph_pipelines_arm( &self, deferred_operation: DeferredOperationKHR, pipeline_cache: PipelineCache, p_create_infos: &[DataGraphPipelineCreateInfoARM], allocator: Option<&AllocationCallbacks>, ) -> VkResult<Vec<Pipeline>>

Wraps vkCreateDataGraphPipelinesARM. Provided by VK_ARM_data_graph.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkCreateDataGraphPipelinesARM was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Creates one or more data graph pipelines for ARM’s data graph extension. Supports deferred compilation via a DeferredOperationKHR handle and pipeline caching. Data graph pipelines define GPU-side data processing graphs.

Requires VK_ARM_data_graph.

pub unsafe fn create_data_graph_pipeline_session_arm( &self, p_create_info: &DataGraphPipelineSessionCreateInfoARM, allocator: Option<&AllocationCallbacks>, ) -> VkResult<DataGraphPipelineSessionARM>

Wraps vkCreateDataGraphPipelineSessionARM. Provided by VK_ARM_data_graph.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkCreateDataGraphPipelineSessionARM was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Creates a session for executing a data graph pipeline. A session holds the runtime state and memory bindings needed to dispatch the graph. Must be bound to memory before dispatch.

Requires VK_ARM_data_graph.

pub unsafe fn get_data_graph_pipeline_session_bind_point_requirements_arm( &self, p_info: &DataGraphPipelineSessionBindPointRequirementsInfoARM, ) -> VkResult<Vec<DataGraphPipelineSessionBindPointRequirementARM>>

Wraps vkGetDataGraphPipelineSessionBindPointRequirementsARM. Provided by VK_ARM_data_graph.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetDataGraphPipelineSessionBindPointRequirementsARM was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Queries the memory bind point requirements for a data graph pipeline session. Uses the two-call idiom. Each returned requirement describes a bind point that must be satisfied with bind_data_graph_pipeline_session_memory_arm before dispatch.

Requires VK_ARM_data_graph.

pub unsafe fn get_data_graph_pipeline_session_memory_requirements_arm( &self, p_info: &DataGraphPipelineSessionMemoryRequirementsInfoARM, p_memory_requirements: &mut MemoryRequirements2, )

Wraps vkGetDataGraphPipelineSessionMemoryRequirementsARM. Provided by VK_ARM_data_graph.

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetDataGraphPipelineSessionMemoryRequirementsARM was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Queries the memory requirements for a specific bind point within a data graph pipeline session. Returns a MemoryRequirements2 describing the size, alignment, and compatible memory types.

Requires VK_ARM_data_graph.

pub unsafe fn bind_data_graph_pipeline_session_memory_arm( &self, p_bind_infos: &[BindDataGraphPipelineSessionMemoryInfoARM], ) -> VkResult<()>

Wraps vkBindDataGraphPipelineSessionMemoryARM. Provided by VK_ARM_data_graph.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkBindDataGraphPipelineSessionMemoryARM was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Binds device memory to a data graph pipeline session at the bind points returned by get_data_graph_pipeline_session_bind_point_requirements_arm. Must be called before dispatching the session.

Requires VK_ARM_data_graph.

pub unsafe fn destroy_data_graph_pipeline_session_arm( &self, session: DataGraphPipelineSessionARM, allocator: Option<&AllocationCallbacks>, )

Wraps vkDestroyDataGraphPipelineSessionARM. Provided by VK_ARM_data_graph.

Safety

• device (self) must be valid and not destroyed.
• session must be externally synchronized.

Panics

Panics if vkDestroyDataGraphPipelineSessionARM was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Destroys a data graph pipeline session and frees associated host resources. The session must not be in use by any command buffer.

Requires VK_ARM_data_graph.

pub unsafe fn cmd_dispatch_data_graph_arm( &self, command_buffer: CommandBuffer, session: DataGraphPipelineSessionARM, p_info: Option<&DataGraphPipelineDispatchInfoARM>, )

Wraps vkCmdDispatchDataGraphARM. Provided by VK_ARM_data_graph.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdDispatchDataGraphARM was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Records a data graph pipeline dispatch into a command buffer. The session must have been created and bound to memory before dispatch. Optional dispatch info can configure execution parameters.

