mirror of
https://github.com/ROCm/composable_kernel.git
synced 2026-05-03 05:01:25 +00:00
321627aec5
* Fix example * fix build error * update pk_i4 & moe test case * fix all instance build (examples) * fix batched_gemm_gemm (example) * disable example_gemm_bias_softmax_gemm_permute on gfx11 * remove unnecessary disable gfx11 * update tests * update tests2
49 lines
1.6 KiB
C++
49 lines
1.6 KiB
C++
// SPDX-License-Identifier: MIT
|
|
// Copyright (c) 2023-2025, Advanced Micro Devices, Inc. All rights reserved.
|
|
|
|
#include "common.hpp"
|
|
|
|
#include "ck/tensor_operation/gpu/device/impl/device_gemm_xdl_cshuffle_v3.hpp"
|
|
|
|
using ADataType = ck::f8_t;
|
|
using BDataType = ck::f8_t;
|
|
using AccDataType = float;
|
|
using CShuffleDataType = ck::half_t;
|
|
using CDataType = ck::half_t;
|
|
|
|
using ALayout = Row;
|
|
using BLayout = Col;
|
|
using CLayout = Row;
|
|
|
|
using AElementOp = PassThrough;
|
|
using BElementOp = PassThrough;
|
|
using CElementOp = PassThrough;
|
|
|
|
static constexpr auto GemmDefault = ck::tensor_operation::device::GemmSpecialization::Default;
|
|
|
|
// clang-format off
|
|
using DeviceGemmV2Instance =
|
|
ck::tensor_operation::device::DeviceGemm_Xdl_CShuffleV3<
|
|
ALayout, BLayout, CLayout,
|
|
ADataType, BDataType, CDataType, AccDataType, CShuffleDataType,
|
|
PassThrough, PassThrough, PassThrough, GemmDefault,
|
|
256,
|
|
224, 256,
|
|
128, 16, 16,
|
|
16, 16,
|
|
7, 8,
|
|
S<8, 32, 1>, S<1, 0, 2>, S<1, 0, 2>,
|
|
2, 16, 16, 0,
|
|
S<8, 32, 1>, S<1, 0, 2>, S<1, 0, 2>,
|
|
2, 16, 16, 0,
|
|
1, 2, S<1, 32, 1, 8>, 4,
|
|
ck::BlockGemmPipelineScheduler::Intrawave,ck::BlockGemmPipelineVersion::v3, ck::f8_t>;
|
|
// clang-format on
|
|
|
|
using ReferenceGemmInstance = ck::tensor_operation::host::
|
|
ReferenceGemm<ADataType, BDataType, CDataType, AccDataType, AElementOp, BElementOp, CElementOp>;
|
|
|
|
#include "run_gemm_example_v2.inc"
|
|
|
|
int main(int argc, char* argv[]) { return !run_gemm_splitk_example(argc, argv); }
|