Enum HardwareCompatibilityLevel 

#[repr(i32)]
pub enum HardwareCompatibilityLevel {
    kNONE = 0,
    kAMPERE_PLUS = 1,
    kSAME_COMPUTE_CAPABILITY = 2,
}

HardwareCompatibilityLevel

Describes requirements of compatibility with GPU architectures other than that of the GPU on which the engine was built.

Note that compatibility with future hardware depends on CUDA forward compatibility support.

Variants

kNONE = 0

Do not require hardware compatibility with GPU architectures other than that of the GPU on which the engine was built.

kAMPERE_PLUS = 1

Require that the engine is compatible with Ampere and newer GPUs. This will limit the combined usage of driver reserved and backend kernel max shared memory to 48KiB, may reduce the number of available tactics for each layer, and may prevent some fusions from occurring. Thus this can decrease the performance, especially for tf32 models. This option will disable cuDNN, cuBLAS, and cuBLASLt as tactic sources.

This option is only supported for engines built on NVIDIA Ampere and later GPUs.

The driver reserved shared memory can be queried from cuDeviceGetAttribute(&reservedShmem, CU_DEVICE_ATTRIBUTE_RESERVED_SHARED_MEMORY_PER_BLOCK).

kSAME_COMPUTE_CAPABILITY = 2

Require that the engine is compatible with GPUs that have the same Compute Capability (https://developer.nvidia.com/cuda-gpus) as the one it was built on. This may decrease the performance compared to an engine with no compatibility.

This option will disable cuDNN, cuBLAS, and cuBLASLt as tactic sources.

This option is only supported for engines built on NVIDIA Turing and later GPUs.

Trait Implementations

impl Clone for HardwareCompatibilityLevel

fn clone(&self) -> HardwareCompatibilityLevel

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl ExternType for HardwareCompatibilityLevel

type Id = (n, v, i, n, f, e, r, _1, (), H, a, r, d, w, a, r, e, C, o, m, p, a, t, i, b, i, l, i, t, y, L, e, v, e, l)

A type-level representation of the type’s C++ namespace and type name.

type Kind = Trivial

Either cxx::kind::Opaque or cxx::kind::Trivial.

impl From<HardwareCompatibilityLevel> for HardwareCompatibilityLevel

fn from(value: HardwareCompatibilityLevel) -> Self

Converts to this type from the input type.

impl Hash for HardwareCompatibilityLevel

fn hash<__H: Hasher>(&self, state: &mut __H)

Feeds this value into the given Hasher.

fn hash_slice<H>(data: &[Self], state: &mut H)

where H: Hasher, Self: Sized,

Feeds a slice of this type into the given Hasher.

impl Into<HardwareCompatibilityLevel> for HardwareCompatibilityLevel

fn into(self) -> HardwareCompatibilityLevel

Converts this type into the (usually inferred) input type.

impl PartialEq for HardwareCompatibilityLevel

fn eq(&self, other: &HardwareCompatibilityLevel) -> bool

Tests for self and other values to be equal, and is used by ==.

fn ne(&self, other: &Rhs) -> bool

Tests for !=. The default implementation is almost always sufficient, and should not be overridden without very good reason.

impl Eq for HardwareCompatibilityLevel

impl SharedPtrTarget for HardwareCompatibilityLevel

impl StructuralPartialEq for HardwareCompatibilityLevel

impl UniquePtrTarget for HardwareCompatibilityLevel

impl VectorElement for HardwareCompatibilityLevel

impl WeakPtrTarget for HardwareCompatibilityLevel