MTLAccelerationStructureCommandEncoder

Trait MTLAccelerationStructureCommandEncoder 

Source
pub unsafe trait MTLAccelerationStructureCommandEncoder: MTLCommandEncoder {
Show 14 methods // Provided methods fn buildAccelerationStructure_descriptor_scratchBuffer_scratchBufferOffset( &self, acceleration_structure: &ProtocolObject<dyn MTLAccelerationStructure>, descriptor: &MTLAccelerationStructureDescriptor, scratch_buffer: &ProtocolObject<dyn MTLBuffer>, scratch_buffer_offset: NSUInteger, ) where Self: Sized + Message { ... } unsafe fn refitAccelerationStructure_descriptor_destination_scratchBuffer_scratchBufferOffset( &self, source_acceleration_structure: &ProtocolObject<dyn MTLAccelerationStructure>, descriptor: &MTLAccelerationStructureDescriptor, destination_acceleration_structure: Option<&ProtocolObject<dyn MTLAccelerationStructure>>, scratch_buffer: Option<&ProtocolObject<dyn MTLBuffer>>, scratch_buffer_offset: NSUInteger, ) where Self: Sized + Message { ... } unsafe fn refitAccelerationStructure_descriptor_destination_scratchBuffer_scratchBufferOffset_options( &self, source_acceleration_structure: &ProtocolObject<dyn MTLAccelerationStructure>, descriptor: &MTLAccelerationStructureDescriptor, destination_acceleration_structure: Option<&ProtocolObject<dyn MTLAccelerationStructure>>, scratch_buffer: Option<&ProtocolObject<dyn MTLBuffer>>, scratch_buffer_offset: NSUInteger, options: MTLAccelerationStructureRefitOptions, ) where Self: Sized + Message { ... } unsafe fn copyAccelerationStructure_toAccelerationStructure( &self, source_acceleration_structure: &ProtocolObject<dyn MTLAccelerationStructure>, destination_acceleration_structure: &ProtocolObject<dyn MTLAccelerationStructure>, ) where Self: Sized + Message { ... } fn writeCompactedAccelerationStructureSize_toBuffer_offset( &self, acceleration_structure: &ProtocolObject<dyn MTLAccelerationStructure>, buffer: &ProtocolObject<dyn MTLBuffer>, offset: NSUInteger, ) where Self: Sized + Message { ... } unsafe fn writeCompactedAccelerationStructureSize_toBuffer_offset_sizeDataType( &self, acceleration_structure: &ProtocolObject<dyn MTLAccelerationStructure>, buffer: &ProtocolObject<dyn MTLBuffer>, offset: NSUInteger, size_data_type: MTLDataType, ) where Self: Sized + Message { ... } fn copyAndCompactAccelerationStructure_toAccelerationStructure( &self, source_acceleration_structure: &ProtocolObject<dyn MTLAccelerationStructure>, destination_acceleration_structure: &ProtocolObject<dyn MTLAccelerationStructure>, ) where Self: Sized + Message { ... } fn updateFence(&self, fence: &ProtocolObject<dyn MTLFence>) where Self: Sized + Message { ... } fn waitForFence(&self, fence: &ProtocolObject<dyn MTLFence>) where Self: Sized + Message { ... } fn useResource_usage( &self, resource: &ProtocolObject<dyn MTLResource>, usage: MTLResourceUsage, ) where Self: Sized + Message { ... } unsafe fn useResources_count_usage( &self, resources: NonNull<NonNull<ProtocolObject<dyn MTLResource>>>, count: NSUInteger, usage: MTLResourceUsage, ) where Self: Sized + Message { ... } fn useHeap(&self, heap: &ProtocolObject<dyn MTLHeap>) where Self: Sized + Message { ... } unsafe fn useHeaps_count( &self, heaps: NonNull<NonNull<ProtocolObject<dyn MTLHeap>>>, count: NSUInteger, ) where Self: Sized + Message { ... } unsafe fn sampleCountersInBuffer_atSampleIndex_withBarrier( &self, sample_buffer: &ProtocolObject<dyn MTLCounterSampleBuffer>, sample_index: NSUInteger, barrier: bool, ) where Self: Sized + Message { ... }
}
Available on crate features MTLAccelerationStructureCommandEncoder and MTLCommandEncoder only.
Expand description