Requires VK_ARM_data_graph.

pub unsafe fn get_data_graph_pipeline_available_properties_arm( &self, p_pipeline_info: &DataGraphPipelineInfoARM, ) -> VkResult<Vec<DataGraphPipelinePropertyARM>>

Wraps vkGetDataGraphPipelineAvailablePropertiesARM. Provided by VK_ARM_data_graph.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetDataGraphPipelineAvailablePropertiesARM was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Enumerates the queryable properties for a data graph pipeline. Uses the two-call idiom. The returned property descriptors can then be queried with get_data_graph_pipeline_properties_arm.

Requires VK_ARM_data_graph.

pub unsafe fn get_data_graph_pipeline_properties_arm( &self, p_pipeline_info: &DataGraphPipelineInfoARM, properties_count: u32, p_properties: *mut DataGraphPipelinePropertyQueryResultARM, ) -> VkResult<()>

Wraps vkGetDataGraphPipelinePropertiesARM. Provided by VK_ARM_data_graph.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetDataGraphPipelinePropertiesARM was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Queries specific property values for a data graph pipeline. First enumerate available properties with get_data_graph_pipeline_available_properties_arm, then query the ones you need.

Requires VK_ARM_data_graph.

pub unsafe fn get_native_buffer_properties_ohos( &self, buffer: *const c_void, p_properties: &mut NativeBufferPropertiesOHOS, ) -> VkResult<()>

Wraps vkGetNativeBufferPropertiesOHOS. Provided by VK_OHOS_external_memory.

Errors

• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_INVALID_EXTERNAL_HANDLE_KHR
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetNativeBufferPropertiesOHOS was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Queries Vulkan memory properties (memory type bits, size) for an OHOS native buffer. Use before importing external memory to determine compatible memory types. OHOS only.

Requires VK_OHOS_external_memory.

pub unsafe fn get_memory_native_buffer_ohos( &self, p_info: &MemoryGetNativeBufferInfoOHOS, p_buffer: *mut *mut c_void, ) -> VkResult<()>

Wraps vkGetMemoryNativeBufferOHOS. Provided by VK_OHOS_external_memory.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetMemoryNativeBufferOHOS was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Exports a Vulkan device memory allocation as an OHOS native buffer handle for sharing with other OpenHarmony services. OHOS only.

Requires VK_OHOS_external_memory.

pub unsafe fn get_swapchain_gralloc_usage_ohos( &self, format: Format, image_usage: ImageUsageFlags, ) -> VkResult<u64>

Wraps vkGetSwapchainGrallocUsageOHOS.

Errors

• VK_ERROR_INITIALIZATION_FAILED
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetSwapchainGrallocUsageOHOS was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Queries the OHOS gralloc usage flags needed for swapchain images with the given format and Vulkan image usage. Used internally by the OHOS WSI implementation. OHOS only.

Requires VK_OHOS_native_buffer.

pub unsafe fn acquire_image_ohos( &self, image: Image, native_fence_fd: i32, semaphore: Semaphore, fence: Fence, ) -> VkResult<()>

Wraps vkAcquireImageOHOS.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkAcquireImageOHOS was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Acquires ownership of a swapchain image on OpenHarmony OS. Takes a native fence file descriptor for synchronisation and can signal a Vulkan semaphore or fence on completion. OHOS only.

Requires VK_OHOS_native_buffer.

pub unsafe fn queue_signal_release_image_ohos( &self, queue: Queue, p_wait_semaphores: &[Semaphore], image: Image, ) -> VkResult<i32>

Wraps vkQueueSignalReleaseImageOHOS.

Errors

• VK_ERROR_INITIALIZATION_FAILED
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• queue (self) must be valid and not destroyed.

Panics

Panics if vkQueueSignalReleaseImageOHOS was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Releases a swapchain image back to the OHOS compositor after rendering. Waits on the given semaphores and returns a native fence FD for external synchronisation. OHOS only.

Requires VK_OHOS_native_buffer.

pub unsafe fn cmd_set_compute_occupancy_priority_nv( &self, command_buffer: CommandBuffer, p_parameters: &ComputeOccupancyPriorityParametersNV, )

Wraps vkCmdSetComputeOccupancyPriorityNV. Provided by VK_NV_compute_occupancy_priority.

Safety

• commandBuffer (self) must be valid and not destroyed.

