mirror of
https://github.com/ROCm/composable_kernel.git
synced 2026-04-20 06:49:15 +00:00
Support Wave32 in CK_TILE - Part 1 (#2594)
* Support wave32/wave64 in CK_TILE - Part 1 * remove blocksize in kernel launch * fix build error * fix clang format * fix clang format 2 * fix clang format 3 * fix fmha build error * fix fmha build 2 * fix fmha build 3 * fix build error 4 * address review comment * update change log * replace KernelBlockSize with kBlockSize * fix CI fail * fix clang format * address review comment and rebase code. * fix universal test fail --------- Co-authored-by: Lin, Qun <Quentin.Lin+amdeng@amd.com> Co-authored-by: Thomas Ning <Thomas.Ning@amd.com>
This commit is contained in:
@@ -297,7 +297,7 @@ struct GemmKernel {{
|
||||
throw std::runtime_error("Wrong! Arguments not supported! Skipping gemm!");
|
||||
}}
|
||||
|
||||
constexpr dim3 blocks = Kernel::BlockSize();
|
||||
const dim3 blocks = Kernel::BlockSize();
|
||||
const dim3 grids = {'Kernel::MaxOccupancyGridSize(stream)' if persistent == 'true' else 'Kernel::GridSize(args.M, args.N, args.k_batch)'};
|
||||
|
||||
if(stream.log_level_ > 0)
|
||||
@@ -346,12 +346,12 @@ struct GemmKernel {{
|
||||
ave_time = ck_tile::launch_kernel_time_mask(
|
||||
stream,
|
||||
run_flush_cache,
|
||||
ck_tile::make_kernel<blocks.x, kBlockPerCu>(
|
||||
ck_tile::make_kernel<kBlockPerCu>(
|
||||
Kernel{{}}, grids, blocks, 0, kargs));
|
||||
}}
|
||||
else{{
|
||||
ave_time = ck_tile::launch_kernel(stream,
|
||||
ck_tile::make_kernel<blocks.x, kBlockPerCu>(
|
||||
ck_tile::make_kernel<kBlockPerCu>(
|
||||
Kernel{{}}, grids, blocks, 0, kargs));
|
||||
}}
|
||||
return ave_time;
|
||||
|
||||
Reference in New Issue
Block a user