ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-05-12 09:16:52 +00:00
Code Issues Packages Projects Releases Wiki Activity

Improve 4k gemm perf (#1047)

Browse Source 
* improve 4k gemm perf

* add f8 instances

* format

---------

Co-authored-by: Jing Zhang <jizha@amd.com>
This commit is contained in:
zjing14
2023-11-17 07:06:24 -06:00
committed by GitHub
parent e1fa00917c
commit e8cddfdc3b
7 changed files with 198 additions and 109 deletions
Download Patch File Download Diff File Expand all files Collapse all files

36
library/src/tensor_operation_instance/gpu/gemm/device_gemm_xdl_c_shuffle_f16_f16_f16_mk_kn_mn_instance.cpp
View File

@@ -58,27 +58,27 @@ using device_gemm_xdl_c_shuffle_f16_f16_f16_mk_kn_mn_instances = std::tuple<
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F16, F16, F16, F32, F16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 128, 64, 32, 8, 8, 32, 32, 2, 1, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 1, 8, 1, 1, 1, S<1, 32, 1, 8>, 8, LoopScheduler::Default, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F16, F16, F16, F32, F16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 64, 128, 32, 8, 2, 32, 32, 1, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<8, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 2, 0, 1, 1, S<1, 32, 1, 8>, 8, LoopScheduler::Default, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F16, F16, F16, F32, F16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 64, 128, 32, 8, 8, 32, 32, 1, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 2, 8, 1, 1, 1, S<1, 32, 1, 8>, 8, LoopScheduler::Default, PipelineVersion::v1>
#if CK_EXPERIMENTAL_INTER_WAVE_INSTANCES
#if CK_EXPERIMENTAL_INTER_WAVE_INSTANCES
// pipeline v1, 2 waves
,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F16, F16, F16, F32, F16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 256, 128, 32, 8, 2, 32, 32, 4, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<8, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 2, 0, 1, 1, S<1, 32, 1, 8>, 8, LoopScheduler::Interwave, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F16, F16, F16, F32, F16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 256, 128, 32, 8, 8, 32, 32, 4, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 2, 8, 1, 1, 1, S<1, 32, 1, 8>, 8, LoopScheduler::Interwave, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F16, F16, F16, F32, F16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 128, 256, 32, 8, 2, 32, 32, 2, 4, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 2, 0, 1, 1, S<1, 32, 1, 8>, 8, LoopScheduler::Interwave, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F16, F16, F16, F32, F16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 128, 256, 32, 8, 8, 32, 32, 2, 4, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 8, 1, 1, 1, S<1, 32, 1, 8>, 8, LoopScheduler::Interwave, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F16, F16, F16, F32, F16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 128, 128, 128, 32, 8, 2, 32, 32, 4, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 2, 0, 1, 1, S<1, 16, 1, 8>, 8, LoopScheduler::Interwave, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F16, F16, F16, F32, F16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 128, 128, 128, 32, 8, 8, 32, 32, 4, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 8, 1, 1, 1, S<1, 16, 1, 8>, 8, LoopScheduler::Interwave, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F16, F16, F16, F32, F16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 128, 128, 32, 8, 2, 32, 32, 2, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<8, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 2, 0, 1, 1, S<1, 32, 1, 8>, 8, LoopScheduler::Interwave, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F16, F16, F16, F32, F16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 128, 128, 32, 8, 8, 32, 32, 2, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 2, 8, 1, 1, 1, S<1, 32, 1, 8>, 8, LoopScheduler::Interwave, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F16, F16, F16, F32, F16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 128, 128, 64, 32, 8, 2, 32, 32, 2, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<8, 16, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 2, 0, 1, 1, S<1, 32, 1, 4>, 8, LoopScheduler::Interwave, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F16, F16, F16, F32, F16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 128, 128, 64, 32, 8, 8, 32, 32, 2, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 2, 8, 1, 1, 1, S<1, 32, 1, 4>, 8, LoopScheduler::Interwave, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F16, F16, F16, F32, F16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 128, 64, 128, 32, 8, 2, 32, 32, 2, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 2, 0, 1, 1, S<1, 16, 1, 8>, 8, LoopScheduler::Interwave, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F16, F16, F16, F32, F16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 128, 64, 128, 32, 8, 8, 32, 32, 2, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 8, 1, 1, 1, S<1, 16, 1, 8>, 8, LoopScheduler::Interwave, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F16, F16, F16, F32, F16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 128, 64, 32, 8, 2, 32, 32, 2, 1, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<16,16, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 2, 0, 1, 1, S<1, 32, 1, 8>, 8, LoopScheduler::Interwave, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F16, F16, F16, F32, F16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 128, 64, 32, 8, 8, 32, 32, 2, 1, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 1, 8, 1, 1, 1, S<1, 32, 1, 8>, 8, LoopScheduler::Interwave, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F16, F16, F16, F32, F16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 64, 128, 32, 8, 2, 32, 32, 1, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<8, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 2, 0, 1, 1, S<1, 32, 1, 8>, 8, LoopScheduler::Interwave, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F16, F16, F16, F32, F16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 64, 128, 32, 8, 8, 32, 32, 1, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 2, 8, 1, 1, 1, S<1, 32, 1, 8>, 8, LoopScheduler::Interwave, PipelineVersion::v1>
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F16, F16, F16, F32, F16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 256, 128, 32, 8, 2, 32, 32, 4, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<8, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 2, 0, 1, 1, S<1, 32, 1, 8>, 8, LoopScheduler::Interwave, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F16, F16, F16, F32, F16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 256, 128, 32, 8, 8, 32, 32, 4, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 2, 8, 1, 1, 1, S<1, 32, 1, 8>, 8, LoopScheduler::Interwave, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F16, F16, F16, F32, F16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 128, 256, 32, 8, 2, 32, 32, 2, 4, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 2, 0, 1, 1, S<1, 32, 1, 8>, 8, LoopScheduler::Interwave, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F16, F16, F16, F32, F16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 128, 256, 32, 8, 8, 32, 32, 2, 4, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 8, 1, 1, 1, S<1, 32, 1, 8>, 8, LoopScheduler::Interwave, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F16, F16, F16, F32, F16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 128, 128, 128, 32, 8, 2, 32, 32, 4, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 2, 0, 1, 1, S<1, 16, 1, 8>, 8, LoopScheduler::Interwave, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F16, F16, F16, F32, F16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 128, 128, 128, 32, 8, 8, 32, 32, 4, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 8, 1, 1, 1, S<1, 16, 1, 8>, 8, LoopScheduler::Interwave, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F16, F16, F16, F32, F16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 128, 128, 32, 8, 2, 32, 32, 2, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<8, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 2, 0, 1, 1, S<1, 32, 1, 8>, 8, LoopScheduler::Interwave, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F16, F16, F16, F32, F16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 128, 128, 32, 8, 8, 32, 32, 2, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 2, 8, 1, 1, 1, S<1, 32, 1, 8>, 8, LoopScheduler::Interwave, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F16, F16, F16, F32, F16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 128, 128, 64, 32, 8, 2, 32, 32, 2, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<8, 16, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 2, 0, 1, 1, S<1, 32, 1, 4>, 8, LoopScheduler::Interwave, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F16, F16, F16, F32, F16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 128, 128, 64, 32, 8, 8, 32, 32, 2, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 2, 8, 1, 1, 1, S<1, 32, 1, 4>, 8, LoopScheduler::Interwave, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F16, F16, F16, F32, F16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 128, 64, 128, 32, 8, 2, 32, 32, 2, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 2, 0, 1, 1, S<1, 16, 1, 8>, 8, LoopScheduler::Interwave, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F16, F16, F16, F32, F16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 128, 64, 128, 32, 8, 8, 32, 32, 2, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 8, 1, 1, 1, S<1, 16, 1, 8>, 8, LoopScheduler::Interwave, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F16, F16, F16, F32, F16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 128, 64, 32, 8, 2, 32, 32, 2, 1, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<16,16, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 2, 0, 1, 1, S<1, 32, 1, 8>, 