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

initial stream-k implementation with example (#699)

Browse Source 
* initial stream-k implementation with example

* fix unexpected change in err

* improve a little bit performance by reorganize pipeline.

* improve perf a little bit by swizzle block idx

* add profiler

* update example

* fix spelling

* shrink karg for streamk

* support dynamic buffer using memory coherence glc_slc bit from template

* control memory coherence while construct dynamic buffer

* update reduction for streamk(not ready yet)

* Add template parameter to make_dynamic_buffer to support amd_buffer coherence setting

* fix build issue

* fix several bug

* now result is correct, everything works (but has scratch)

* remove scratch by manually reset coordinate

* update device code

* fix a bug in final reduce

* fix something in example

* update async memset

* fix enum as camel case

* modify coherence enum name

* clean code and use atomic streamk by default

* remove unused var

* throw exception if have empty pointer

* fix format

* fix CI warning

* fix type in init

* modify CI error

* filter out on gfx10+

* restore changed example code

---------

Co-authored-by: Qianfeng Zhang <Qianfeng.Zhang@amd.com>
This commit is contained in:
carlushuang
2023-07-27 03:18:15 +08:00
committed by GitHub
parent 9195435c77
commit e7dca79d27
28 changed files with 4234 additions and 36 deletions
Download Patch File Download Diff File Expand all files Collapse all files

121
library/include/ck/library/tensor_operation_instance/gpu/gemm_streamk.hpp Normal file
View File

@@ -0,0 +1,121 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
#pragma once
#include <cstdlib>
#include "ck/ck.hpp"
#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
#include "ck/tensor_operation/gpu/device/device_gemm_streamk.hpp"
#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
#include "ck/library/tensor_operation_instance/device_operation_instance_factory.hpp"
namespace ck {
namespace tensor_operation {
namespace device {
namespace instance {
void add_device_gemm_xdl_streamk_f16_f16_f16_mk_kn_mn_instances(
std::vector<std::unique_ptr<
DeviceGemmStreamK<Row, Row, Row, F16, F16, F16, PassThrough, PassThrough, PassThrough>>>&
instances);
template <typename ADataType,
typename BDataType,
typename CDataType,
typename ALayout,
typename BLayout,
typename CLayout>
struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGemmStreamK<
ALayout,
BLayout,
CLayout,
ADataType,
BDataType,
CDataType,
ck::tensor_operation::element_wise::PassThrough,
ck::tensor_operation::element_wise::PassThrough,
ck::tensor_operation::element_wise::PassThrough>>
{
using DeviceOp = DeviceGemmStreamK<ALayout,
BLayout,
CLayout,
ADataType,
BDataType,
CDataType,
ck::tensor_operation::element_wise::PassThrough,
ck::tensor_operation::element_wise::PassThrough,
ck::tensor_operation::element_wise::PassThrough>;
static auto GetInstances()
{
std::vector<std::unique_ptr<DeviceOp>> op_ptrs;
#if 0
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_splitk_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_splitk_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_splitk_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_splitk_f32_f32_f32_km_nk_mn_instances(op_ptrs);
}
}
else 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_splitk_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_splitk_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_splitk_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_splitk_f16_f16_f16_km_nk_mn_instances(op_ptrs);
}
}
#endif
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_streamk_f16_f16_f16_mk_kn_mn_instances(op_ptrs);
}
}
return op_ptrs;
}
};
} // namespace instance
} // namespace device
} // namespace tensor_operation
} // namespace ck

3
library/include/ck/library/utility/device_memory.hpp
View File

@@ -20,8 +20,9 @@ __global__ void set_buffer_value(T* p, T x, uint64_t buffer_element_size)
*/
struct DeviceMem
{
DeviceMem() = delete;
DeviceMem() : mpDeviceBuf(nullptr), mMemSize(0) {}
DeviceMem(std::size_t mem_size);
void Realloc(std::size_t mem_size);
void* GetDeviceBuffer() const;
std::size_t GetBufferSize() const;
void ToDevice(const void* p) const;