Skip to content

GitLab

Commit 82a7b60d authored by Po-Yen, Chen's avatar Po-Yen, Chen
Browse files

Add instance for layout=TN

parent 0345963e
Hide whitespace changes
Inline Side-by-side
Showing with 287 additions and 246 deletions
include/ck/ck.hpp
View file @ 82a7b60d
......@@ -140,7 +140,7 @@
// this will let make_default_loop_scheduler() return interwave scheduling flag by default
#define CK_EXPERIMENTAL_DEFAULT_TO_INTER_WAVE_SCHEDULING 0
// experimental feature: add instances using interwave scheduling
#define CK_EXPERIMENTAL_INTER_WAVE_INSTANCES 1
#define CK_EXPERIMENTAL_INTER_WAVE_INSTANCES 0
// experimental feature: add instances using pipeline v2
#define CK_EXPERIMENTAL_PIPELINE_V2_INSTANCES 1
......
include/ck/tensor_operation/gpu/grid/gridwise_gemm_pipeline_v2.hpp
View file @ 82a7b60d
......@@ -79,6 +79,10 @@ struct GridwiseGemmPipeline_v2
do
{
#ifdef USE_IGLP_OPT
__builtin_amdgcn_iglp_opt(1);
#endif
block_sync_lds();
// GEMM i
......
include/ck/tensor_operation/gpu/grid/gridwise_gemm_xdlops_v2r3.hpp
View file @ 82a7b60d
......@@ -27,6 +27,9 @@ template <typename GridwiseGemm,
typename CElementwiseOperation,
typename Block2CTileMap,
bool HasMainKBlockLoop>
#ifdef USE_WAVES_PER_EU
__attribute__((amdgpu_waves_per_eu(1, 1)))
#endif
__global__ void
#if CK_USE_LAUNCH_BOUNDS
__launch_bounds__(CK_MAX_THREAD_PER_BLOCK, CK_MIN_BLOCK_PER_CU)
......
library/include/ck/library/tensor_operation_instance/gpu/gemm.hpp
View file @ 82a7b60d
......@@ -257,123 +257,122 @@ struct DeviceOperationInstanceFactory<
{
std::vector<std::unique_ptr<DeviceOp>> op_ptrs;
if constexpr(is_same_v<ADataType, float> && is_same_v<BDataType, float> &&
is_same_v<CDataType, float>)
{
if constexpr(is_same_v<ALayout, Row> && is_same_v<BLayout, Row> &&
is_same_v<CLayout, Row>)
{
add_device_gemm_xdl_f32_f32_f32_mk_kn_mn_instances(op_ptrs);
add_device_gemm_dl_f32_f32_f32_mk_kn_mn_instances(op_ptrs);
add_device_gemm_xdl_c_shuffle_f32_f32_f32_mk_kn_mn_instances(op_ptrs);
}
else if constexpr(is_same_v<ALayout, Row> && is_same_v<BLayout, Col> &&
is_same_v<CLayout, Row>)
{
add_device_gemm_xdl_f32_f32_f32_mk_nk_mn_instances(op_ptrs);
add_device_gemm_dl_f32_f32_f32_mk_nk_mn_instances(op_ptrs);
add_device_gemm_xdl_c_shuffle_f32_f32_f32_mk_nk_mn_instances(op_ptrs);
}
else if constexpr(is_same_v<ALayout, Col> && is_same_v<BLayout, Row> &&
is_same_v<CLayout, Row>)
{
add_device_gemm_xdl_f32_f32_f32_km_kn_mn_instances(op_ptrs);
add_device_gemm_dl_f32_f32_f32_km_kn_mn_instances(op_ptrs);
add_device_gemm_xdl_c_shuffle_f32_f32_f32_km_kn_mn_instances(op_ptrs);
}
else if constexpr(is_same_v<ALayout, Col> && is_same_v<BLayout, Col> &&
is_same_v<CLayout, Row>)
{
add_device_gemm_xdl_f32_f32_f32_km_nk_mn_instances(op_ptrs);
add_device_gemm_dl_f32_f32_f32_km_nk_mn_instances(op_ptrs);
add_device_gemm_xdl_c_shuffle_f32_f32_f32_km_nk_mn_instances(op_ptrs);
}
}
else if constexpr(is_same_v<ADataType, half_t> && is_same_v<BDataType, half_t> &&
// if constexpr(is_same_v<ADataType, float> && is_same_v<BDataType, float> &&
// is_same_v<CDataType, float>)
// {
// if constexpr(is_same_v<ALayout, Row> && is_same_v<BLayout, Row> &&
// is_same_v<CLayout, Row>)
// {
// add_device_gemm_xdl_f32_f32_f32_mk_kn_mn_instances(op_ptrs);
// add_device_gemm_dl_f32_f32_f32_mk_kn_mn_instances(op_ptrs);
// add_device_gemm_xdl_c_shuffle_f32_f32_f32_mk_kn_mn_instances(op_ptrs);
// }
// else if constexpr(is_same_v<ALayout, Row> && is_same_v<BLayout, Col> &&
// is_same_v<CLayout, Row>)
// {
// add_device_gemm_xdl_f32_f32_f32_mk_nk_mn_instances(op_ptrs);
// add_device_gemm_dl_f32_f32_f32_mk_nk_mn_instances(op_ptrs);
// add_device_gemm_xdl_c_shuffle_f32_f32_f32_mk_nk_mn_instances(op_ptrs);
// }
// else if constexpr(is_same_v<ALayout, Col> && is_same_v<BLayout, Row> &&
// is_same_v<CLayout, Row>)
// {
// add_device_gemm_xdl_f32_f32_f32_km_kn_mn_instances(op_ptrs);
// add_device_gemm_dl_f32_f32_f32_km_kn_mn_instances(op_ptrs);
// add_device_gemm_xdl_c_shuffle_f32_f32_f32_km_kn_mn_instances(op_ptrs);
// }
// else if constexpr(is_same_v<ALayout, Col> && is_same_v<BLayout, Col> &&
// is_same_v<CLayout, Row>)
// {
// add_device_gemm_xdl_f32_f32_f32_km_nk_mn_instances(op_ptrs);
// add_device_gemm_dl_f32_f32_f32_km_nk_mn_instances(op_ptrs);
// add_device_gemm_xdl_c_shuffle_f32_f32_f32_km_nk_mn_instances(op_ptrs);
// }
// }
if constexpr(is_same_v<ADataType, half_t> && is_same_v<BDataType, half_t> &&
is_same_v<CDataType, half_t>)
{
if constexpr(is_same_v<ALayout, Row> && is_same_v<BLayout, Row> &&
is_same_v<CLayout, Row>)
{
add_device_gemm_xdl_f16_f16_f16_mk_kn_mn_instances(op_ptrs);
add_device_gemm_dl_f16_f16_f16_mk_kn_mn_instances(op_ptrs);
add_device_gemm_xdl_c_shuffle_f16_f16_f16_mk_kn_mn_instances(op_ptrs);
// add_device_gemm_xdl_f16_f16_f16_mk_kn_mn_instances(op_ptrs);
// add_device_gemm_dl_f16_f16_f16_mk_kn_mn_instances(op_ptrs);
// add_device_gemm_xdl_c_shuffle_f16_f16_f16_mk_kn_mn_instances(op_ptrs);
}
else if constexpr(is_same_v<ALayout, Row> && is_same_v<BLayout, Col> &&
is_same_v<CLayout, Row>)
{
add_device_gemm_xdl_f16_f16_f16_mk_nk_mn_instances(op_ptrs);
add_device_gemm_dl_f16_f16_f16_mk_nk_mn_instances(op_ptrs);
