ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-05-24 14:54:47 +00:00
Code Issues Packages Projects Releases Wiki Activity

Merge commit 'b7a806f2442ed04db9e835e3e4e14aaebe3db9b4' into develop

Browse Source
This commit is contained in:
assistant-librarian[bot]
2025-09-16 16:13:22 +00:00
parent 1191876ea6
commit 0daa4023fc
16 changed files with 203 additions and 96 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
test/ck_tile/add_rmsnorm2d_rdquant/instances/add_rmsnorm2d_rdquant_fwd_instance_common.hpp
View File

@@ -58,7 +58,7 @@ float add_rmsnorm2d_rdquant_fwd_(const S& s, A a)
using Kernel = ck_tile::AddRmsnorm2dRdquantFwd<Pipeline>;
const dim3 grids = Kernel::GridSize(a);
constexpr dim3 blocks = Kernel::BlockSize();
const dim3 blocks = Kernel::BlockSize();
constexpr ck_tile::index_t kBlockPerCu = 1;
auto kargs = Kernel::MakeKargs(a);

2
test/ck_tile/moe_smoothquant/instances/moe_smoothquant_instance_common.hpp
View File

@@ -53,7 +53,7 @@ float moe_smoothquant_(const S& s, A a)
using Kernel = ck_tile::MoeSmoothquant<Pipeline>;
const dim3 grids = Kernel::GridSize(a);
constexpr dim3 blocks = Kernel::BlockSize();
const dim3 blocks = Kernel::BlockSize();
constexpr ck_tile::index_t kBlockPerCu = 1;
auto kargs = Kernel::MakeKargs(a);

2
test/ck_tile/rmsnorm2d/generate.py
View File

@@ -201,7 +201,7 @@ float rmsnorm2d_fwd_(const S& s, A a)
using Kernel = ck_tile::Rmsnorm2dFwd<Pipeline, Epilogue>;
const dim3 grids = Kernel::GridSize(a);
constexpr dim3 blocks = Kernel::BlockSize();
const dim3 blocks = Kernel::BlockSize();
constexpr ck_tile::index_t kBlockPerCu = 1;
auto kargs = Kernel::MakeKargs(a);

2
test/ck_tile/smoothquant/instances/smoothquant_instance_common.hpp
View File

@@ -49,7 +49,7 @@ float smoothquant_(const S& s, A a)
using Kernel = ck_tile::Smoothquant<Pipeline>;
const dim3 grids = Kernel::GridSize(a);
constexpr dim3 blocks = Kernel::BlockSize();
const dim3 blocks = Kernel::BlockSize();
constexpr ck_tile::index_t kBlockPerCu = 1;
auto kargs = Kernel::MakeKargs(a);