Provided Methods§

Source

fn buildAccelerationStructure_descriptor_scratchBuffer_scratchBufferOffset( &self, acceleration_structure: &ProtocolObject<dyn MTLAccelerationStructure>, descriptor: &MTLAccelerationStructureDescriptor, scratch_buffer: &ProtocolObject<dyn MTLBuffer>, scratch_buffer_offset: NSUInteger, )
where Self: Sized + Message,

Available on crate features MTLAccelerationStructure and MTLAllocation and MTLBuffer and MTLResource only.

Encode an acceleration structure build into the command buffer. All bottom-level acceleration structure builds must have completed before a top-level acceleration structure build may begin. The resulting acceleration structure will not retain any references to the input vertex buffer, instance buffer, etc.

The acceleration structure build will not be completed until the command buffer has been committed and finished executing. However, it is safe to encode ray tracing work against the acceleration structure as long as the command buffers are scheduled and synchronized such that the command buffer will have completed by the time the ray tracing starts.

The acceleration structure and scratch buffer must be at least the size returned by the [MTLDevice accelerationStructureSizesWithDescriptor:] query.

Parameter accelerationStructure: Acceleration structure storage to build into

Parameter descriptor: Object describing the acceleration structure to build

Parameter scratchBuffer: Scratch buffer to use while building the acceleration structure. The contents may be overwritten and are undefined after the build has started/completed.

Parameter scratchBufferOffset: Offset into the scratch buffer

Source

unsafe fn refitAccelerationStructure_descriptor_destination_scratchBuffer_scratchBufferOffset( &self, source_acceleration_structure: &ProtocolObject<dyn MTLAccelerationStructure>, descriptor: &MTLAccelerationStructureDescriptor, destination_acceleration_structure: Option<&ProtocolObject<dyn MTLAccelerationStructure>>, scratch_buffer: Option<&ProtocolObject<dyn MTLBuffer>>, scratch_buffer_offset: NSUInteger, )
where Self: Sized + Message,

Available on crate features MTLAccelerationStructure and MTLAllocation and MTLBuffer and MTLResource only.

Encode an acceleration structure refit into the command buffer. Refitting can be used to update the acceleration structure when geometry changes and is much faster than rebuilding from scratch. However, the quality of the acceleration structure and the subsequent ray tracing performance will degrade depending on how much the geometry changes.

Refitting can not be used after certain changes, such as adding or removing geometry. Acceleration structures can be refit in place by specifying the same source and destination acceleration structures or by providing a nil destination acceleration structure. If the source and destination acceleration structures are not the same, they must not overlap in memory.

The destination acceleration structure must be at least as large as the source acceleration structure, unless the source acceleration structure has been compacted, in which case the destination acceleration structure must be at least as large as the compacted size of the source acceleration structure.

The scratch buffer must be at least the size returned by the accelerationStructureSizesWithDescriptor method of the MTLDevice.

Parameter descriptor: Object describing the acceleration structure to build

Parameter sourceAccelerationStructure: Acceleration structure to copy from

Parameter destinationAccelerationStructure: Acceleration structure to copy to

Parameter scratchBuffer: Scratch buffer to use while refitting the acceleration structure. The contents may be overwritten and are undefined after the refit has started/completed.

Parameter scratchBufferOffset: Offset into the scratch buffer.

§Safety
  • source_acceleration_structure may need to be synchronized.
  • source_acceleration_structure may be unretained, you must ensure it is kept alive while in use.
  • destination_acceleration_structure may need to be synchronized.
  • destination_acceleration_structure may be unretained, you must ensure it is kept alive while in use.
  • scratch_buffer may need to be synchronized.
  • scratch_buffer may be unretained, you must ensure it is kept alive while in use.
  • scratch_buffer contents should be of the correct type.
  • scratchBufferOffset might not be bounds-checked.
Source

unsafe fn refitAccelerationStructure_descriptor_destination_scratchBuffer_scratchBufferOffset_options( &self, source_acceleration_structure: &ProtocolObject<dyn MTLAccelerationStructure>, descriptor: &MTLAccelerationStructureDescriptor, destination_acceleration_structure: Option<&ProtocolObject<dyn MTLAccelerationStructure>>, scratch_buffer: Option<&ProtocolObject<dyn MTLBuffer>>, scratch_buffer_offset: NSUInteger, options: MTLAccelerationStructureRefitOptions, )
where Self: Sized + Message,

Available on crate features MTLAccelerationStructure and MTLAllocation and MTLBuffer and MTLResource only.