add_device_gemm_xdl_c_shuffle_f16_f16_f16_mk_nk_mn_instances(op_ptrs);
add_device_gemm_xdl_c_shuffle_2_stage_f16_f16_f16_mk_nk_mn_instances(op_ptrs);
}
else if constexpr(is_same_v<ALayout, Col> && is_same_v<BLayout, Row> &&
is_same_v<CLayout, Row>)
{
add_device_gemm_xdl_f16_f16_f16_km_kn_mn_instances(op_ptrs);
add_device_gemm_dl_f16_f16_f16_km_kn_mn_instances(op_ptrs);
add_device_gemm_xdl_c_shuffle_f16_f16_f16_km_kn_mn_instances(op_ptrs);
}
else if constexpr(is_same_v<ALayout, Col> && is_same_v<BLayout, Col> &&
is_same_v<CLayout, Row>)
{
add_device_gemm_xdl_f16_f16_f16_km_nk_mn_instances(op_ptrs);
add_device_gemm_dl_f16_f16_f16_km_nk_mn_instances(op_ptrs);
add_device_gemm_xdl_c_shuffle_f16_f16_f16_km_nk_mn_instances(op_ptrs);
}
}
else if constexpr(is_same_v<ADataType, ck::bhalf_t> && is_same_v<BDataType, ck::bhalf_t> &&
is_same_v<CDataType, ck::bhalf_t>)
{
if constexpr(is_same_v<ALayout, Row> && is_same_v<BLayout, Row> &&
is_same_v<CLayout, Row>)
{
add_device_gemm_xdl_c_shuffle_bf16_bf16_bf16_mk_kn_mn_instances(op_ptrs);
}
else if constexpr(is_same_v<ALayout, Row> && is_same_v<BLayout, Col> &&
is_same_v<CLayout, Row>)
{
add_device_gemm_xdl_c_shuffle_bf16_bf16_bf16_mk_nk_mn_instances(op_ptrs);
}
else if constexpr(is_same_v<ALayout, Col> && is_same_v<BLayout, Row> &&
is_same_v<CLayout, Row>)
{
add_device_gemm_xdl_c_shuffle_bf16_bf16_bf16_km_kn_mn_instances(op_ptrs);
}
else if constexpr(is_same_v<ALayout, Col> && is_same_v<BLayout, Col> &&
is_same_v<CLayout, Row>)
{
add_device_gemm_xdl_c_shuffle_bf16_bf16_bf16_km_nk_mn_instances(op_ptrs);
}
}
else if constexpr(is_same_v<ADataType, int8_t> && is_same_v<BDataType, int8_t> &&
is_same_v<CDataType, int8_t>)
{
if constexpr(is_same_v<ALayout, Row> && is_same_v<BLayout, Row> &&
is_same_v<CLayout, Row>)
{
add_device_gemm_xdl_c_shuffle_i8_i8_i8_mk_kn_mn_instances(op_ptrs);
add_device_gemm_dl_i8_i8_i8_mk_kn_mn_instances(op_ptrs);
}
else if constexpr(is_same_v<ALayout, Row> && is_same_v<BLayout, Col> &&
is_same_v<CLayout, Row>)
{
add_device_gemm_xdl_c_shuffle_i8_i8_i8_mk_nk_mn_instances(op_ptrs);
add_device_gemm_dl_i8_i8_i8_mk_nk_mn_instances(op_ptrs);
// add_device_gemm_xdl_c_shuffle_f16_f16_f16_mk_nk_mn_instances(op_ptrs);
// add_device_gemm_xdl_c_shuffle_2_stage_f16_f16_f16_mk_nk_mn_instances(op_ptrs);
}
else if constexpr(is_same_v<ALayout, Col> && is_same_v<BLayout, Row> &&
is_same_v<CLayout, Row>)
{
add_device_gemm_xdl_c_shuffle_i8_i8_i8_km_kn_mn_instances(op_ptrs);
add_device_gemm_dl_i8_i8_i8_km_kn_mn_instances(op_ptrs);
// add_device_gemm_xdl_f16_f16_f16_km_kn_mn_instances(op_ptrs);
// add_device_gemm_dl_f16_f16_f16_km_kn_mn_instances(op_ptrs);
// add_device_gemm_xdl_c_shuffle_f16_f16_f16_km_kn_mn_instances(op_ptrs);
}
else if constexpr(is_same_v<ALayout, Col> && is_same_v<BLayout, Col> &&
is_same_v<CLayout, Row>)
{
add_device_gemm_xdl_c_shuffle_i8_i8_i8_km_nk_mn_instances(op_ptrs);
add_device_gemm_dl_i8_i8_i8_km_nk_mn_instances(op_ptrs);
// add_device_gemm_xdl_f16_f16_f16_km_nk_mn_instances(op_ptrs);
// add_device_gemm_dl_f16_f16_f16_km_nk_mn_instances(op_ptrs);
// add_device_gemm_xdl_c_shuffle_f16_f16_f16_km_nk_mn_instances(op_ptrs);
}
}
// else if constexpr(is_same_v<ADataType, ck::bhalf_t> && is_same_v<BDataType, ck::bhalf_t> &&
// is_same_v<CDataType, ck::bhalf_t>)
// {
// if constexpr(is_same_v<ALayout, Row> && is_same_v<BLayout, Row> &&
// is_same_v<CLayout, Row>)
// {
// add_device_gemm_xdl_c_shuffle_bf16_bf16_bf16_mk_kn_mn_instances(op_ptrs);
// }
// else if constexpr(is_same_v<ALayout, Row> && is_same_v<BLayout, Col> &&
// is_same_v<CLayout, Row>)
// {
// add_device_gemm_xdl_c_shuffle_bf16_bf16_bf16_mk_nk_mn_instances(op_ptrs);
// }
// else if constexpr(is_same_v<ALayout, Col> && is_same_v<BLayout, Row> &&
// is_same_v<CLayout, Row>)
// {
// add_device_gemm_xdl_c_shuffle_bf16_bf16_bf16_km_kn_mn_instances(op_ptrs);
// }
// else if constexpr(is_same_v<ALayout, Col> && is_same_v<BLayout, Col> &&
// is_same_v<CLayout, Row>)
// {
// add_device_gemm_xdl_c_shuffle_bf16_bf16_bf16_km_nk_mn_instances(op_ptrs);
// }
// }
// else if constexpr(is_same_v<ADataType, int8_t> && is_same_v<BDataType, int8_t> &&
// is_same_v<CDataType, int8_t>)
// {
// if constexpr(is_same_v<ALayout, Row> && is_same_v<BLayout, Row> &&
// is_same_v<CLayout, Row>)
// {
// add_device_gemm_xdl_c_shuffle_i8_i8_i8_mk_kn_mn_instances(op_ptrs);
// add_device_gemm_dl_i8_i8_i8_mk_kn_mn_instances(op_ptrs);
// }
// else if constexpr(is_same_v<ALayout, Row> && is_same_v<BLayout, Col> &&
// is_same_v<CLayout, Row>)
// {
// add_device_gemm_xdl_c_shuffle_i8_i8_i8_mk_nk_mn_instances(op_ptrs);
// add_device_gemm_dl_i8_i8_i8_mk_nk_mn_instances(op_ptrs);
// }
// else if constexpr(is_same_v<ALayout, Col> && is_same_v<BLayout, Row> &&
// is_same_v<CLayout, Row>)
// {
// add_device_gemm_xdl_c_shuffle_i8_i8_i8_km_kn_mn_instances(op_ptrs);
// add_device_gemm_dl_i8_i8_i8_km_kn_mn_instances(op_ptrs);
// }
// else if constexpr(is_same_v<ALayout, Col> && is_same_v<BLayout, Col> &&
// is_same_v<CLayout, Row>)
// {
// add_device_gemm_xdl_c_shuffle_i8_i8_i8_km_nk_mn_instances(op_ptrs);
// add_device_gemm_dl_i8_i8_i8_km_nk_mn_instances(op_ptrs);
// }
// }
return op_ptrs;
}
......
