Parameters object constructable on the host.
#include <gemm_traits.h>
Inheritance diagram for cutlass::gemm::GemmTraits< GemmConfig_, GlobalLoadStreamA_, GlobalLoadStreamB_, SharedLoadStreamA_, SharedLoadStreamB_, Epilogue_, BlockSwizzle_, Index_, ClearAccumulators_ >::Params:
Public Member Functions | |
| template<typename GemmDesc_ > | |
| CUTLASS_HOST_DEVICE int | initialize (GemmDesc_ const &desc) |
| Initialize the parameters. More... | |
| CUTLASS_HOST_DEVICE int | initialize (Index m, Index n, Index k, typename Epilogue::Scalar alpha, ScalarA const *d_a, Index lda, ScalarB const *d_b, Index ldb, typename Epilogue::Scalar beta, ScalarC const *d_c, Index ldc, ScalarD *d_d, Index ldd) |
| Helper to construct a GEMM params using a BLAS-like API. More... | |
| CUTLASS_HOST_DEVICE int | initialize (Index m, Index n, Index k, typename Epilogue::Scalar alpha, ScalarA const *d_a, Index lda, long long int batch_stride_A, ScalarB const *d_b, Index ldb, long long int batch_stride_B, typename Epilogue::Scalar beta, ScalarC const *d_c, Index ldc, long long int batch_stride_C, ScalarD *d_d, Index ldd, long long int batch_stride_D, Index batch_count) |
| Helper to construct a batched GEMM params. More... | |
 Public Member Functions inherited from cutlass::KernelLaunchConfiguration | |
| CUTLASS_HOST_DEVICE | KernelLaunchConfiguration (dim3 _grid=dim3(1, 1, 1), dim3 _block=dim3(1, 1, 1), size_t _dynamic_smem=0) |
| Constructs a KernellaunchConfiguration object. More... | |
Public Attributes | |
| GemmCoord | problem_size |
| GEMM problem size. More... | |
| GlobalLoadStream::Params | global_to_shared_stream |
| Parameters object for the global load stream. More... | |
| SharedStream::Params | shared_stream |
| Parameters object for the shared load stream. More... | |
| Epilogue::Params | epilogue |
| The params for the epilogue. More... | |
| Public Attributes inherited from cutlass::KernelLaunchConfiguration | |
| dim3 | grid |
| CUDA grid dimensions. More... | |
| dim3 | block |
| CUDA threablock dimensions. More... | |
| size_t | dynamic_smem |
| Bytes of dynamically allocated SMEM in addition to static SMEM. More... | |
Member Function Documentation
◆ initialize() [1/3]
template<typename GemmConfig_, typename GlobalLoadStreamA_, typename GlobalLoadStreamB_, typename SharedLoadStreamA_, typename SharedLoadStreamB_, typename Epilogue_, typename BlockSwizzle_ = IdentityBlockSwizzle, typename Index_ = int, typename ClearAccumulators_ = ClearAccumulators<typename GemmConfig_::Accumulators::Element>>
template<typename GemmDesc_ >
|
inline |
◆ initialize() [2/3]
template<typename GemmConfig_, typename GlobalLoadStreamA_, typename GlobalLoadStreamB_, typename SharedLoadStreamA_, typename SharedLoadStreamB_, typename Epilogue_, typename BlockSwizzle_ = IdentityBlockSwizzle, typename Index_ = int, typename ClearAccumulators_ = ClearAccumulators<typename GemmConfig_::Accumulators::Element>>
|
inline |
◆ initialize() [3/3]
template<typename GemmConfig_, typename GlobalLoadStreamA_, typename GlobalLoadStreamB_, typename SharedLoadStreamA_, typename SharedLoadStreamB_, typename Epilogue_, typename BlockSwizzle_ = IdentityBlockSwizzle, typename Index_ = int, typename ClearAccumulators_ = ClearAccumulators<typename GemmConfig_::Accumulators::Element>>
|
inline |
Member Data Documentation
◆ epilogue
template<typename GemmConfig_, typename GlobalLoadStreamA_, typename GlobalLoadStreamB_, typename SharedLoadStreamA_, typename SharedLoadStreamB_, typename Epilogue_, typename BlockSwizzle_ = IdentityBlockSwizzle, typename Index_ = int, typename ClearAccumulators_ = ClearAccumulators<typename GemmConfig_::Accumulators::Element>>
| Epilogue::Params cutlass::gemm::GemmTraits< GemmConfig_, GlobalLoadStreamA_, GlobalLoadStreamB_, SharedLoadStreamA_, SharedLoadStreamB_, Epilogue_, BlockSwizzle_, Index_, ClearAccumulators_ >::Params::epilogue |
◆ global_to_shared_stream
template<typename GemmConfig_, typename GlobalLoadStreamA_, typename GlobalLoadStreamB_, typename SharedLoadStreamA_, typename SharedLoadStreamB_, typename Epilogue_, typename BlockSwizzle_ = IdentityBlockSwizzle, typename Index_ = int, typename ClearAccumulators_ = ClearAccumulators<typename GemmConfig_::Accumulators::Element>>
| GlobalLoadStream::Params cutlass::gemm::GemmTraits< GemmConfig_, GlobalLoadStreamA_, GlobalLoadStreamB_, SharedLoadStreamA_, SharedLoadStreamB_, Epilogue_, BlockSwizzle_, Index_, ClearAccumulators_ >::Params::global_to_shared_stream |
◆ problem_size
template<typename GemmConfig_, typename GlobalLoadStreamA_, typename GlobalLoadStreamB_, typename SharedLoadStreamA_, typename SharedLoadStreamB_, typename Epilogue_, typename BlockSwizzle_ = IdentityBlockSwizzle, typename Index_ = int, typename ClearAccumulators_ = ClearAccumulators<typename GemmConfig_::Accumulators::Element>>
| GemmCoord cutlass::gemm::GemmTraits< GemmConfig_, GlobalLoadStreamA_, GlobalLoadStreamB_, SharedLoadStreamA_, SharedLoadStreamB_, Epilogue_, BlockSwizzle_, Index_, ClearAccumulators_ >::Params::problem_size |
◆ shared_stream
template<typename GemmConfig_, typename GlobalLoadStreamA_, typename GlobalLoadStreamB_, typename SharedLoadStreamA_, typename SharedLoadStreamB_, typename Epilogue_, typename BlockSwizzle_ = IdentityBlockSwizzle, typename Index_ = int, typename ClearAccumulators_ = ClearAccumulators<typename GemmConfig_::Accumulators::Element>>
| SharedStream::Params cutlass::gemm::GemmTraits< GemmConfig_, GlobalLoadStreamA_, GlobalLoadStreamB_, SharedLoadStreamA_, SharedLoadStreamB_, Epilogue_, BlockSwizzle_, Index_, ClearAccumulators_ >::Params::shared_stream |
The documentation for this struct was generated from the following file:
