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

4
include/ck_tile/ops/gemm_quant/kernel/grouped_gemm_quant_kernel.hpp
View File

@@ -208,7 +208,7 @@ struct QuantGroupedGemmKernel
*/
CK_TILE_HOST static auto MaxOccupancyGridSize(const stream_config& s) -> dim3
{
using ConstantPointer = const void CK_CONSTANT_ADDRESS_SPACE*;
using ConstantPointer = const void CK_TILE_CONSTANT_ADDRESS_SPACE*;
const auto kernel_func = kentry<1, Kernel, ConstantPointer, index_t>;
int occupancy;
HIP_CHECK_ERROR(
@@ -499,7 +499,7 @@ struct QuantGroupedGemmKernel
template <bool U = UsePersistentKernel,
typename = std::enable_if_t<U>,
typename = void> // extra template parameter to avoid redefinition
CK_TILE_DEVICE void operator()(const void CK_CONSTANT_ADDRESS_SPACE* gemm_descs_const,
CK_TILE_DEVICE void operator()(const void CK_TILE_CONSTANT_ADDRESS_SPACE* gemm_descs_const,
const index_t group_count) const
{
const index_t grid_size = ck_tile::get_grid_size();