Panics

Panics if vkCmdSetComputeOccupancyPriorityNV was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Sets the compute occupancy priority for subsequent dispatch commands. Higher priority may increase the number of warps resident on an SM, trading off per-warp resources for greater parallelism.

Requires VK_NV_compute_occupancy_priority.

pub unsafe fn write_sampler_descriptors_ext( &self, p_samplers: &[SamplerCreateInfo], p_descriptors: &HostAddressRangeEXT, ) -> VkResult<()>

Wraps vkWriteSamplerDescriptorsEXT. Provided by VK_EXT_descriptor_heap.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkWriteSamplerDescriptorsEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Writes sampler descriptors to host-visible memory. This is the descriptor heap equivalent for sampler descriptors, instead of using descriptor sets, samplers are written directly to heap memory.

Provided by VK_EXT_descriptor_heap.

pub unsafe fn write_resource_descriptors_ext( &self, p_resources: &[ResourceDescriptorInfoEXT], p_descriptors: &HostAddressRangeEXT, ) -> VkResult<()>

Wraps vkWriteResourceDescriptorsEXT. Provided by VK_EXT_descriptor_heap.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkWriteResourceDescriptorsEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Writes resource descriptors (buffers, images, acceleration structures) to host-visible memory. This is the descriptor heap equivalent of writing descriptors, instead of using descriptor sets, descriptors are written directly to heap memory.

Provided by VK_EXT_descriptor_heap.

pub unsafe fn cmd_bind_sampler_heap_ext( &self, command_buffer: CommandBuffer, p_bind_info: &BindHeapInfoEXT, )

Wraps vkCmdBindSamplerHeapEXT. Provided by VK_EXT_descriptor_heap.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdBindSamplerHeapEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Binds a sampler descriptor heap for use in subsequent draw and dispatch commands. The BindHeapInfoEXT specifies the heap to bind.

Sampler heaps hold sampler descriptors. Resource descriptors are bound separately with cmd_bind_resource_heap_ext.

Provided by VK_EXT_descriptor_heap.

pub unsafe fn cmd_bind_resource_heap_ext( &self, command_buffer: CommandBuffer, p_bind_info: &BindHeapInfoEXT, )

Wraps vkCmdBindResourceHeapEXT. Provided by VK_EXT_descriptor_heap.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdBindResourceHeapEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Binds a resource descriptor heap for use in subsequent draw and dispatch commands. The BindHeapInfoEXT specifies the heap to bind.

Resource heaps hold descriptors for buffers, images, and acceleration structures. Samplers are bound separately with cmd_bind_sampler_heap_ext.

Provided by VK_EXT_descriptor_heap.

pub unsafe fn cmd_push_data_ext( &self, command_buffer: CommandBuffer, p_push_data_info: &PushDataInfoEXT, )

Wraps vkCmdPushDataEXT. Provided by VK_EXT_descriptor_heap.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdPushDataEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Pushes inline data to the command buffer for use in shaders. The PushDataInfoEXT specifies the pipeline layout, stage flags, offset, and data bytes.

Similar to push constants but used with the descriptor heap model. Data is accessible in shaders via the push data mechanism.

Provided by VK_EXT_descriptor_heap.

pub unsafe fn register_custom_border_color_ext( &self, p_border_color: &SamplerCustomBorderColorCreateInfoEXT, request_index: bool, ) -> VkResult<u32>

Wraps vkRegisterCustomBorderColorEXT. Provided by VK_EXT_descriptor_heap.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_TOO_MANY_OBJECTS
• VK_ERROR_INVALID_OPAQUE_CAPTURE_ADDRESS
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkRegisterCustomBorderColorEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Registers a custom border color for use with samplers. Returns a handle that can be referenced when creating samplers with BORDER_COLOR_CUSTOM modes.

The device has a limited number of custom border color slots (query maxCustomBorderColors).

Unregister with unregister_custom_border_color_ext when no longer needed.

Provided by VK_EXT_descriptor_heap.

pub unsafe fn unregister_custom_border_color_ext(&self, index: u32)

Wraps vkUnregisterCustomBorderColorEXT. Provided by VK_EXT_descriptor_heap.

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkUnregisterCustomBorderColorEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Unregisters a custom border color previously registered with register_custom_border_color_ext, freeing the slot for reuse.

Ensure no samplers referencing this border color are in use before unregistering.