library/src/tensor_operation_instance/gpu/gemm/CMakeLists.txt
View file @ 82a7b60d
add_instance_library(device_gemm_instance
device_gemm_xdl_f64_f64_f64_mk_kn_mn_instance.cpp
device_gemm_xdl_f64_f64_f64_mk_nk_mn_instance.cpp
device_gemm_xdl_f64_f64_f64_km_kn_mn_instance.cpp
device_gemm_xdl_f64_f64_f64_km_nk_mn_instance.cpp
device_gemm_xdl_f32_f32_f32_mk_kn_mn_instance.cpp
device_gemm_xdl_f32_f32_f32_mk_nk_mn_instance.cpp
device_gemm_xdl_f32_f32_f32_km_kn_mn_instance.cpp
device_gemm_xdl_f32_f32_f32_km_nk_mn_instance.cpp
device_gemm_xdl_f16_f16_f16_mk_kn_mn_instance.cpp
# device_gemm_xdl_f64_f64_f64_mk_kn_mn_instance.cpp
# device_gemm_xdl_f64_f64_f64_mk_nk_mn_instance.cpp
# device_gemm_xdl_f64_f64_f64_km_kn_mn_instance.cpp
# device_gemm_xdl_f64_f64_f64_km_nk_mn_instance.cpp
# device_gemm_xdl_f32_f32_f32_mk_kn_mn_instance.cpp
# device_gemm_xdl_f32_f32_f32_mk_nk_mn_instance.cpp
# device_gemm_xdl_f32_f32_f32_km_kn_mn_instance.cpp
# device_gemm_xdl_f32_f32_f32_km_nk_mn_instance.cpp
# device_gemm_xdl_f16_f16_f16_mk_kn_mn_instance.cpp
device_gemm_xdl_f16_f16_f16_mk_nk_mn_instance.cpp
device_gemm_xdl_f16_f16_f16_km_kn_mn_instance.cpp
device_gemm_xdl_f16_f16_f16_km_nk_mn_instance.cpp
device_gemm_xdl_c_shuffle_i8_i8_i8_mk_kn_mn_instance.cpp
device_gemm_xdl_c_shuffle_i8_i8_i8_mk_nk_mn_instance.cpp
device_gemm_xdl_c_shuffle_i8_i8_i8_km_kn_mn_instance.cpp
device_gemm_xdl_c_shuffle_i8_i8_i8_km_nk_mn_instance.cpp
device_gemm_xdl_c_shuffle_bf16_bf16_bf16_mk_kn_mn_instance.cpp
device_gemm_xdl_c_shuffle_bf16_bf16_bf16_mk_nk_mn_instance.cpp
device_gemm_xdl_c_shuffle_bf16_bf16_bf16_km_kn_mn_instance.cpp
device_gemm_xdl_c_shuffle_bf16_bf16_bf16_km_nk_mn_instance.cpp
device_gemm_xdl_c_shuffle_f16_f16_f16_mk_kn_mn_instance.cpp
device_gemm_xdl_c_shuffle_f16_f16_f16_mk_nk_mn_instance.cpp
device_gemm_xdl_c_shuffle_f16_f16_f16_km_kn_mn_instance.cpp
device_gemm_xdl_c_shuffle_f16_f16_f16_km_nk_mn_instance.cpp
device_gemm_xdl_c_shuffle_f32_f32_f32_mk_kn_mn_instance.cpp
device_gemm_xdl_c_shuffle_f32_f32_f32_mk_nk_mn_instance.cpp
device_gemm_xdl_c_shuffle_f32_f32_f32_km_kn_mn_instance.cpp
device_gemm_xdl_c_shuffle_f32_f32_f32_km_nk_mn_instance.cpp
device_gemm_xdl_c_shuffle_2_stage_f16_f16_f16_mk_nk_mn_instance.cpp
device_gemm_dl_f32_f32_f32_mk_kn_mn_instance.cpp
device_gemm_dl_f32_f32_f32_mk_nk_mn_instance.cpp
device_gemm_dl_f32_f32_f32_km_kn_mn_instance.cpp
device_gemm_dl_f32_f32_f32_km_nk_mn_instance.cpp
device_gemm_dl_f16_f16_f16_mk_kn_mn_instance.cpp
device_gemm_dl_f16_f16_f16_mk_nk_mn_instance.cpp
device_gemm_dl_f16_f16_f16_km_kn_mn_instance.cpp
device_gemm_dl_f16_f16_f16_km_nk_mn_instance.cpp
device_gemm_dl_i8_i8_i8_mk_kn_mn_instance.cpp
device_gemm_dl_i8_i8_i8_mk_nk_mn_instance.cpp
device_gemm_dl_i8_i8_i8_km_kn_mn_instance.cpp
device_gemm_dl_i8_i8_i8_km_nk_mn_instance.cpp
# device_gemm_xdl_f16_f16_f16_km_kn_mn_instance.cpp
# device_gemm_xdl_f16_f16_f16_km_nk_mn_instance.cpp
# device_gemm_xdl_c_shuffle_i8_i8_i8_mk_kn_mn_instance.cpp
# device_gemm_xdl_c_shuffle_i8_i8_i8_mk_nk_mn_instance.cpp
# device_gemm_xdl_c_shuffle_i8_i8_i8_km_kn_mn_instance.cpp
# device_gemm_xdl_c_shuffle_i8_i8_i8_km_nk_mn_instance.cpp
# device_gemm_xdl_c_shuffle_bf16_bf16_bf16_mk_kn_mn_instance.cpp
# device_gemm_xdl_c_shuffle_bf16_bf16_bf16_mk_nk_mn_instance.cpp
# device_gemm_xdl_c_shuffle_bf16_bf16_bf16_km_kn_mn_instance.cpp
# device_gemm_xdl_c_shuffle_bf16_bf16_bf16_km_nk_mn_instance.cpp
# device_gemm_xdl_c_shuffle_f16_f16_f16_mk_kn_mn_instance.cpp
## device_gemm_xdl_c_shuffle_f16_f16_f16_mk_nk_mn_instance.cpp
# device_gemm_xdl_c_shuffle_f16_f16_f16_km_kn_mn_instance.cpp
# device_gemm_xdl_c_shuffle_f16_f16_f16_km_nk_mn_instance.cpp
# device_gemm_xdl_c_shuffle_f32_f32_f32_mk_kn_mn_instance.cpp
# device_gemm_xdl_c_shuffle_f32_f32_f32_mk_nk_mn_instance.cpp
# device_gemm_xdl_c_shuffle_f32_f32_f32_km_kn_mn_instance.cpp
# device_gemm_xdl_c_shuffle_f32_f32_f32_km_nk_mn_instance.cpp
## device_gemm_xdl_c_shuffle_2_stage_f16_f16_f16_mk_nk_mn_instance.cpp
# device_gemm_dl_f32_f32_f32_mk_kn_mn_instance.cpp
# device_gemm_dl_f32_f32_f32_mk_nk_mn_instance.cpp
# device_gemm_dl_f32_f32_f32_km_kn_mn_instance.cpp
# device_gemm_dl_f32_f32_f32_km_nk_mn_instance.cpp
# device_gemm_dl_f16_f16_f16_mk_kn_mn_instance.cpp
## device_gemm_dl_f16_f16_f16_mk_nk_mn_instance.cpp
# device_gemm_dl_f16_f16_f16_km_kn_mn_instance.cpp
# device_gemm_dl_f16_f16_f16_km_nk_mn_instance.cpp
# device_gemm_dl_i8_i8_i8_mk_kn_mn_instance.cpp
# device_gemm_dl_i8_i8_i8_mk_nk_mn_instance.cpp
# device_gemm_dl_i8_i8_i8_km_kn_mn_instance.cpp
# device_gemm_dl_i8_i8_i8_km_nk_mn_instance.cpp
)
set(ENABLE_IGLP_OPT ON)
if(ENABLE_IGLP_OPT)
set_source_files_properties(device_gemm_xdl_f16_f16_f16_mk_nk_mn_instance.cpp PROPERTIES
COMPILE_OPTIONS ";--save-temps;-Wno-gnu-line-marker;-mllvm;-amdgpu-enable-max-ilp-scheduling-strategy;"
COMPILE_DEFINITIONS ";USE_IGLP_OPT;USE_WAVES_PER_EU;")
endif(ENABLE_IGLP_OPT)
library/src/tensor_operation_instance/gpu/gemm/device_gemm_xdl_f16_f16_f16_mk_nk_mn_instance.cpp
View file @ 82a7b60d
......@@ -26,7 +26,7 @@ using S = ck::Sequence<Is...>;
using PassThrough = ck::tensor_operation::element_wise::PassThrough;
static constexpr auto GemmDefault = ck::tensor_operation::device::GemmSpecialization::Default;
static constexpr auto GemmMNPadding = ck::tensor_operation::device::GemmSpecialization::MNPadding;
// static constexpr auto GemmMNPadding = ck::tensor_operation::device::GemmSpecialization::MNPadding;
// Compilation parameters for a[m, k] * b[n, k] = c[m, n]
using device_gemm_xdl_f16_f16_f16_mk_nk_mn_instances = std::tuple<
......@@ -36,19 +36,32 @@ using device_gemm_xdl_f16_f16_f16_mk_nk_mn_instances = std::tuple<
//###########| | | | | | | | Operation| Operation| Operation| | | | | | | | | Wave| Wave| Lengths_K0_M_K1| ArrangeOrder| | | PerVector| PerVector_K1| | Lengths_K0_N_K1| ArrangeOrder| | | PerVector| PerVector_K1| | | PerVector| | | |
//###########| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | |
// pipeline v1, 1 wave
DeviceGemmXdl< F16, F16, F16, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, GemmDefault, 256, 256, 128, 4, 8, 32, 32, 4, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 7, 1, 1, LoopScheduler::Default, PipelineVersion::v1>,
DeviceGemmXdl< F16, F16, F16, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, GemmDefault, 256, 128, 256, 4, 8, 32, 32, 2, 4, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 7, 1, 1, LoopScheduler::Default, PipelineVersion::v1>,
