ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-05-04 05:31:24 +00:00
Code Issues Packages Projects Releases Wiki Activity

fixing ambiguous shuffle definitions (#3175)

Browse Source 
Co-authored-by: Thomas Ning <Thomas.Ning@amd.com>
This commit is contained in:
Khushbu Agarwal
2025-11-12 23:44:12 -08:00
committed by GitHub
parent 6fd8ddabe7
commit fb41a7b73b
3 changed files with 2 additions and 148 deletions
Download Patch File Download Diff File Expand all files Collapse all files

43
test/ck_tile/gemm_weight_preshuffle/test_gemm_pipeline_util.hpp
View File

@@ -9,6 +9,7 @@
#include "ck_tile/host.hpp"
#include "ck_tile/host/kernel_launch.hpp"
#include "ck_tile/host/permute_pk_int4.hpp"
#include "ck_tile/host/tensor_shuffle_utils.hpp"
#include "ck_tile/ops/epilogue.hpp"
#include "ck_tile/ops/gemm.hpp"
@@ -259,48 +260,6 @@ class TestCkTileGemmPipeline : public ::testing::Test
k_batches_ = {1};
}
template <typename GemmConfig, typename T>
auto shuffle_b(const ck_tile::HostTensor<T>& t)
{
assert(t.get_lengths().size() == 2);
int n_ = t.get_lengths()[1];
int k_ = t.get_lengths()[0];
if(ck_tile::is_gfx12_supported())
{
constexpr int divisor = 2;
constexpr int kABK1PerLane = 8;
constexpr int kABK0PerLane = GemmConfig::K_Warp_Tile / divisor / kABK1PerLane;
ck_tile::HostTensor<T> t_view({n_ / GemmConfig::N_Warp_Tile,
GemmConfig::N_Warp_Tile,
k_ / GemmConfig::K_Warp_Tile,
kABK0PerLane,
divisor,
kABK1PerLane});
std::copy(t.begin(), t.end(), t_view.begin());
return ck_tile::reference_permute(t_view, {0, 2, 4, 1, 3, 5});
}
else
{
int divisor = 1;
if(ck_tile::is_gfx11_supported())
{
divisor = 1;
}
else
{
assert(is_wave32() == false);
divisor = GemmConfig::N_Warp_Tile == 32 ? 2 : 4;
}
ck_tile::HostTensor<T> t_view({n_ / GemmConfig::N_Warp_Tile,
GemmConfig::N_Warp_Tile,
k_ / GemmConfig::K_Warp_Tile,
divisor,
GemmConfig::K_Warp_Tile / divisor});
std::copy(t.begin(), t.end(), t_view.begin());
return ck_tile::reference_permute(t_view, {0, 2, 3, 1, 4});
}
}
template <bool PadM = true, bool PadN = true, bool PadK = true, bool Preshuffle = false>
void Run(const int M,
const int N,