ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-05-01 12:11:19 +00:00
Code Issues Packages Projects Releases Wiki Activity

Revert "[CK-Tile] Add the API to load SGPR (#2878)" (#2904)

Browse Source 
This reverts commit 2cbbf5dcb3.
This commit is contained in:
asleepzzz
2025-09-24 05:33:51 +08:00
committed by GitHub
parent 959df2a155
commit f161b5b738
40 changed files with 167 additions and 273 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
include/ck_tile/ops/rmsnorm2d/pipeline/rmsnorm2d_fwd_pipeline_two_pass.hpp
View File

@@ -82,7 +82,7 @@ struct Rmsnorm2dFwdPipelineTwoPass
// Problem::BlockShape
static constexpr index_t Block_N = Problem::BlockShape::Block_N;
index_t num_n_tile_iteration =
amd_wave_read_first_lane(integer_divide_ceil(row_size, Block_N));
__builtin_amdgcn_readfirstlane(integer_divide_ceil(row_size, Block_N));
auto reduce_square_sum_func = ReduceOp::SquareAdd{};
auto reduce_sum_func = ReduceOp::Add{};
@@ -95,7 +95,7 @@ struct Rmsnorm2dFwdPipelineTwoPass
auto square_sum = block_reduce2d.template MakeYBlockTile<ComputeTensorType>();
set_tile(square_sum, reduce_square_sum_func.GetIdentityValue<ComputeDataType>());
for(int iN = amd_wave_read_first_lane(0); iN < num_n_tile_iteration; ++iN)
for(int iN = __builtin_amdgcn_readfirstlane(0); iN < num_n_tile_iteration; ++iN)
{
auto x = load_tile(x_window);
auto x_resi = load_tile(x_residual_window);
@@ -151,7 +151,7 @@ struct Rmsnorm2dFwdPipelineTwoPass
move_tile_window(y_window, {0, stride_to_right_most_window});
// rmsnorm computation
for(int iN = amd_wave_read_first_lane(0); iN < num_n_tile_iteration; ++iN)
for(int iN = __builtin_amdgcn_readfirstlane(0); iN < num_n_tile_iteration; ++iN)
{
auto acc = make_static_distributed_tensor<ComputeDataType>(
decltype(load_tile(x_window))::get_tile_distribution());