DeviceGemmXdl< F16, F16, F16, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, GemmDefault, 128, 128, 128, 4, 8, 32, 32, 4, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 7, 1, 1, LoopScheduler::Default, PipelineVersion::v1>,
DeviceGemmXdl< F16, F16, F16, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, GemmDefault, 256, 128, 128, 4, 8, 32, 32, 2, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 7, 1, 1, LoopScheduler::Default, PipelineVersion::v1>,
DeviceGemmXdl< F16, F16, F16, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, GemmDefault, 128, 128, 64, 4, 8, 32, 32, 2, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 7, 1, 1, LoopScheduler::Default, PipelineVersion::v1>,
DeviceGemmXdl< F16, F16, F16, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, GemmDefault, 128, 64, 128, 4, 8, 32, 32, 2, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 7, 1, 1, LoopScheduler::Default, PipelineVersion::v1>,
DeviceGemmXdl< F16, F16, F16, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, GemmDefault, 64, 64, 64, 4, 8, 32, 32, 2, 2, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 7, 1, 1, LoopScheduler::Default, PipelineVersion::v1>,
DeviceGemmXdl< F16, F16, F16, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, GemmDefault, 256, 128, 64, 4, 8, 32, 32, 2, 1, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 7, 1, 1, LoopScheduler::Default, PipelineVersion::v1>,
DeviceGemmXdl< F16, F16, F16, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, GemmDefault, 256, 64, 128, 4, 8, 32, 32, 1, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 7, 1, 1, LoopScheduler::Default, PipelineVersion::v1>,
DeviceGemmXdl< F16, F16, F16, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, GemmDefault, 128, 128, 32, 4, 8, 32, 32, 2, 1, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 7, 1, 1, LoopScheduler::Default, PipelineVersion::v1>,
DeviceGemmXdl< F16, F16, F16, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, GemmDefault, 128, 32, 128, 4, 8, 32, 32, 1, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 7, 1, 1, LoopScheduler::Default, PipelineVersion::v1>,
DeviceGemmXdl< F16, F16, F16, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, GemmDefault, 64, 64, 32, 4, 8, 32, 32, 2, 1, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 7, 1, 1, LoopScheduler::Default, PipelineVersion::v1>,
DeviceGemmXdl< F16, F16, F16, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, GemmDefault, 64, 32, 64, 4, 8, 32, 32, 1, 2, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 7, 1, 1, LoopScheduler::Default, PipelineVersion::v1>
// DeviceGemmXdl< F16, F16, F16, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, GemmDefault, 256, 256, 128, 4, 8, 32, 32, 4, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 7, 1, 1, LoopScheduler::Default, PipelineVersion::v1>,
// DeviceGemmXdl< F16, F16, F16, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, GemmDefault, 256, 128, 256, 4, 8, 32, 32, 2, 4, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 7, 1, 1, LoopScheduler::Default, PipelineVersion::v1>,
// DeviceGemmXdl< F16, F16, F16, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, GemmDefault, 128, 128, 128, 4, 8, 32, 32, 4, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 7, 1, 1, LoopScheduler::Default, PipelineVersion::v1>,
// DeviceGemmXdl< F16, F16, F16, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, GemmDefault, 256, 128, 128, 4, 8, 32, 32, 2, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 7, 1, 1, LoopScheduler::Default, PipelineVersion::v1>,
// DeviceGemmXdl< F16, F16, F16, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, GemmDefault, 128, 128, 64, 4, 8, 32, 32, 2, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 7, 1, 1, LoopScheduler::Default, PipelineVersion::v1>,
// DeviceGemmXdl< F16, F16, F16, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, GemmDefault, 128, 64, 128, 4, 8, 32, 32, 2, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 7, 1, 1, LoopScheduler::Default, PipelineVersion::v1>,
// DeviceGemmXdl< F16, F16, F16, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, GemmDefault, 64, 64, 64, 4, 8, 32, 32, 2, 2, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 7, 1, 1, LoopScheduler::Default, PipelineVersion::v1>,
// DeviceGemmXdl< F16, F16, F16, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, GemmDefault, 256, 128, 64, 4, 8, 32, 32, 2, 1, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 7, 1, 1, LoopScheduler::Default, PipelineVersion::v1>,
// DeviceGemmXdl< F16, F16, F16, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, GemmDefault, 256, 64, 128, 4, 8, 32, 32, 1, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 7, 1, 1, LoopScheduler::Default, PipelineVersion::v1>,
// DeviceGemmXdl< F16, F16, F16, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, GemmDefault, 128, 128, 32, 4, 8, 32, 32, 2, 1, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 7, 1, 1, LoopScheduler::Default, PipelineVersion::v1>,
// DeviceGemmXdl< F16, F16, F16, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, GemmDefault, 128, 32, 128, 4, 8, 32, 32, 1, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 7, 1, 1, LoopScheduler::Default, PipelineVersion::v1>,
// DeviceGemmXdl< F16, F16, F16, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, GemmDefault, 64, 64, 32, 4, 8, 32, 32, 2, 1, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 7, 1, 1, LoopScheduler::Default, PipelineVersion::v1>,
// DeviceGemmXdl< F16, F16, F16, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, GemmDefault, 64, 32, 64, 4, 8, 32, 32, 1, 2, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 7, 1, 1, LoopScheduler::Default, PipelineVersion::v1>,
// DeviceGemmXdl< F16, F16, F16, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, GemmDefault, 256, 256, 128, 8, 8, 32, 32, 4, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 7, 1, 1, LoopScheduler::Default, PipelineVersion::v1>,
// DeviceGemmXdl< F16, F16, F16, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, GemmDefault, 256, 128, 256, 8, 8, 32, 32, 2, 4, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 7, 1, 1, LoopScheduler::Default, PipelineVersion::v1>,
// DeviceGemmXdl< F16, F16, F16, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, GemmDefault, 128, 128, 128, 8, 8, 32, 32, 4, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 7, 1, 1, LoopScheduler::Default, PipelineVersion::v1>,
// DeviceGemmXdl< F16, F16, F16, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, GemmDefault, 256, 128, 128, 8, 8, 32, 32, 2, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 7, 1, 1, LoopScheduler::Default, PipelineVersion::v1>,