8, LoopScheduler::Interwave, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F16, F16, F16, F32, F16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 128, 64, 32, 8, 8, 32, 32, 2, 1, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 1, 8, 1, 1, 1, S<1, 32, 1, 8>, 8, LoopScheduler::Interwave, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F16, F16, F16, F32, F16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 64, 128, 32, 8, 2, 32, 32, 1, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<8, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 2, 0, 1, 1, S<1, 32, 1, 8>, 8, LoopScheduler::Interwave, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F16, F16, F16, F32, F16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 64, 128, 32, 8, 8, 32, 32, 1, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 2, 8, 1, 1, 1, S<1, 32, 1, 8>, 8, LoopScheduler::Interwave, PipelineVersion::v1>
#endif
#if CK_EXPERIMENTAL_PIPELINE_V2_INSTANCES
#if CK_EXPERIMENTAL_PIPELINE_V2_INSTANCES
// pipeline v2, 1 wave
,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F16, F16, F16, F32, F16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 256, 128, 32, 8, 2, 32, 32, 4, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<8, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 2, 0, 1, 1, S<1, 32, 1, 8>, 8, LoopScheduler::Default, PipelineVersion::v2>,

91
library/src/tensor_operation_instance/gpu/gemm/device_gemm_xdl_c_shuffle_fp8_fp8_fp8_mk_kn_mn_instance.cpp
View File

@@ -31,31 +31,72 @@ static constexpr auto MNKPadding = ck::tensor_operation::device::GemmSpecializat
// Compilation parameters for a[m, k] * b[k, n] = c[m, n]
template <ck::tensor_operation::device::GemmSpecialization GemmSpec>
using device_gemm_xdl_c_shuffle_f8_f8_f8_mk_kn_mn_instances =
std::tuple<
// clang-format off
//#####################| ALayout| BLayout| CLayout| AData| BData| CData| AccData| CShuffle| A| B| C| GEMM| NumGemmK| Block| MPer| NPer| KPer| AK1| BK1| MPer| NPer| MXdl| NXdl| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds| BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds| CShuffle| CShuffle| CBlockTransferClusterLengths| CBlockTransfer|
//#####################| | | | Type| Type| Type| Type| DataType| Elementwise| Elementwise| Elementwise| Specialization| Prefetch| Size| Block| Block| Block| | | XDL| XDL| Per| Per| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| AddExtraM| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| _MBlock_MWaveMPerXdl| ScalarPerVector|
//#####################| | | | | | | | | Operation| Operation| Operation| | Stage| | | | | | | | | Wave| Wave| Lengths_K0_M_K1| ArrangeOrder| | | PerVector| PerVector_K1| | Lengths_K0_N_K1| ArrangeOrder| | | PerVector| PerVector_K1| | PerShuffle| PerShuffle| _NBlock_NWaveNPerXdl| _NWaveNPerXdl|
//#####################| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | |
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F8, F8, F8, F32, F8, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 256, 128, 64, 16, 4, 32, 32, 4, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, 1, S<8, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 4, 0, 1, 1, S<1, 64, 1, 4>, 16>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F8, F8, F8, F32, F8, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 256, 128, 64, 16, 16, 32, 32, 4, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, 1, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 2, 16, 1, 1, 1, S<1, 64, 1, 4>, 16>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F8, F8, F8, F32, F8, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 128, 256, 64, 16, 4, 32, 32, 2, 4, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, 1, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 4, 0, 1, 1, S<1, 64, 1, 4>, 16>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F8, F8, F8, F32, F8, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 128, 256, 64, 16, 16, 32, 32, 2, 4, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, 1, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 16, 1, 1, 1, S<1, 64, 1, 4>, 16>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F8, F8, F8, F32, F8, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 128, 128, 128, 64, 16, 4, 32, 32, 4, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, 1, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 4, 0, 1, 1, S<1, 32, 1, 4>, 16>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F8, F8, F8, F32, F8, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 128, 128, 128, 64, 16, 16, 32, 32, 4, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, 1, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 16, 1, 1, 1, S<1, 32, 1, 4>, 16>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F8, F8, F8, F32, F8, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 128, 128, 64, 16, 4, 32, 32, 2, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, 1, S<8, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 4, 0, 1, 1, S<1, 64, 1, 4>, 16>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F8, F8, F8, F32, F8, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 