Provided by VK_EXT_descriptor_heap.

pub unsafe fn get_image_opaque_capture_data_ext( &self, p_images: &[Image], ) -> VkResult<Vec<HostAddressRangeEXT>>

Wraps vkGetImageOpaqueCaptureDataEXT. Provided by VK_EXT_descriptor_heap.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetImageOpaqueCaptureDataEXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Retrieves opaque capture data for one or more images. The returned HostAddressRangeEXT data can be used to reconstruct image resource bindings during capture/replay.

Provided by VK_EXT_descriptor_heap.

pub unsafe fn get_tensor_opaque_capture_data_arm( &self, p_tensors: &[TensorARM], ) -> VkResult<Vec<HostAddressRangeEXT>>

Wraps vkGetTensorOpaqueCaptureDataARM. Provided by VK_EXT_descriptor_heap.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_OUT_OF_DEVICE_MEMORY
• VK_ERROR_UNKNOWN
• VK_ERROR_VALIDATION_FAILED

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkGetTensorOpaqueCaptureDataARM was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Retrieves opaque capture data for one or more ARM tensors. The returned data can be used to reconstruct tensor resource bindings during capture/replay.

Provided by VK_ARM_tensors.

pub unsafe fn cmd_copy_memory_khr( &self, command_buffer: CommandBuffer, p_copy_memory_info: Option<&CopyDeviceMemoryInfoKHR>, )

Wraps vkCmdCopyMemoryKHR. Provided by VK_KHR_device_address_commands.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdCopyMemoryKHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Copies data between two device address ranges on the GPU. This is the device-address equivalent of cmd_copy_buffer, instead of buffer handles, source and destination are specified as device addresses in CopyDeviceMemoryInfoKHR.

Useful for copying data between arbitrary device memory locations without needing buffer objects.

Requires VK_KHR_device_address_commands.

pub unsafe fn cmd_copy_memory_to_image_khr( &self, command_buffer: CommandBuffer, p_copy_memory_info: Option<&CopyDeviceMemoryImageInfoKHR>, )

Wraps vkCmdCopyMemoryToImageKHR. Provided by VK_KHR_device_address_commands.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdCopyMemoryToImageKHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Device-address variant of cmd_copy_buffer_to_image. Copies data from a device address range into an image, instead of reading from a buffer handle.

The CopyDeviceMemoryImageInfoKHR struct specifies the source device address, destination image, and region descriptions.

This is the device-address counterpart, for the host-side equivalent, see copy_memory_to_image (core 1.4).

Requires VK_KHR_device_address_commands.

pub unsafe fn cmd_copy_image_to_memory_khr( &self, command_buffer: CommandBuffer, p_copy_memory_info: Option<&CopyDeviceMemoryImageInfoKHR>, )

Wraps vkCmdCopyImageToMemoryKHR. Provided by VK_KHR_device_address_commands.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdCopyImageToMemoryKHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Device-address variant of cmd_copy_image_to_buffer. Copies image texel data to a device address range instead of a buffer handle.

The CopyDeviceMemoryImageInfoKHR struct specifies the source image, destination device address, and region descriptions.

This is the device-address counterpart, for the host-side equivalent, see copy_image_to_memory (core 1.4).

Requires VK_KHR_device_address_commands.

pub unsafe fn cmd_update_memory_khr( &self, command_buffer: CommandBuffer, p_dst_range: &DeviceAddressRangeKHR, dst_flags: AddressCommandFlagsKHR, data_size: u64, p_data: *const c_void, )

Wraps vkCmdUpdateMemoryKHR. Provided by VK_KHR_device_address_commands.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdUpdateMemoryKHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Device-address variant of cmd_update_buffer. Writes a small amount of inline data to a device address range, instead of targeting a buffer handle.

data_size must be ≤ 65536 bytes and a multiple of 4. For larger transfers, use cmd_copy_memory_khr.

The DeviceAddressRangeKHR specifies the destination address.

Requires VK_KHR_device_address_commands.

pub unsafe fn cmd_fill_memory_khr( &self, command_buffer: CommandBuffer, p_dst_range: &DeviceAddressRangeKHR, dst_flags: AddressCommandFlagsKHR, data: u32, )

Wraps vkCmdFillMemoryKHR. Provided by VK_KHR_device_address_commands.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdFillMemoryKHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Device-address variant of cmd_fill_buffer. Fills a device address range with a repeating 32-bit data value, instead of targeting a buffer handle.