// DeviceGemmXdl< F16, F16, F16, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, GemmDefault, 128, 128, 64, 8, 8, 32, 32, 2, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 7, 1, 1, LoopScheduler::Default, PipelineVersion::v1>,
// DeviceGemmXdl< F16, F16, F16, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, GemmDefault, 128, 64, 128, 8, 8, 32, 32, 2, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 7, 1, 1, LoopScheduler::Default, PipelineVersion::v1>,
// DeviceGemmXdl< F16, F16, F16, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, GemmDefault, 64, 64, 64, 8, 8, 32, 32, 2, 2, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 7, 1, 1, LoopScheduler::Default, PipelineVersion::v1>,
// DeviceGemmXdl< F16, F16, F16, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, GemmDefault, 256, 128, 64, 8, 8, 32, 32, 2, 1, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 7, 1, 1, LoopScheduler::Default, PipelineVersion::v1>,
// DeviceGemmXdl< F16, F16, F16, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, GemmDefault, 256, 64, 128, 8, 8, 32, 32, 1, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 7, 1, 1, LoopScheduler::Default, PipelineVersion::v1>,
// DeviceGemmXdl< F16, F16, F16, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, GemmDefault, 128, 128, 32, 8, 8, 32, 32, 2, 1, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 7, 1, 1, LoopScheduler::Default, PipelineVersion::v1>,
// DeviceGemmXdl< F16, F16, F16, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, GemmDefault, 128, 32, 128, 8, 8, 32, 32, 1, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 7, 1, 1, LoopScheduler::Default, PipelineVersion::v1>,
// DeviceGemmXdl< F16, F16, F16, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, GemmDefault, 64, 64, 32, 8, 8, 32, 32, 2, 1, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 7, 1, 1, LoopScheduler::Default, PipelineVersion::v1>,
// DeviceGemmXdl< F16, F16, F16, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, GemmDefault, 64, 32, 64, 8, 8, 32, 32, 1, 2, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 7, 1, 1, LoopScheduler::Default, PipelineVersion::v1>
#if CK_EXPERIMENTAL_INTER_WAVE_INSTANCES
// pipeline v1, 2 waves
,
......@@ -68,60 +81,75 @@ using device_gemm_xdl_f16_f16_f16_mk_nk_mn_instances = std::tuple<
#endif
#if CK_EXPERIMENTAL_PIPELINE_V2_INSTANCES
// pipeline v2, 1 wave
,
DeviceGemmXdl< F16, F16, F16, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, GemmDefault, 256, 256, 128, 4, 8, 32, 32, 4, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 7, 1, 1, LoopScheduler::Default, PipelineVersion::v2>,
DeviceGemmXdl< F16, F16, F16, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, GemmDefault, 256, 128, 256, 4, 8, 32, 32, 2, 4, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 7, 1, 1, LoopScheduler::Default, PipelineVersion::v2>,
DeviceGemmXdl< F16, F16, F16, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, GemmDefault, 128, 128, 128, 4, 8, 32, 32, 4, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 7, 1, 1, LoopScheduler::Default, PipelineVersion::v2>,
DeviceGemmXdl< F16, F16, F16, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, GemmDefault, 256, 128, 128, 4, 8, 32, 32, 2, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 7, 1, 1, LoopScheduler::Default, PipelineVersion::v2>,
DeviceGemmXdl< F16, F16, F16, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, GemmDefault, 128, 128, 64, 4, 8, 32, 32, 2, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 7, 1, 1, LoopScheduler::Default, PipelineVersion::v2>,
DeviceGemmXdl< F16, F16, F16, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, GemmDefault, 128, 64, 128, 4, 8, 32, 32, 2, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 7, 1, 1, LoopScheduler::Default, PipelineVersion::v2>,
DeviceGemmXdl< F16, F16, F16, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, GemmDefault, 64, 64, 64, 4, 8, 32, 32, 2, 2, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 7, 1, 1, LoopScheduler::Default, PipelineVersion::v2>,
DeviceGemmXdl< F16, F16, F16, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, GemmDefault, 256, 128, 64, 4, 8, 32, 32, 2, 1, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 7, 1, 1, LoopScheduler::Default, PipelineVersion::v2>,
// ,
// DeviceGemmXdl< F16, F16, F16, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, GemmDefault, 256, 256, 128, 4, 8, 32, 32, 4, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 7, 1, 1, LoopScheduler::Default, PipelineVersion::v2>,
// DeviceGemmXdl< F16, F16, F16, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, GemmDefault, 256, 128, 256, 4, 8, 32, 32, 2, 4, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 7, 1, 1, LoopScheduler::Default, PipelineVersion::v2>,
// DeviceGemmXdl< F16, F16, F16, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, GemmDefault, 128, 128, 128, 4, 8, 32, 32, 4, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 7, 1, 1, LoopScheduler::Default, PipelineVersion::v2>,
// DeviceGemmXdl< F16, F16, F16, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, GemmDefault, 256, 128, 128, 4, 8, 32, 32, 2, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 7, 1, 1, LoopScheduler::Default, PipelineVersion::v2>,
// DeviceGemmXdl< F16, F16, F16, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, GemmDefault, 128, 128, 64, 4, 8, 32, 32, 2, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 7, 1, 1, LoopScheduler::Default, PipelineVersion::v2>,
// DeviceGemmXdl< F16, F16, F16, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, GemmDefault, 128, 64, 128, 4, 8, 32, 32, 2, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 7, 1, 1, LoopScheduler::Default, PipelineVersion::v2>,
// DeviceGemmXdl< F16, F16, F16, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, GemmDefault, 64, 64, 64, 4, 8, 32, 32, 2, 2, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 7, 1, 1, LoopScheduler::Default, PipelineVersion::v2>,
// DeviceGemmXdl< F16, F16, F16, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, GemmDefault, 256, 128, 64, 4, 8, 32, 32, 2, 1, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 7, 1, 1, LoopScheduler::Default, PipelineVersion::v2>,
DeviceGemmXdl< F16, F16, F16, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, GemmDefault, 256, 64, 128, 4, 8, 32, 32, 1, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 7, 1, 1, LoopScheduler::Default, PipelineVersion::v2>,
DeviceGemmXdl< F16, F16, F16, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, GemmDefault, 128, 128, 32, 4, 8, 32, 32, 2, 1, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 7, 1, 1, LoopScheduler::Default, PipelineVersion::v2>,
DeviceGemmXdl< F16, F16, F16, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, GemmDefault, 128, 32, 128, 4, 8, 32, 32, 1, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 7, 1, 1, LoopScheduler::Default, PipelineVersion::v2>,
DeviceGemmXdl< F16, F16, F16, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, GemmDefault, 64, 64, 32, 4, 8, 32, 32, 2, 1, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 7, 1, 1, LoopScheduler::Default, PipelineVersion::v2>,
DeviceGemmXdl< F16, F16, F16, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, GemmDefault, 64, 32, 64, 4, 8, 32, 32, 1, 2, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 7, 1, 1, LoopScheduler::Default, PipelineVersion::v2>
#endif
// clang-format on
>;
// DeviceGemmXdl< F16, F16, F16, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, GemmDefault, 128, 128, 32, 4, 8, 32, 32, 2, 1, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 7, 1, 1, LoopScheduler::Default, PipelineVersion::v2>,
// DeviceGemmXdl< F16, F16, F16, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, GemmDefault, 128, 32, 128, 4, 8, 32, 32, 1, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 7, 1, 1, LoopScheduler::Default, PipelineVersion::v2>,
// DeviceGemmXdl< F16, F16, F16, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, GemmDefault, 64, 64, 32, 4, 8, 32, 32, 2, 1, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 7, 1, 1, LoopScheduler::Default, PipelineVersion::v2>,
// DeviceGemmXdl< F16, F16, F16, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, GemmDefault, 64, 32, 64, 4, 8, 32, 32, 1, 2, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 7, 1, 1, LoopScheduler::Default, PipelineVersion::v2>,
// DeviceGemmXdl< F16, F16, F16, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, GemmDefault, 256, 256, 128, 8, 8, 32, 32, 4, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 7, 1, 1, LoopScheduler::Default, PipelineVersion::v2>,
// DeviceGemmXdl< F16, F16, F16, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, GemmDefault, 256, 128, 256, 8, 8, 32, 32, 2, 4, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 7, 1, 1, LoopScheduler::Default, PipelineVersion::v2>,
// DeviceGemmXdl< F16, F16, F16, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, GemmDefault, 128, 128, 128, 8, 8, 32, 32, 4, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 7, 1, 1, LoopScheduler::Default, PipelineVersion::v2>,
// DeviceGemmXdl< F16, F16, F16, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, GemmDefault, 256, 128, 128, 8, 8, 32, 32, 2, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 7, 1, 1, LoopScheduler::Default, PipelineVersion::v2>,
// DeviceGemmXdl< F16, F16, F16, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, GemmDefault, 128, 128, 64, 8, 8, 32, 32, 2, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 7, 1, 1, LoopScheduler::Default, PipelineVersion::v2>,
// DeviceGemmXdl< F16, F16, F16, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, GemmDefault, 128, 64, 128, 8, 8, 32, 32, 2, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 7, 1, 1, LoopScheduler::Default, PipelineVersion::v2>,
// DeviceGemmXdl< F16, F16, F16, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, GemmDefault, 64, 64, 64, 8, 8, 32, 32, 2, 2, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 7, 1, 1, LoopScheduler::Default, PipelineVersion::v2>,
// DeviceGemmXdl< F16, F16, F16, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, GemmDefault, 256, 128, 64, 8, 8, 32, 32, 2, 1, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 7, 1, 1, LoopScheduler::Default, PipelineVersion::v2>,
DeviceGemmXdl< F16, F16, F16, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, GemmDefault, 256, 64, 128, 8, 8, 32, 32, 1, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 7, 1, 1, LoopScheduler::Default, PipelineVersion::v2>
// DeviceGemmXdl< F16, F16, F16, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, GemmDefault, 128, 128, 32, 8, 8, 32, 32, 2, 1, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 7, 1, 1, LoopScheduler::Default, PipelineVersion::v2>,
// DeviceGemmXdl< F16, F16, F16, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, GemmDefault, 128, 32, 128, 8, 8, 32, 32, 1, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 7, 1, 1, LoopScheduler::Default, PipelineVersion::v2>,
// DeviceGemmXdl< F16, F16, F16, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, GemmDefault, 64, 64, 32, 8, 8, 32, 32, 2, 1, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 7, 1, 1, LoopScheduler::Default, PipelineVersion::v2>,
// DeviceGemmXdl< F16, F16, F16, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, GemmDefault, 64, 32, 64, 8, 8, 32, 32, 1, 2, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 7, 1, 1, LoopScheduler::Default, PipelineVersion::v2>
// irregular tile size
using device_gemm_xdl_f16_f16_f16_mk_nk_mn_irregular_tile_instances = std::tuple<
// clang-format off
//###########| AData| BData| CData| AccData| ALayout| BLayout| CLayout| A| B| C| GEMM| Block| MPer| NPer| K0Per| K1| MPer| NPer| MXdl| NXdl| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds| BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds| CThreadTransfer| CThreadTransfer| NumPrefetch| LoopScheduler| Pipeline|
//###########| Type| Type| Type| Type| | | | Elementwise| Elementwise| Elementwise|Specialization| Size| Block| Block| Block| | XDL| XDL| Per| Per| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| AddExtraM| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| AddExtraN| SrcDstVectorDim| DstScalar| | | |
//###########| | | | | | | | Operation| Operation| Operation| | | | | | | | | Wave| Wave| Lengths_K0_M_K1| ArrangeOrder| | | PerVector| PerVector_K1| | Lengths_K0_N_K1| ArrangeOrder| | | PerVector| PerVector_K1| | | PerVector| | | |
//###########| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | |
// pipeline v1, 1 wave
DeviceGemmXdl< F16, F16, F16, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, GemmMNPadding, 64, 64, 32, 4, 8, 32, 32, 2, 1, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 7, 1, 1, LoopScheduler::Default, PipelineVersion::v1>,
DeviceGemmXdl< F16, F16, F16, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, GemmMNPadding, 256, 128, 144, 8, 8, 16, 16, 2, 9, S<8, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<8, 8, 4>, S<1, 0, 2>, S<1, 0, 2>, 2, 2, 2, true, 7, 1, 1, LoopScheduler::Default, PipelineVersion::v1>,
