ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-05-17 11:30:02 +00:00
Code Issues Packages Projects Releases Wiki Activity

Fused GEMM+GEMM (#351)

Browse Source 
* initial stub for gemm_gemm_xdl_cshuffle

* set up example code

* compiles

* prevent integer overflow

* harmonize interface between ref_gemm and ref_batched_gemm

* batched_gemm_gemm

* fix example

* host tensor gen: diagonal pattern in lowest two-dimensions only

* make c descriptors containing only integral constants

* clean up

* add BlockwiseGemmXdlops_v2 while exploring an unified approach

* implement proper interface

* tidy up example

* fix compilation warnings

* coarsely controlled 2nd gemm padding

* remove rocm-cmake's hard requirement for certain revision

* clang-format

* resolve merge conflict

* fix compilation error on gfx10

* adds acc0 elementwise op to interface

* add gemm_gemm instances and tests

* avoid LDS data hazard

* fix build

Co-authored-by: Chao Liu <chao.liu2@amd.com>

[ROCm/composable_kernel commit: c20a75b07d]
This commit is contained in:
Anthony Chang
2022-08-13 22:18:58 +08:00
committed by GitHub
parent 4dab000987
commit bd30eaf33b
32 changed files with 2822 additions and 689 deletions
Download Patch File Download Diff File Expand all files Collapse all files

93
library/include/ck/library/tensor_operation_instance/gpu/batched_gemm_gemm.hpp Normal file
View File

@@ -0,0 +1,93 @@
// 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_batched_gemm_gemm.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_batched_gemm_gemm_xdl_cshuffle_f16_f16_f16_f16_gmk_gnk_gno_gmo_instance(
std::vector<std::unique_ptr<DeviceBatchedGemmGemm<Row,
Col,
Row,
Row,
F16,
F16,
F16,
F16,
PassThrough,
PassThrough,
PassThrough,
PassThrough,
PassThrough>>>& instances);
template <typename ALayout,
typename B0Layout,
typename B1Layout,
typename CLayout,
typename ADataType,
typename B0DataType,
typename B1DataType,
typename CDataType>
struct DeviceOperationInstanceFactory<
ck::tensor_operation::device::DeviceBatchedGemmGemm<ALayout,
B0Layout,
B1Layout,
CLayout,
ADataType,
B0DataType,
B1DataType,
CDataType,
PassThrough,
PassThrough,
PassThrough,
PassThrough,
PassThrough>>
{
using DeviceOp = DeviceBatchedGemmGemm<ALayout,
B0Layout,
B1Layout,
CLayout,
ADataType,
B0DataType,
B1DataType,
CDataType,
PassThrough,
PassThrough,
PassThrough,
PassThrough,
PassThrough>;
static auto GetInstances()
{
std::vector<std::unique_ptr<DeviceOp>> op_ptrs;
if constexpr(is_same_v<ADataType, half_t> && is_same_v<B0DataType, half_t> &&
is_same_v<B1DataType, half_t> && is_same_v<CDataType, half_t>)
{
if constexpr(is_same_v<ALayout, Row> && is_same_v<B0Layout, Col> &&
is_same_v<B1Layout, Row> && is_same_v<CLayout, Row>)
{
add_device_batched_gemm_gemm_xdl_cshuffle_f16_f16_f16_f16_gmk_gnk_gno_gmo_instance(
op_ptrs);
}
}
return op_ptrs;
}
};
} // namespace instance
} // namespace device
} // namespace tensor_operation
} // namespace ck