The DeviceAddressRangeKHR specifies the destination address and size. The fill size must be a multiple of 4 bytes.

Requires VK_KHR_device_address_commands.

pub unsafe fn cmd_copy_query_pool_results_to_memory_khr( &self, command_buffer: CommandBuffer, query_pool: QueryPool, first_query: u32, query_count: u32, p_dst_range: &StridedDeviceAddressRangeKHR, dst_flags: AddressCommandFlagsKHR, query_result_flags: QueryResultFlags, )

Wraps vkCmdCopyQueryPoolResultsToMemoryKHR. Provided by VK_KHR_device_address_commands.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdCopyQueryPoolResultsToMemoryKHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Device-address variant of cmd_copy_query_pool_results. Copies query results directly to a device address range instead of a buffer handle.

first_query and query_count select which queries to copy. query_result_flags controls formatting (_64_BIT, WAIT, WITH_AVAILABILITY, PARTIAL).

The StridedDeviceAddressRangeKHR specifies the destination address and stride between results.

Requires VK_KHR_device_address_commands.

pub unsafe fn cmd_begin_conditional_rendering2_ext( &self, command_buffer: CommandBuffer, p_conditional_rendering_begin: &ConditionalRenderingBeginInfo2EXT, )

Wraps vkCmdBeginConditionalRendering2EXT. Provided by VK_KHR_device_address_commands.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdBeginConditionalRendering2EXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Device-address variant of cmd_begin_conditional_rendering_ext. Instead of a buffer handle + offset, the condition is read from a DeviceAddress specified in ConditionalRenderingBeginInfo2EXT.

When the 32-bit value at the address is zero, subsequent rendering and dispatch commands are discarded (or the inverse, if INVERTED is set). End the conditional block with cmd_end_conditional_rendering_ext.

Requires VK_KHR_device_address_commands and VK_EXT_conditional_rendering.

pub unsafe fn cmd_bind_transform_feedback_buffers2_ext( &self, command_buffer: CommandBuffer, first_binding: u32, p_binding_infos: &[BindTransformFeedbackBuffer2InfoEXT], )

Wraps vkCmdBindTransformFeedbackBuffers2EXT. Provided by VK_KHR_device_address_commands.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdBindTransformFeedbackBuffers2EXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Device-address variant of cmd_bind_transform_feedback_buffers_ext. Binds transform feedback output buffers using device addresses instead of buffer handles.

Each BindTransformFeedbackBuffer2InfoEXT specifies a device address and size for one binding slot starting at first_binding.

Requires VK_KHR_device_address_commands and VK_EXT_transform_feedback.

pub unsafe fn cmd_begin_transform_feedback2_ext( &self, command_buffer: CommandBuffer, first_counter_range: u32, p_counter_infos: &[BindTransformFeedbackBuffer2InfoEXT], )

Wraps vkCmdBeginTransformFeedback2EXT. Provided by VK_KHR_device_address_commands.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdBeginTransformFeedback2EXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Device-address variant of cmd_begin_transform_feedback_ext. Activates transform feedback using counter buffers specified via device addresses in BindTransformFeedbackBuffer2InfoEXT rather than buffer handles.

first_counter_range and the info array identify which transform feedback counter ranges to resume from. Pass empty counter infos to start from offset zero.

End the transform feedback pass with cmd_end_transform_feedback2_ext.

Requires VK_KHR_device_address_commands and VK_EXT_transform_feedback.

pub unsafe fn cmd_end_transform_feedback2_ext( &self, command_buffer: CommandBuffer, first_counter_range: u32, p_counter_infos: &[BindTransformFeedbackBuffer2InfoEXT], )

Wraps vkCmdEndTransformFeedback2EXT. Provided by VK_KHR_device_address_commands.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdEndTransformFeedback2EXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Device-address variant of cmd_end_transform_feedback_ext. Stops transform feedback and writes counter values to device addresses specified in BindTransformFeedbackBuffer2InfoEXT.

These saved counter values can be passed to cmd_begin_transform_feedback2_ext to resume feedback in a later render pass.