DeviceGemmXdl< F16, F16, F16, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, GemmMNPadding, 256, 128, 144, 4, 8, 16, 16, 2, 9, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 16, 4>, S<1, 0, 2>, S<1, 0, 2>, 2, 2, 2, true, 7, 1, 1, LoopScheduler::Default, PipelineVersion::v1>
#if CK_EXPERIMENTAL_INTER_WAVE_INSTANCES
// pipeline v1, 2 waves
,
DeviceGemmXdl< F16, F16, F16, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, GemmMNPadding, 64, 64, 32, 4, 8, 32, 32, 2, 1, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 7, 1, 1, LoopScheduler::Interwave, PipelineVersion::v1>,
DeviceGemmXdl< F16, F16, F16, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, GemmMNPadding, 256, 128, 144, 8, 8, 16, 16, 2, 9, S<8, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<8, 8, 4>, S<1, 0, 2>, S<1, 0, 2>, 2, 2, 2, true, 7, 1, 1, LoopScheduler::Interwave, PipelineVersion::v1>,
DeviceGemmXdl< F16, F16, F16, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, GemmMNPadding, 256, 128, 144, 4, 8, 16, 16, 2, 9, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 16, 4>, S<1, 0, 2>, S<1, 0, 2>, 2, 2, 2, true, 7, 1, 1, LoopScheduler::Interwave, PipelineVersion::v1>
#endif
#if CK_EXPERIMENTAL_PIPELINE_V2_INSTANCES
// pipeline v2, 1 wave
,
DeviceGemmXdl< F16, F16, F16, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, GemmMNPadding, 64, 64, 32, 4, 8, 32, 32, 2, 1, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 7, 1, 1, LoopScheduler::Default, PipelineVersion::v2>,
DeviceGemmXdl< F16, F16, F16, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, GemmMNPadding, 256, 128, 144, 8, 8, 16, 16, 2, 9, S<8, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<8, 8, 4>, S<1, 0, 2>, S<1, 0, 2>, 2, 2, 2, true, 7, 1, 1, LoopScheduler::Default, PipelineVersion::v2>,
DeviceGemmXdl< F16, F16, F16, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, GemmMNPadding, 256, 128, 144, 4, 8, 16, 16, 2, 9, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 16, 4>, S<1, 0, 2>, S<1, 0, 2>, 2, 2, 2, true, 7, 1, 1, LoopScheduler::Default, PipelineVersion::v2>
#endif
// clang-format on
>;
// // irregular tile size
// using device_gemm_xdl_f16_f16_f16_mk_nk_mn_irregular_tile_instances = std::tuple<
// // clang-format off
// //###########| AData| BData| CData| AccData| ALayout| BLayout| CLayout| A| B| C| GEMM| Block| MPer| NPer| K0Per| K1| MPer| NPer| MXdl| NXdl| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds| BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds| CThreadTransfer| CThreadTransfer| NumPrefetch| LoopScheduler| Pipeline|
// //###########| Type| Type| Type| Type| | | | Elementwise| Elementwise| Elementwise|Specialization| Size| Block| Block| Block| | XDL| XDL| Per| Per| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| AddExtraM| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| AddExtraN| SrcDstVectorDim| DstScalar| | | |
// //###########| | | | | | | | Operation| Operation| Operation| | | | | | | | | Wave| Wave| Lengths_K0_M_K1| ArrangeOrder| | | PerVector| PerVector_K1| | Lengths_K0_N_K1| ArrangeOrder| | | PerVector| PerVector_K1| | | PerVector| | | |
// //###########| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | |
// // pipeline v1, 1 wave
// DeviceGemmXdl< F16, F16, F16, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, GemmMNPadding, 64, 64, 32, 4, 8, 32, 32, 2, 1, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 7, 1, 1, LoopScheduler::Default, PipelineVersion::v1>,
// DeviceGemmXdl< F16, F16, F16, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, GemmMNPadding, 256, 128, 144, 8, 8, 16, 16, 2, 9, S<8, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<8, 8, 4>, S<1, 0, 2>, S<1, 0, 2>, 2, 2, 2, true, 7, 1, 1, LoopScheduler::Default, PipelineVersion::v1>,
// DeviceGemmXdl< F16, F16, F16, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, GemmMNPadding, 256, 128, 144, 4, 8, 16, 16, 2, 9, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 16, 4>, S<1, 0, 2>, S<1, 0, 2>, 2, 2, 2, true, 7, 1, 1, LoopScheduler::Default, PipelineVersion::v1>
// #if CK_EXPERIMENTAL_INTER_WAVE_INSTANCES
// // pipeline v1, 2 waves
// ,
// DeviceGemmXdl< F16, F16, F16, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, GemmMNPadding, 64, 64, 32, 4, 8, 32, 32, 2, 1, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 7, 1, 1, LoopScheduler::Interwave, PipelineVersion::v1>,
// DeviceGemmXdl< F16, F16, F16, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, GemmMNPadding, 256, 128, 144, 8, 8, 16, 16, 2, 9, S<8, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<8, 8, 4>, S<1, 0, 2>, S<1, 0, 2>, 2, 2, 2, true, 7, 1, 1, LoopScheduler::Interwave, PipelineVersion::v1>,
// DeviceGemmXdl< F16, F16, F16, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, GemmMNPadding, 256, 128, 144, 4, 8, 16, 16, 2, 9, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 16, 4>, S<1, 0, 2>, S<1, 0, 2>, 2, 2, 2, true, 7, 1, 1, LoopScheduler::Interwave, PipelineVersion::v1>
// #endif
// #if CK_EXPERIMENTAL_PIPELINE_V2_INSTANCES
// // pipeline v2, 1 wave
// ,
// DeviceGemmXdl< F16, F16, F16, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, GemmMNPadding, 64, 64, 32, 4, 8, 32, 32, 2, 1, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 7, 1, 1, LoopScheduler::Default, PipelineVersion::v2>,
// DeviceGemmXdl< F16, F16, F16, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, GemmMNPadding, 256, 128, 144, 8, 8, 16, 16, 2, 9, S<8, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<8, 8, 4>, S<1, 0, 2>, S<1, 0, 2>, 2, 2, 2, true, 7, 1, 1, LoopScheduler::Default, PipelineVersion::v2>,
// DeviceGemmXdl< F16, F16, F16, F32, Row, Col, Row, PassThrough, PassThrough, PassThrough, GemmMNPadding, 256, 128, 144, 4, 8, 16, 16, 2, 9, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 16, 4>, S<1, 0, 2>, S<1, 0, 2>, 2, 2, 2, true, 7, 1, 1, LoopScheduler::Default, PipelineVersion::v2>
// #endif
// // clang-format on
// >;
void add_device_gemm_xdl_f16_f16_f16_mk_nk_mn_instances(
std::vector<std::unique_ptr<
DeviceGemm<Row, Col, Row, F16, F16, F16, PassThrough, PassThrough, PassThrough>>>&
instances)
{
add_device_operation_instances(instances, device_gemm_xdl_f16_f16_f16_mk_nk_mn_instances{});
add_device_operation_instances(instances,
device_gemm_xdl_f16_f16_f16_mk_nk_mn_irregular_tile_instances{});
// add_device_operation_instances(instances,
// device_gemm_xdl_f16_f16_f16_mk_nk_mn_irregular_tile_instances{});
}
} // namespace instance
......