Encode an acceleration structure refit into the command buffer. Refitting can be used to update the acceleration structure when geometry changes and is much faster than rebuilding from scratch. However, the quality of the acceleration structure and the subsequent ray tracing performance will degrade depending on how much the geometry changes.

Refitting can not be used after certain changes, such as adding or removing geometry. Acceleration structures can be refit in place by specifying the same source and destination acceleration structures or by providing a nil destination acceleration structure. If the source and destination acceleration structures are not the same, they must not overlap in memory.

The destination acceleration structure must be at least as large as the source acceleration structure, unless the source acceleration structure has been compacted, in which case the destination acceleration structure must be at least as large as the compacted size of the source acceleration structure.

The scratch buffer must be at least the size returned by the accelerationStructureSizesWithDescriptor method of the MTLDevice.

Parameter descriptor: Object describing the acceleration structure to build

Parameter sourceAccelerationStructure: Acceleration structure to copy from

Parameter destinationAccelerationStructure: Acceleration structure to copy to

Parameter scratchBuffer: Scratch buffer to use while refitting the acceleration structure. The contents may be overwritten and are undefined after the refit has started/completed.

Parameter scratchBufferOffset: Offset into the scratch buffer.

Parameter options: Options specifying the elements of the acceleration structure to refit.

§Safety
  • source_acceleration_structure may need to be synchronized.
  • source_acceleration_structure may be unretained, you must ensure it is kept alive while in use.
  • destination_acceleration_structure may need to be synchronized.
  • destination_acceleration_structure may be unretained, you must ensure it is kept alive while in use.
  • scratch_buffer may need to be synchronized.
  • scratch_buffer may be unretained, you must ensure it is kept alive while in use.
  • scratch_buffer contents should be of the correct type.
  • scratchBufferOffset might not be bounds-checked.
Source

unsafe fn copyAccelerationStructure_toAccelerationStructure( &self, source_acceleration_structure: &ProtocolObject<dyn MTLAccelerationStructure>, destination_acceleration_structure: &ProtocolObject<dyn MTLAccelerationStructure>, )
where Self: Sized + Message,

Available on crate features MTLAccelerationStructure and MTLAllocation and MTLResource only.

Copy an acceleration structure. The source and destination acceleration structures must not overlap in memory. If this is a top level acceleration structure, references to bottom level acceleration structures will be preserved.

The destination acceleration structure must be at least as large as the source acceleration structure, unless the source acceleration structure has been compacted, in which case the destination acceleration structure must be at least as large as the compacted size of the source acceleration structure.

Parameter sourceAccelerationStructure: Acceleration structure to copy from

Parameter destinationAccelerationStructure: Acceleration structure to copy to

§Safety
  • source_acceleration_structure may need to be synchronized.
  • source_acceleration_structure may be unretained, you must ensure it is kept alive while in use.
  • destination_acceleration_structure may need to be synchronized.
  • destination_acceleration_structure may be unretained, you must ensure it is kept alive while in use.
Source

fn writeCompactedAccelerationStructureSize_toBuffer_offset( &self, acceleration_structure: &ProtocolObject<dyn MTLAccelerationStructure>, buffer: &ProtocolObject<dyn MTLBuffer>, offset: NSUInteger, )
where Self: Sized + Message,

Available on crate features MTLAccelerationStructure and MTLAllocation and MTLBuffer and MTLResource only.

Compute the compacted size for an acceleration structure and write it into a buffer.

This size is potentially smaller than the source acceleration structure. To perform compaction, read this size from the buffer once the command buffer has completed and use it to allocate a smaller acceleration structure. Then create another encoder and call the copyAndCompactAccelerationStructure method.

