mirror of
https://github.com/ROCm/composable_kernel.git
synced 2026-05-03 05:01:25 +00:00
453ca37347
* Gemm Kernel Refactor part1 * Gemm Kernel Refactor common gemm pipeline part2 * [CK TILE] Refactor batched gemm to reuse GemmKernel * [CK TILE] Refactor GemmKernel - review changes part1 * [CK TILE] Refactor GemmKernel - references fix * [CK TILE] Refactor GemmKernel - naming changes, add problem * [CK_TILE] Refactor GemmKernel - update tests * [CK_TILE] Refactor GemmKernel - review changes * [CK_TILE] Refactor GemmKernel - update test * [CK_TILE] Refactor GemmKernel - constness fixes * [CK_TILE] Refactor GemmKernel - update tests
79 lines
2.3 KiB
C++
79 lines
2.3 KiB
C++
|
|
// SPDX-License-Identifier: MIT
|
|
// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
|
|
|
|
#pragma once
|
|
|
|
#include <string>
|
|
|
|
#include "ck_tile/core.hpp"
|
|
#include "ck_tile/host/kernel_launch.hpp"
|
|
|
|
template <typename DataType>
|
|
struct GemmBasicTypeConfig;
|
|
|
|
template <>
|
|
struct GemmBasicTypeConfig<ck_tile::half_t>
|
|
{
|
|
using ADataType = ck_tile::half_t;
|
|
using BDataType = ck_tile::half_t;
|
|
using AccDataType = float;
|
|
using CDataType = ck_tile::half_t;
|
|
// ToDo: Add more bias config to support different categories of GEMM.
|
|
};
|
|
|
|
template <typename T>
|
|
struct DataTypeTraits;
|
|
|
|
template <>
|
|
struct DataTypeTraits<float>
|
|
{
|
|
static constexpr const char* name = "fp32";
|
|
};
|
|
|
|
template <>
|
|
struct DataTypeTraits<double>
|
|
{
|
|
static constexpr const char* name = "fp64";
|
|
};
|
|
|
|
template <>
|
|
struct DataTypeTraits<ck_tile::half_t>
|
|
{
|
|
static constexpr const char* name = "fp16";
|
|
};
|
|
|
|
using Types = GemmBasicTypeConfig<ck_tile::half_t>;
|
|
|
|
// Specific type aliases for easy access
|
|
using ADataType = Types::ADataType;
|
|
using BDataType = Types::BDataType;
|
|
using AccDataType = Types::AccDataType;
|
|
using CDataType = Types::CDataType;
|
|
|
|
auto create_args(int argc, char* argv[])
|
|
{
|
|
ck_tile::ArgParser arg_parser;
|
|
arg_parser.insert("b", "1", "batch size")
|
|
.insert("m", "3840", "m dimension")
|
|
.insert("n", "4096", "n dimension")
|
|
.insert("k", "2048", "k dimension")
|
|
.insert("a_layout", "R", "A tensor data layout - Row by default")
|
|
.insert("b_layout", "R", "B tensor data layout - Row by default")
|
|
.insert("c_layout", "R", "C tensor data layout - Row by default")
|
|
.insert("stride_a", "0", "Tensor A stride")
|
|
.insert("stride_b", "0", "Tensor B stride")
|
|
.insert("stride_c", "0", "Tensor C stride")
|
|
.insert("v", "2", "0. No validation, 1. Validation on CPU, 2. Validation on GPU")
|
|
.insert("prec", "fp16", "data type. fp16/bf16/fp8/bf8")
|
|
.insert("warmup", "50", "number of iterations before benchmark the kernel")
|
|
.insert("repeat", "100", "number of iterations to benchmark the kernel")
|
|
.insert("timer", "gpu", "gpu:gpu timer, cpu:cpu timer");
|
|
|
|
bool result = arg_parser.parse(argc, argv);
|
|
return std::make_tuple(result, arg_parser);
|
|
}
|
|
|
|
// host API
|
|
float gemm_calc(const ck_tile::GemmHostArgs& args, const ck_tile::stream_config& s);
|