128, 128, 64, 16, 16, 32, 32, 2, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, 1, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 2, 16, 1, 1, 1, S<1, 64, 1, 4>, 16>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F8, F8, F8, F32, F8, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 128, 128, 64, 64, 16, 4, 32, 32, 2, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, 1, S<8, 16, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 4, 0, 1, 1, S<1, 64, 1, 2>, 16>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F8, F8, F8, F32, F8, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 128, 128, 64, 64, 16, 16, 32, 32, 2, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, 1, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 2, 16, 1, 1, 1, S<1, 64, 1, 2>, 16>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F8, F8, F8, F32, F8, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 128, 64, 128, 64, 16, 4, 32, 32, 2, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, 1, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 4, 0, 1, 1, S<1, 32, 1, 4>, 16>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F8, F8, F8, F32, F8, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 128, 64, 128, 64, 16, 16, 32, 32, 2, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, 1, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 16, 1, 1, 1, S<1, 32, 1, 4>, 16>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F8, F8, F8, F32, F8, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 128, 64, 64, 16, 4, 32, 32, 2, 1, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, 1, S<16,16, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 4, 0, 1, 1, S<1, 64, 1, 4>, 16>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F8, F8, F8, F32, F8, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 128, 64, 64, 16, 16, 32, 32, 2, 1, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, 1, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 1, 16, 1, 1, 1, S<1, 64, 1, 4>, 16>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F8, F8, F8, F32, F8, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 64, 128, 64, 16, 4, 32, 32, 1, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, 1, S<8, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 4, 0, 1, 1, S<1, 64, 1, 4>, 16>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F8, F8, F8, F32, F8, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 64, 128, 64, 16, 16, 32, 32, 1, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, 1, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 2, 16, 1, 1, 1, S<1, 64, 1, 4>, 16>
// clang-format on
>;
using device_gemm_xdl_c_shuffle_f8_f8_f8_mk_kn_mn_instances = std::tuple<
// clang-format off
//#####################| ALayout| BLayout| CLayout| AData| BData| CData| AccData| CShuffle| A| B| C| GEMM| NumGemmK| Block| MPer| NPer| KPer| AK1| BK1| MPer| NPer| MXdl| NXdl| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds| BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds| CShuffle| CShuffle| CBlockTransferClusterLengths| CBlockTransfer| LoopScheduler| Pipeline|
//#####################| | | | Type| Type| Type| Type| DataType| Elementwise| Elementwise| Elementwise| Specialization| Prefetch| Size| Block| Block| Block| | | XDL| XDL| Per| Per| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| AddExtraM| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| _MBlock_MWaveMPerXdl| ScalarPerVector| | |
//#####################| | | | | | | | | Operation| Operation| Operation| | Stage| | | | | | | | | Wave| Wave| Lengths_K0_M_K1| ArrangeOrder| | | PerVector| PerVector_K1| | Lengths_K0_N_K1| ArrangeOrder| | | PerVector| PerVector_K1| | PerShuffle| PerShuffle| _NBlock_NWaveNPerXdl| _NWaveNPerXdl| | |
//#####################| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | |
// pipeline v1, 1 wave
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F8, F8, F8, F32, F8, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 256, 128, 64, 16, 4, 32, 32, 4, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, 1, S<8, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 4, 0, 1, 1, S<1, 64, 1, 4>, 16, LoopScheduler::Default, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F8, F8, F8, F32, F8, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 256, 128, 64, 16, 16, 32, 32, 4, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, 1, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 2, 16, 1, 1, 1, S<1, 64, 1, 4>, 16, LoopScheduler::Default, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F8, F8, F8, F32, F8, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 128, 256, 64, 16, 4, 32, 32, 2, 4, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, 1, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 4, 0, 1, 1, S<1, 64, 1, 4>, 16, LoopScheduler::Default, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F8, F8, F8, F32, F8, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 128, 256, 64, 16, 16, 32, 32, 2, 4, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, 1, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 16, 1, 1, 1, S<1, 64, 1, 4>, 16, LoopScheduler::Default, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F8, F8, F8, F32, F8, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 128, 128, 128, 64, 16, 4, 32, 32, 4, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, 1, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 4, 0, 1, 1, S<1, 32, 1, 4>, 16, LoopScheduler::Default, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F8, F8, F8, F32, F8, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 128, 128, 128, 64, 16, 16, 32, 32, 4, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, 1, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 16, 1, 1, 1, S<1, 32, 1, 4>, 16, LoopScheduler::Default, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F8, F8, F8, F32, F8, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 128, 128, 64, 16, 4, 32, 32, 2, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, 1, S<8, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 4, 0, 1, 1, S<1, 64, 1, 4>, 16, LoopScheduler::Default, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F8, F8, F8, F32, F8, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 128, 128, 64, 16, 16, 32, 32, 2, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, 1, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 2, 16, 1, 1, 1, S<1, 64, 1, 4>, 16, LoopScheduler::Default, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F8, F8, F8, F32, F8, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 128, 128, 64, 64, 16, 4, 32, 32, 2, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, 1, S<8, 16, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 4, 0, 1, 1, S<1, 64, 1, 2>, 16, LoopScheduler::Default, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F8, F8, F8, F32, F8, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 128, 128, 64, 64, 16, 16, 32, 32, 2, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, 1, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 2, 16, 1, 1, 1, S<1, 64, 1, 2>, 16, LoopScheduler::Default, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F8, F8, F8, F32, F8, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 128, 64, 128, 64, 16, 4, 32, 32, 2, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, 1, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 4, 0, 1, 1, S<1, 32, 1, 4>, 16, LoopScheduler::Default, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F8, F8, F8, F32, F8, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 128, 64, 128, 64, 16, 16, 32, 32, 2, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, 1, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 16, 1, 1, 1, S<1, 32, 1, 4>, 16, LoopScheduler::Default, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F8, F8, F8, F32, F8, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 128, 64, 64, 16, 4, 32, 32, 2, 1, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, 1, S<16,16, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 4, 0, 1, 1, S<1, 64, 1, 4>, 16, LoopScheduler::Default, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F8, F8, F8, F32, F8, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 128, 64, 64, 16, 16, 32, 32, 2, 1, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, 1, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 1, 16, 1, 1, 1, S<1, 64, 1, 4>, 16, LoopScheduler::Default, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F8, F8, F8, F32, F8, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 64, 128, 64, 16, 4, 32, 32, 1, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, 1, S<8, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 4, 0, 1, 1, S<1, 64, 1, 4>, 16, LoopScheduler::Default, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F8, F8, F8, F32, F8, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 64, 128, 64, 16, 16, 32, 32, 1, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, 1, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 2, 16, 1, 1, 1, S<1, 64, 1, 4>, 16, LoopScheduler::Default, PipelineVersion::v1>
#if CK_EXPERIMENTAL_INTER_WAVE_INSTANCES
// pipeline v1, 2 waves
,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F8, F8, F8, F32, F8, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 256, 128, 64, 16, 4, 32, 32, 4, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, 1, S<8, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 4, 0, 1, 1, S<1, 64, 1, 4>, 16, LoopScheduler::Interwave, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F8, F8, F8, F32, F8, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 256, 128, 64, 16, 16, 32, 32, 4, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, 1, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 2, 16, 1, 1, 1, S<1, 64, 1, 4>, 16, LoopScheduler::Interwave, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F8, F8, F8, F32, F8, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 128, 256, 64, 16, 4, 32, 32, 2, 4, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, 1, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 4, 0, 1, 1, S<1, 64, 1, 4>, 16, LoopScheduler::Interwave, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F8, F8, F8, F32, F8, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 128, 256, 64, 16, 16, 32, 32, 2, 4, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, 1, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 16, 1, 1, 1, S<1, 64, 1, 4>, 16, LoopScheduler::Interwave, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F8, F8, F8, F32, F8, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 128, 128, 128, 64, 16, 4, 32, 32, 4, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, 1, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 4, 0, 1, 1, S<1, 32, 1, 4>, 16, LoopScheduler::Interwave, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F8, F8, F8, F32, F8, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 128, 128, 128, 64, 16, 16, 32, 32, 4, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, 1, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 16, 1, 1, 1, S<1, 32, 1, 4>, 16, LoopScheduler::Interwave, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F8, F8, F8, F32, F8, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 128, 128, 64, 16, 4, 32, 32, 2, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, 1, S<8, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 4, 0, 1, 1, S<1, 64, 1, 4>, 16, LoopScheduler::Interwave, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F8, F8, F8, F32, F8, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 128, 128, 64, 16, 16, 32, 32, 2, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, 1, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 2, 16, 1, 1, 1, S<1, 64, 1, 4>, 16, LoopScheduler::Interwave, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F8, F8, F8, F32, F8, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 128, 128, 64, 64, 16, 4, 32, 32, 2, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, 1, S<8, 16, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 4, 0, 1, 1, S<1, 64, 1, 2>, 16, LoopScheduler::Interwave, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F8, F8, F8, F32, F8, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 128, 128, 64, 64, 16, 16, 32, 32, 2, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, 1, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 2, 16, 1, 1, 1, S<1, 64, 1, 2>, 16, LoopScheduler::Interwave, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F8, F8, F8, F32, F8, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 128, 64, 128, 64, 16, 4, 32, 32, 2, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, 1, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 4, 0, 1, 1, S<1, 32, 1, 4>, 16, LoopScheduler::Interwave, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F8, F8, F8, F32, F8, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 128, 64, 128, 64, 16, 16, 32, 32, 2, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, 1, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 16, 1, 1, 1, S<1, 32, 1, 4>, 16, LoopScheduler::Interwave, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F8, F8, F8, F32, F8, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 128, 64, 64, 16, 4, 32, 32, 2, 1, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, 1, S<16,16, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 4, 0, 1, 1, S<1, 64, 1, 4>, 16, LoopScheduler::Interwave, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F8, F8, F8, F32, F8, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 128, 64, 64, 16, 16, 32, 32, 2, 1, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, 1, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 1, 16, 1, 1, 1, S<1, 64, 1, 4>, 16, LoopScheduler::Interwave, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F8, F8, F8, F32, F8, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 64, 128, 64, 16, 4, 32, 32, 1, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, 1, S<8, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 4, 0, 1, 1, S<1, 64, 1, 4>, 16, LoopScheduler::Interwave, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F8, F8, F8, F32, F8, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 64, 128, 64, 16, 16, 32, 32, 1, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, 1, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 2, 16, 1, 1, 1, S<1, 64, 1, 4>, 16, LoopScheduler::Interwave, PipelineVersion::v1>
#endif
#if CK_EXPERIMENTAL_PIPELINE_V2_INSTANCES
// pipeline v2, 1 wave
