ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-04-20 06:49:15 +00:00
Code Issues Packages Projects Releases Wiki Activity

[CK_BUILDER] Add grouped conv fwd ck tile traits (#3183)

Browse Source 
* [CK BUILDER] Add grouped conv fwd ck tile traits

* Update instance_traits_tile_grouped_convolution_forward.hpp

* Update grouped_convolution_forward_kernel.hpp
This commit is contained in:
Bartłomiej Kocot
2025-11-11 22:55:33 +01:00
committed by GitHub
parent b145a5fe80
commit 92c1f4981a
18 changed files with 433 additions and 15 deletions
Download Patch File Download Diff File Expand all files Collapse all files

10
include/ck_tile/core/arch/arch.hpp Executable file → Normal file
View File

@@ -299,12 +299,12 @@ CK_TILE_DEVICE void s_nop(index_t cnt = 0)
#endif
}
#define CK_CONSTANT_ADDRESS_SPACE \
__attribute__((address_space( \
#define CK_TILE_CONSTANT_ADDRESS_SPACE \
__attribute__((address_space( \
static_cast<safe_underlying_type_t<address_space_enum>>(address_space_enum::constant))))
template <typename T>
__device__ T* cast_pointer_to_generic_address_space(T CK_CONSTANT_ADDRESS_SPACE* p)
__device__ T* cast_pointer_to_generic_address_space(T CK_TILE_CONSTANT_ADDRESS_SPACE* p)
{
// cast a pointer in "Constant" address space (4) to "Generic" address space (0)
// only c-style pointer cast seems be able to be compiled
@@ -315,13 +315,13 @@ __device__ T* cast_pointer_to_generic_address_space(T CK_CONSTANT_ADDRESS_SPACE*
}
template <typename T>
__host__ __device__ T CK_CONSTANT_ADDRESS_SPACE* cast_pointer_to_constant_address_space(T* p)
__host__ __device__ T CK_TILE_CONSTANT_ADDRESS_SPACE* cast_pointer_to_constant_address_space(T* p)
{
// cast a pointer in "Generic" address space (0) to "Constant" address space (4)
// only c-style pointer cast seems be able to be compiled;
#pragma clang diagnostic push
#pragma clang diagnostic ignored "-Wold-style-cast"
return (T CK_CONSTANT_ADDRESS_SPACE*)p; // NOLINT(old-style-cast)
return (T CK_TILE_CONSTANT_ADDRESS_SPACE*)p; // NOLINT(old-style-cast)
#pragma clang diagnostic pop
}