Requires VK_KHR_device_address_commands and VK_EXT_transform_feedback.

pub unsafe fn cmd_draw_indirect_byte_count2_ext( &self, command_buffer: CommandBuffer, instance_count: u32, first_instance: u32, p_counter_info: &BindTransformFeedbackBuffer2InfoEXT, counter_offset: u32, vertex_stride: u32, )

Wraps vkCmdDrawIndirectByteCount2EXT. Provided by VK_KHR_device_address_commands.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdDrawIndirectByteCount2EXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Device-address variant of cmd_draw_indirect_byte_count_ext. Draws vertices using a byte count from a transform feedback counter, with the counter buffer specified via device address instead of a buffer handle.

counter_offset is the byte offset within the counter value to account for any header. vertex_stride determines how many bytes each vertex consumes.

Requires VK_KHR_device_address_commands and VK_EXT_transform_feedback.

pub unsafe fn cmd_write_marker_to_memory_amd( &self, command_buffer: CommandBuffer, p_info: &MemoryMarkerInfoAMD, )

Wraps vkCmdWriteMarkerToMemoryAMD. Provided by VK_KHR_device_address_commands.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdWriteMarkerToMemoryAMD was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Device-address variant of the AMD buffer marker extension. Writes a 32-bit marker value to a device address at a specific pipeline stage.

The MemoryMarkerInfoAMD specifies the pipeline stage, device address, and marker value. Useful for GPU crash debugging, markers that were written indicate how far the GPU progressed before a fault.

Requires VK_KHR_device_address_commands. This is the device-address counterpart of cmd_write_buffer_marker_amd.

pub unsafe fn cmd_bind_index_buffer3_khr( &self, command_buffer: CommandBuffer, p_info: &BindIndexBuffer3InfoKHR, )

Wraps vkCmdBindIndexBuffer3KHR. Provided by VK_KHR_device_address_commands.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdBindIndexBuffer3KHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Device-address variant of cmd_bind_index_buffer. Binds an index buffer for subsequent indexed draw commands using a device address instead of a buffer handle.

The BindIndexBuffer3InfoKHR struct specifies the device address, size, and index type (UINT16, UINT32, or UINT8 if enabled).

Supersedes cmd_bind_index_buffer and cmd_bind_index_buffer2 when using VK_KHR_device_address_commands.

pub unsafe fn cmd_bind_vertex_buffers3_khr( &self, command_buffer: CommandBuffer, first_binding: u32, p_binding_infos: &[BindVertexBuffer3InfoKHR], )

Wraps vkCmdBindVertexBuffers3KHR. Provided by VK_KHR_device_address_commands.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdBindVertexBuffers3KHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Device-address variant of cmd_bind_vertex_buffers. Binds vertex buffers for subsequent draw commands using device addresses instead of buffer handles.

Each BindVertexBuffer3InfoKHR specifies a device address, size, and stride for one binding slot starting at first_binding.

Supersedes cmd_bind_vertex_buffers, cmd_bind_vertex_buffers2, and the core 1.4 equivalent when using VK_KHR_device_address_commands.

pub unsafe fn cmd_draw_indirect2_khr( &self, command_buffer: CommandBuffer, p_info: &DrawIndirect2InfoKHR, )

Wraps vkCmdDrawIndirect2KHR. Provided by VK_KHR_device_address_commands.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdDrawIndirect2KHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Device-address variant of cmd_draw_indirect. Reads non-indexed draw parameters from a device address instead of a buffer handle

• offset.

The DrawIndirect2InfoKHR specifies the device address, draw count, and stride.

Requires VK_KHR_device_address_commands.

pub unsafe fn cmd_draw_indexed_indirect2_khr( &self, command_buffer: CommandBuffer, p_info: &DrawIndirect2InfoKHR, )

Wraps vkCmdDrawIndexedIndirect2KHR. Provided by VK_KHR_device_address_commands.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdDrawIndexedIndirect2KHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Device-address variant of cmd_draw_indexed_indirect. Reads indexed draw parameters from a device address instead of a buffer handle + offset.

The DrawIndirect2InfoKHR specifies the device address, draw count, and stride.

Requires VK_KHR_device_address_commands.

pub unsafe fn cmd_draw_indirect_count2_khr( &self, command_buffer: CommandBuffer, p_info: &DrawIndirectCount2InfoKHR, )

Wraps vkCmdDrawIndirectCount2KHR. Provided by VK_KHR_device_address_commands.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdDrawIndirectCount2KHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Device-address variant of cmd_draw_indirect_count. Reads non-indexed draw parameters and the draw count from device addresses instead of buffer handles.