,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F8, F8, F8, F32, F8, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 256, 128, 64, 16, 4, 32, 32, 4, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, 1, S<8, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 4, 0, 1, 1, S<1, 64, 1, 4>, 16, LoopScheduler::Default, PipelineVersion::v2>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F8, F8, F8, F32, F8, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 256, 128, 64, 16, 16, 32, 32, 4, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, 1, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 2, 16, 1, 1, 1, S<1, 64, 1, 4>, 16, LoopScheduler::Default, PipelineVersion::v2>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F8, F8, F8, F32, F8, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 128, 256, 64, 16, 4, 32, 32, 2, 4, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, 1, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 4, 0, 1, 1, S<1, 64, 1, 4>, 16, LoopScheduler::Default, PipelineVersion::v2>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F8, F8, F8, F32, F8, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 128, 256, 64, 16, 16, 32, 32, 2, 4, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, 1, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 16, 1, 1, 1, S<1, 64, 1, 4>, 16, LoopScheduler::Default, PipelineVersion::v2>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F8, F8, F8, F32, F8, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 128, 128, 128, 64, 16, 4, 32, 32, 4, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, 1, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 4, 0, 1, 1, S<1, 32, 1, 4>, 16, LoopScheduler::Default, PipelineVersion::v2>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F8, F8, F8, F32, F8, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 128, 128, 128, 64, 16, 16, 32, 32, 4, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, 1, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 16, 1, 1, 1, S<1, 32, 1, 4>, 16, LoopScheduler::Default, PipelineVersion::v2>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F8, F8, F8, F32, F8, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 128, 128, 64, 16, 4, 32, 32, 2, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, 1, S<8, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 4, 0, 1, 1, S<1, 64, 1, 4>, 16, LoopScheduler::Default, PipelineVersion::v2>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F8, F8, F8, F32, F8, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 128, 128, 64, 16, 16, 32, 32, 2, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, 1, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 2, 16, 1, 1, 1, S<1, 64, 1, 4>, 16, LoopScheduler::Default, PipelineVersion::v2>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F8, F8, F8, F32, F8, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 128, 128, 64, 64, 16, 4, 32, 32, 2, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, 1, S<8, 16, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 4, 0, 1, 1, S<1, 64, 1, 2>, 16, LoopScheduler::Default, PipelineVersion::v2>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F8, F8, F8, F32, F8, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 128, 128, 64, 64, 16, 16, 32, 32, 2, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, 1, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 2, 16, 1, 1, 1, S<1, 64, 1, 2>, 16, LoopScheduler::Default, PipelineVersion::v2>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F8, F8, F8, F32, F8, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 128, 64, 128, 64, 16, 4, 32, 32, 2, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, 1, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 4, 0, 1, 1, S<1, 32, 1, 4>, 16, LoopScheduler::Default, PipelineVersion::v2>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F8, F8, F8, F32, F8, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 128, 64, 128, 64, 16, 16, 32, 32, 2, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, 1, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 16, 1, 1, 1, S<1, 32, 1, 4>, 16, LoopScheduler::Default, PipelineVersion::v2>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F8, F8, F8, F32, F8, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 128, 64, 64, 16, 4, 32, 32, 2, 1, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, 1, S<16,16, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 4, 0, 1, 1, S<1, 64, 1, 4>, 16, LoopScheduler::Default, PipelineVersion::v2>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F8, F8, F8, F32, F8, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 128, 64, 64, 16, 16, 32, 32, 2, 1, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, 1, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 1, 16, 1, 1, 1, S<1, 64, 1, 4>, 16, LoopScheduler::Default, PipelineVersion::v2>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F8, F8, F8, F32, F8, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 64, 128, 64, 16, 4, 32, 32, 1, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, 1, S<8, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 4, 0, 1, 1, S<1, 64, 1, 4>, 16, LoopScheduler::Default, PipelineVersion::v2>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, F8, F8, F8, F32, F8, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 64, 128, 64, 16, 16, 32, 32, 1, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, 1, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 2, 16, 1, 1, 1, S<1, 64, 1, 4>, 16, LoopScheduler::Default, PipelineVersion::v2>
#endif
// clang-format on
>;
void add_device_gemm_xdl_c_shuffle_f8_f8_f8_mk_kn_mn_instances(
std::vector<std::unique_ptr<