ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-07-01 04:07:56 +00:00
Code Issues Packages Projects Releases Wiki Activity

current perf best kernel

Browse Source
This commit is contained in:
joye
2025-06-03 12:56:38 +08:00
parent d50a7ac6cb
commit 850b9adbf9
1 changed files with 2 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
include/ck/tensor_operation/gpu/device/impl/device_grouped_conv_bwd_data_multiple_d_xdl_cshuffle_v1.hpp
View File

@@ -1058,7 +1058,7 @@ struct DeviceGroupedConvBwdDataMultipleD_Xdl_CShuffle_v1
auto launch_kernel = [&]() {
// constexpr bool has_main_loop = has_main_k_block_loop.value;
constexpr index_t GroupPerBlock = 64;
constexpr index_t BatchPerBlock = 64;
constexpr index_t BatchPerBlock = 8;
const auto kernel = kernel_grouped_conv_bwd_data_optimized<ADataType,
EDataType,
GroupPerBlock,
@@ -1066,7 +1066,7 @@ struct DeviceGroupedConvBwdDataMultipleD_Xdl_CShuffle_v1
return launch_and_time_kernel(stream_config,
kernel,
dim3(1344 / GroupPerBlock * 2, 1, 1),
dim3(1344 / GroupPerBlock * 16, 1, 1),
dim3(512),
0,
p_a_grid,