The DrawIndirectCount2InfoKHR specifies both the draw parameter address and the count address, along with max_draw_count and stride.

Requires VK_KHR_device_address_commands.

pub unsafe fn cmd_draw_indexed_indirect_count2_khr( &self, command_buffer: CommandBuffer, p_info: &DrawIndirectCount2InfoKHR, )

Wraps vkCmdDrawIndexedIndirectCount2KHR. Provided by VK_KHR_device_address_commands.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdDrawIndexedIndirectCount2KHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Device-address variant of cmd_draw_indexed_indirect_count. Reads indexed draw parameters and the draw count from device addresses instead of buffer handles.

The DrawIndirectCount2InfoKHR specifies both the draw parameter address and the count address, along with max_draw_count and stride.

Requires VK_KHR_device_address_commands.

pub unsafe fn cmd_draw_mesh_tasks_indirect2_ext( &self, command_buffer: CommandBuffer, p_info: &DrawIndirect2InfoKHR, )

Wraps vkCmdDrawMeshTasksIndirect2EXT. Provided by VK_KHR_device_address_commands.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdDrawMeshTasksIndirect2EXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Device-address variant of cmd_draw_mesh_tasks_indirect_ext. Reads mesh shader dispatch parameters from a device address instead of a buffer handle.

The DrawIndirect2InfoKHR specifies the device address, draw count, and stride.

Requires VK_KHR_device_address_commands and VK_EXT_mesh_shader.

pub unsafe fn cmd_draw_mesh_tasks_indirect_count2_ext( &self, command_buffer: CommandBuffer, p_info: &DrawIndirectCount2InfoKHR, )

Wraps vkCmdDrawMeshTasksIndirectCount2EXT. Provided by VK_KHR_device_address_commands.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdDrawMeshTasksIndirectCount2EXT was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Device-address variant of cmd_draw_mesh_tasks_indirect_count_ext. Reads mesh shader dispatch parameters and the draw count from device addresses instead of buffer handles.

The DrawIndirectCount2InfoKHR specifies both the draw parameter address and the count address, along with max_draw_count and stride.

Requires VK_KHR_device_address_commands and VK_EXT_mesh_shader.

pub unsafe fn cmd_dispatch_indirect2_khr( &self, command_buffer: CommandBuffer, p_info: &DispatchIndirect2InfoKHR, )

Wraps vkCmdDispatchIndirect2KHR. Provided by VK_KHR_device_address_commands.

Safety

• commandBuffer (self) must be valid and not destroyed.
• commandBuffer must be externally synchronized.

Panics

Panics if vkCmdDispatchIndirect2KHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Device-address variant of cmd_dispatch_indirect. Reads the dispatch parameters (group counts) from a device address instead of a buffer handle + offset.

The DispatchIndirect2InfoKHR specifies the device address where the DispatchIndirectCommand struct resides.

Requires VK_KHR_device_address_commands.

pub unsafe fn create_acceleration_structure2_khr( &self, p_create_info: &AccelerationStructureCreateInfo2KHR, allocator: Option<&AllocationCallbacks>, ) -> VkResult<AccelerationStructureKHR>

Wraps vkCreateAccelerationStructure2KHR. Provided by VK_KHR_device_address_commands.

Errors

• VK_ERROR_OUT_OF_HOST_MEMORY
• VK_ERROR_INVALID_OPAQUE_CAPTURE_ADDRESS_KHR
• VK_ERROR_VALIDATION_FAILED
• VK_ERROR_UNKNOWN

Safety

• device (self) must be valid and not destroyed.

Panics

Panics if vkCreateAccelerationStructure2KHR was not loaded. This can happen if the required extension or Vulkan version is not enabled on the instance or device.

Usage Notes

Device-address variant of create_acceleration_structure_khr. Creates a ray tracing acceleration structure backed by a device address range instead of a buffer handle.

The AccelerationStructureCreateInfo2KHR specifies the device address, size, type (top-level or bottom-level), and optional capture/replay address.

Destroy with destroy_acceleration_structure_khr.

Requires VK_KHR_device_address_commands and VK_KHR_acceleration_structure.