profiler/src/CMakeLists.txt
View file @ 82a7b60d
......@@ -2,30 +2,30 @@
set(PROFILER_SOURCES
profiler.cpp
profile_gemm.cpp
profile_gemm_splitk.cpp
profile_gemm_bilinear.cpp
profile_gemm_bias_add_reduce.cpp
profile_gemm_add_add_fastgelu.cpp
profile_gemm_add_fastgelu.cpp
profile_gemm_fastgelu.cpp
profile_gemm_reduce.cpp
profile_batched_gemm.cpp
profile_batched_gemm_gemm.cpp
profile_batched_gemm_add_relu_gemm_add.cpp
profile_batched_gemm_reduce.cpp
profile_grouped_gemm.cpp
profile_conv_fwd.cpp
profile_conv_fwd_bias_relu.cpp
profile_conv_fwd_bias_relu_add.cpp
profile_conv_bwd_data.cpp
profile_grouped_conv_fwd.cpp
profile_grouped_conv_bwd_weight.cpp
profile_reduce.cpp
profile_groupnorm.cpp
profile_layernorm.cpp
profile_softmax.cpp
profile_batchnorm_fwd.cpp
profile_batchnorm_bwd.cpp
# profile_gemm_splitk.cpp
# profile_gemm_bilinear.cpp
# profile_gemm_bias_add_reduce.cpp
# profile_gemm_add_add_fastgelu.cpp
# profile_gemm_add_fastgelu.cpp
# profile_gemm_fastgelu.cpp
# profile_gemm_reduce.cpp
# profile_batched_gemm.cpp
# profile_batched_gemm_gemm.cpp
# profile_batched_gemm_add_relu_gemm_add.cpp
# profile_batched_gemm_reduce.cpp
# profile_grouped_gemm.cpp
# profile_conv_fwd.cpp
# profile_conv_fwd_bias_relu.cpp
# profile_conv_fwd_bias_relu_add.cpp
# profile_conv_bwd_data.cpp
# profile_grouped_conv_fwd.cpp
# profile_grouped_conv_bwd_weight.cpp
# profile_reduce.cpp
# profile_groupnorm.cpp
# profile_layernorm.cpp
# profile_softmax.cpp
# profile_batchnorm_fwd.cpp
# profile_batchnorm_bwd.cpp
)
set(PROFILER_EXECUTABLE ckProfiler)
......@@ -35,33 +35,33 @@ target_compile_options(${PROFILER_EXECUTABLE} PRIVATE -Wno-global-constructors)
target_link_libraries(${PROFILER_EXECUTABLE} PRIVATE utility)
target_link_libraries(${PROFILER_EXECUTABLE} PRIVATE device_gemm_instance)
target_link_libraries(${PROFILER_EXECUTABLE} PRIVATE device_gemm_splitk_instance)
target_link_libraries(${PROFILER_EXECUTABLE} PRIVATE device_gemm_bilinear_instance)
target_link_libraries(${PROFILER_EXECUTABLE} PRIVATE device_gemm_add_add_fastgelu_instance)
target_link_libraries(${PROFILER_EXECUTABLE} PRIVATE device_gemm_add_fastgelu_instance)
target_link_libraries(${PROFILER_EXECUTABLE} PRIVATE device_gemm_fastgelu_instance)
target_link_libraries(${PROFILER_EXECUTABLE} PRIVATE device_gemm_reduce_instance)
target_link_libraries(${PROFILER_EXECUTABLE} PRIVATE device_gemm_bias_add_reduce_instance)
target_link_libraries(${PROFILER_EXECUTABLE} PRIVATE device_batched_gemm_instance)
target_link_libraries(${PROFILER_EXECUTABLE} PRIVATE device_batched_gemm_gemm_instance)
target_link_libraries(${PROFILER_EXECUTABLE} PRIVATE device_batched_gemm_add_relu_gemm_add_instance)
target_link_libraries(${PROFILER_EXECUTABLE} PRIVATE device_batched_gemm_reduce_instance)
target_link_libraries(${PROFILER_EXECUTABLE} PRIVATE device_grouped_gemm_instance)
target_link_libraries(${PROFILER_EXECUTABLE} PRIVATE device_conv2d_fwd_instance)
target_link_libraries(${PROFILER_EXECUTABLE} PRIVATE device_grouped_conv1d_fwd_instance)
target_link_libraries(${PROFILER_EXECUTABLE} PRIVATE device_grouped_conv2d_fwd_instance)
target_link_libraries(${PROFILER_EXECUTABLE} PRIVATE device_grouped_conv3d_fwd_instance)
target_link_libraries(${PROFILER_EXECUTABLE} PRIVATE device_conv1d_bwd_data_instance)
target_link_libraries(${PROFILER_EXECUTABLE} PRIVATE device_conv2d_bwd_data_instance)
target_link_libraries(${PROFILER_EXECUTABLE} PRIVATE device_conv3d_bwd_data_instance)
target_link_libraries(${PROFILER_EXECUTABLE} PRIVATE device_grouped_conv1d_bwd_weight_instance)
target_link_libraries(${PROFILER_EXECUTABLE} PRIVATE device_grouped_conv2d_bwd_weight_instance)
target_link_libraries(${PROFILER_EXECUTABLE} PRIVATE device_grouped_conv3d_bwd_weight_instance)
target_link_libraries(${PROFILER_EXECUTABLE} PRIVATE device_conv2d_fwd_bias_relu_instance)
target_link_libraries(${PROFILER_EXECUTABLE} PRIVATE device_conv2d_fwd_bias_relu_add_instance)
target_link_libraries(${PROFILER_EXECUTABLE} PRIVATE device_normalization_instance)
target_link_libraries(${PROFILER_EXECUTABLE} PRIVATE device_softmax_instance)
target_link_libraries(${PROFILER_EXECUTABLE} PRIVATE device_reduce_instance)
target_link_libraries(${PROFILER_EXECUTABLE} PRIVATE device_batchnorm_instance)
# target_link_libraries(${PROFILER_EXECUTABLE} PRIVATE device_gemm_splitk_instance)
# target_link_libraries(${PROFILER_EXECUTABLE} PRIVATE device_gemm_bilinear_instance)
# target_link_libraries(${PROFILER_EXECUTABLE} PRIVATE device_gemm_add_add_fastgelu_instance)
# target_link_libraries(${PROFILER_EXECUTABLE} PRIVATE device_gemm_add_fastgelu_instance)
# target_link_libraries(${PROFILER_EXECUTABLE} PRIVATE device_gemm_fastgelu_instance)
# target_link_libraries(${PROFILER_EXECUTABLE} PRIVATE device_gemm_reduce_instance)
# target_link_libraries(${PROFILER_EXECUTABLE} PRIVATE device_gemm_bias_add_reduce_instance)
# target_link_libraries(${PROFILER_EXECUTABLE} PRIVATE device_batched_gemm_instance)
# target_link_libraries(${PROFILER_EXECUTABLE} PRIVATE device_batched_gemm_gemm_instance)
# target_link_libraries(${PROFILER_EXECUTABLE} PRIVATE device_batched_gemm_add_relu_gemm_add_instance)
# target_link_libraries(${PROFILER_EXECUTABLE} PRIVATE device_batched_gemm_reduce_instance)
# target_link_libraries(${PROFILER_EXECUTABLE} PRIVATE device_grouped_gemm_instance)
# target_link_libraries(${PROFILER_EXECUTABLE} PRIVATE device_conv2d_fwd_instance)
# target_link_libraries(${PROFILER_EXECUTABLE} PRIVATE device_grouped_conv1d_fwd_instance)
# target_link_libraries(${PROFILER_EXECUTABLE} PRIVATE device_grouped_conv2d_fwd_instance)
# target_link_libraries(${PROFILER_EXECUTABLE} PRIVATE device_grouped_conv3d_fwd_instance)
# target_link_libraries(${PROFILER_EXECUTABLE} PRIVATE device_conv1d_bwd_data_instance)
# target_link_libraries(${PROFILER_EXECUTABLE} PRIVATE device_conv2d_bwd_data_instance)
# target_link_libraries(${PROFILER_EXECUTABLE} PRIVATE device_conv3d_bwd_data_instance)
# target_link_libraries(${PROFILER_EXECUTABLE} PRIVATE device_grouped_conv1d_bwd_weight_instance)
# target_link_libraries(${PROFILER_EXECUTABLE} PRIVATE device_grouped_conv2d_bwd_weight_instance)
# target_link_libraries(${PROFILER_EXECUTABLE} PRIVATE device_grouped_conv3d_bwd_weight_instance)
# target_link_libraries(${PROFILER_EXECUTABLE} PRIVATE device_conv2d_fwd_bias_relu_instance)
# target_link_libraries(${PROFILER_EXECUTABLE} PRIVATE device_conv2d_fwd_bias_relu_add_instance)
# target_link_libraries(${PROFILER_EXECUTABLE} PRIVATE device_normalization_instance)
# target_link_libraries(${PROFILER_EXECUTABLE} PRIVATE device_softmax_instance)
# target_link_libraries(${PROFILER_EXECUTABLE} PRIVATE device_reduce_instance)
# target_link_libraries(${PROFILER_EXECUTABLE} PRIVATE device_batchnorm_instance)
rocm_install(TARGETS ${PROFILER_EXECUTABLE} COMPONENT profiler)
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment