gemm_wmma_fp16.cpp 2.1 KB
Newer Older
aska-0096's avatar
aska-0096 committed
1
// SPDX-License-Identifier: MIT
Illia Silin's avatar
Illia Silin committed
2
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
aska-0096's avatar
aska-0096 committed
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21

#include "common.hpp"

#include "ck/tensor_operation/gpu/device/impl/device_gemm_wmma.hpp"

using ADataType        = ck::half_t;
using BDataType        = ck::half_t;
using AccDataType      = float;
using CShuffleDataType = float;
using CDataType        = ck::half_t;

using ALayout = Row;
using BLayout = Col;
using CLayout = Row;

using AElementOp = PassThrough;
using BElementOp = PassThrough;
using CElementOp = PassThrough;

Jing Zhang's avatar
Jing Zhang committed
22
static constexpr auto GemmDefault = ck::tensor_operation::device::GemmSpecialization::MNKPadding;
aska-0096's avatar
aska-0096 committed
23

Jing Zhang's avatar
Jing Zhang committed
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
// clang-format off
using DeviceGemmInstance = ck::tensor_operation::device::DeviceGemmWmma_CShuffle
         < ALayout,
           BLayout,
           CLayout,
           ADataType,
           BDataType,
           CDataType,
           AccDataType,
           CShuffleDataType,
           AElementOp,
           BElementOp,
           CElementOp,
           GemmDefault,
           1,           // Prefetch stage
           128,         // BlockSize
           64,          // MPerBlock
           128,         // NPerBlock
           64,          // KPerBlock
           8,           // K1
           16,          // MPerWmma
           16,          // NPerWmma
           2,           // M-Repeat // M-PerWmma / M-Repeat = M-Wave
           4,           // N-Repeat // N-PerWmma / N-Repeat = N-Wave
           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,
           1,           // C shuffle (M Repeat) Per store
           1,           // C shuffle (N Repeat) Per store
           S<1, 32, 1,  4>,
           8>;
// clang-format on
67

aska-0096's avatar
aska-0096 committed
68
69
70
71
72
73
using ReferenceGemmInstance = ck::tensor_operation::host::
    ReferenceGemm<ADataType, BDataType, CDataType, AccDataType, AElementOp, BElementOp, CElementOp>;

#include "run_gemm_example.inc"

int main(int argc, char* argv[]) { return !run_gemm_example(argc, argv); }