Parameter accelerationStructure: Source acceleration structure

Parameter buffer: Destination size buffer. The compacted size will be written as a 32 bit unsigned integer representing the compacted size in bytes.

Parameter offset: Offset into the size buffer

Source

unsafe fn writeCompactedAccelerationStructureSize_toBuffer_offset_sizeDataType( &self, acceleration_structure: &ProtocolObject<dyn MTLAccelerationStructure>, buffer: &ProtocolObject<dyn MTLBuffer>, offset: NSUInteger, size_data_type: MTLDataType, )
where Self: Sized + Message,

Available on crate features MTLAccelerationStructure and MTLAllocation and MTLBuffer and MTLDataType and MTLResource only.

Compute the compacted size for an acceleration structure and write it into a buffer.

This size is potentially smaller than the source acceleration structure. To perform compaction, read this size from the buffer once the command buffer has completed and use it to allocate a smaller acceleration structure. Then create another encoder and call the copyAndCompactAccelerationStructure method.

Parameter accelerationStructure: Source acceleration structure

Parameter buffer: Destination size buffer. The compacted size will be written as either a 32 bit or 64 bit value depending on the sizeDataType argument unsigned integer representing the compacted size in bytes.

Parameter offset: Offset into the size buffer

Parameter sizeDataType: Data type of the size to write into the buffer. Must be either MTLDataTypeUInt (32 bit) or MTLDataTypeULong (64 bit)

§Safety
  • acceleration_structure may need to be synchronized.
  • acceleration_structure may be unretained, you must ensure it is kept alive while in use.
  • buffer may need to be synchronized.
  • buffer may be unretained, you must ensure it is kept alive while in use.
  • buffer contents should be of the correct type.
  • offset might not be bounds-checked.
Source

fn copyAndCompactAccelerationStructure_toAccelerationStructure( &self, source_acceleration_structure: &ProtocolObject<dyn MTLAccelerationStructure>, destination_acceleration_structure: &ProtocolObject<dyn MTLAccelerationStructure>, )
where Self: Sized + Message,

Available on crate features MTLAccelerationStructure and MTLAllocation and MTLResource only.

Copy and compact an acceleration structure. The source and destination acceleration structures must not overlap in memory. If this is a top level acceleration structure, references to bottom level acceleration structures will be preserved.

The destination acceleration structure must be at least as large as the compacted size of the source acceleration structure, which is computed by the writeCompactedAccelerationStructureSize method.

Parameter sourceAccelerationStructure: Acceleration structure to copy and compact

Parameter destinationAccelerationStructure: Acceleration structure to copy to

Source

fn updateFence(&self, fence: &ProtocolObject<dyn MTLFence>)
where Self: Sized + Message,

Available on crate feature MTLFence only.

Update the fence to capture all GPU work so far enqueued by this encoder.

The fence is updated at build submission to maintain global order and prevent deadlock. Drivers may delay fence updates until the end of the encoder. Drivers may also wait on fences at the beginning of an encoder. It is therefore illegal to wait on a fence after it has been updated in the same encoder.

Source

fn waitForFence(&self, fence: &ProtocolObject<dyn MTLFence>)
where Self: Sized + Message,

Available on crate feature MTLFence only.

Prevent further GPU work until the fence is reached.

The fence is evaluated at build submission to maintain global order and prevent deadlock. Drivers may delay fence updates until the end of the encoder. Drivers may also wait on fences at the beginning of an encoder. It is therefore illegal to wait on a fence after it has been updated in the same encoder.

Source

fn useResource_usage( &self, resource: &ProtocolObject<dyn MTLResource>, usage: MTLResourceUsage, )
where Self: Sized + Message,

Available on crate features MTLAllocation and MTLResource only.

Declare that a resource may be accessed by the command encoder through an argument buffer

For tracked MTLResources, this method protects against data hazards. This method must be called before encoding any acceleration structure commands which may access the resource through an argument buffer.

Warning: Prior to iOS 13, macOS 10.15, this method does not protect against data hazards. If you are deploying to older versions of macOS or iOS, use fences to ensure data hazards are resolved.

Source

unsafe fn useResources_count_usage( &self, resources: NonNull<NonNull<ProtocolObject<dyn MTLResource>>>, count: NSUInteger, usage: MTLResourceUsage, )
where Self: Sized + Message,

Available on crate features MTLAllocation and MTLResource only.

Declare that an array of resources may be accessed through an argument buffer by the command encoder

For tracked MTL Resources, this method protects against data hazards. This method must be called before encoding any acceleration structure commands which may access the resources through an argument buffer.

Warning: Prior to iOS 13, macOS 10.15, this method does not protect against data hazards. If you are deploying to older versions of macOS or iOS, use fences to ensure data hazards are resolved.

§Safety
  • resources must be a valid pointer.
  • count might not be bounds-checked.
Source

fn useHeap(&self, heap: &ProtocolObject<dyn MTLHeap>)
where Self: Sized + Message,

Available on crate features MTLAllocation and MTLHeap only.

Declare that the resources allocated from a heap may be accessed as readonly by the encoder through an argument buffer

For tracked MTLHeaps, this method protects against data hazards. This method must be called before encoding any acceleration structure commands which may access the resources allocated from the heap through an argument buffer. This method may cause all of the color attachments allocated from the heap to become decompressed. Therefore, it is recommended that the useResource:usage: or useResources:count:usage: methods be used for color attachments instead, with a minimal (i.e. read-only) usage.

Warning: Prior to iOS 13, macOS 10.15, this method does not protect against data hazards. If you are deploying to older versions of macOS or iOS, use fences to ensure data hazards are resolved.

Source

unsafe fn useHeaps_count( &self, heaps: NonNull<NonNull<ProtocolObject<dyn MTLHeap>>>, count: NSUInteger, )
where Self: Sized + Message,

Available on crate features MTLAllocation and MTLHeap only.

Declare that the resources allocated from an array of heaps may be accessed as readonly by the encoder through an argument buffer

For tracked MTLHeaps, this method protects against data hazards. This method must be called before encoding any acceleration structure commands which may access the resources allocated from the heaps through an argument buffer. This method may cause all of the color attachments allocated from the heaps to become decompressed. Therefore, it is recommended that the useResource:usage: or useResources:count:usage: methods be used for color attachments instead, with a minimal (i.e. read-only) usage.

Warning: Prior to iOS 13, macOS 10.15, this method does not protect against data hazards. If you are deploying to older versions of macOS or iOS, use fences to ensure data hazards are resolved.

§Safety
  • heaps must be a valid pointer.
  • count might not be bounds-checked.
Source

unsafe fn sampleCountersInBuffer_atSampleIndex_withBarrier( &self, sample_buffer: &ProtocolObject<dyn MTLCounterSampleBuffer>, sample_index: NSUInteger, barrier: bool, )
where Self: Sized + Message,

Available on crate feature MTLCounters only.

Sample hardware counters at this point in the acceleration structure encoder and store the counter sample into the sample buffer at the specified index.

Parameter sampleBuffer: The sample buffer to sample into

Parameter sampleIndex: The index into the counter buffer to write the sample

Parameter barrier: Insert a barrier before taking the sample. Passing YES will ensure that all work encoded before this operation in the encoder is complete but does not isolate the work with respect to other encoders. Passing NO will allow the sample to be taken concurrently with other operations in this encoder. In general, passing YES will lead to more repeatable counter results but may negatively impact performance. Passing NO will generally be higher performance but counter results may not be repeatable.

§Safety

sampleIndex might not be bounds-checked.

Trait Implementations§

Source§

impl ProtocolType for dyn MTLAccelerationStructureCommandEncoder

Source§

const NAME: &'static str = "MTLAccelerationStructureCommandEncoder"

The name of the Objective-C protocol that this type represents. Read more
Source§

fn protocol() -> Option<&'static AnyProtocol>

Get a reference to the Objective-C protocol object that this type represents. Read more
Source§

impl<T> ImplementedBy<T> for dyn MTLAccelerationStructureCommandEncoder

Implementations on Foreign Types§

Source§

impl<T> MTLAccelerationStructureCommandEncoder for ProtocolObject<T>

Implementors§