1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
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
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
|
/***************************************************************************************************
* Copyright (c) 2017-2018, NVIDIA CORPORATION. All rights reserved.
*
* Redistribution and use in source and binary forms, with or without modification, are permitted
* provided that the following conditions are met:
* * Redistributions of source code must retain the above copyright notice, this list of
* conditions and the following disclaimer.
* * Redistributions in binary form must reproduce the above copyright notice, this list of
* conditions and the following disclaimer in the documentation and/or other materials
* provided with the distribution.
* * Neither the name of the NVIDIA CORPORATION nor the names of its contributors may be used
* to endorse or promote products derived from this software without specific prior written
* permission.
*
* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR
* IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND
* FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL NVIDIA CORPORATION BE LIABLE
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
* BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS;
* OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
* STRICT LIABILITY, OR TOR (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
* OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*
**************************************************************************************************/
/*! \file
\brief Template implementing matrix multiply-add operations on fragments.
*/
#pragma once
#include <cutlass/fragment.h>
namespace cutlass {
namespace gemm {
////////////////////////////////////////////////////////////////////////////////////////////////////
/// Template performing matrix multiply-add operation within a thread
template <typename AccumulatorsPerThread_,
typename ThreadsPerWarp_,
typename ScalarA_,
typename ScalarB_,
typename ScalarC_>
struct ThreadMultiplyAdd {
/// The shape of the instruction.
typedef Shape<1, 1, 1, 1> InstructionShape;
/// The number of accumulators per thread.
typedef AccumulatorsPerThread_ AccumulatorsPerThread;
/// The number of threads per warp.
typedef ThreadsPerWarp_ ThreadsPerWarp;
/// The number of accumulators per warp.
typedef typename ShapeMul<AccumulatorsPerThread, ThreadsPerWarp>::Shape AccumulatorsPerWarp;
/// The type for A.
typedef ScalarA_ ScalarA;
/// The fragment for A.
typedef Fragment<ScalarA, AccumulatorsPerThread::kW> FragmentA;
/// The type for B.
typedef ScalarB_ ScalarB;
/// The fragment for B.
typedef Fragment<ScalarB, AccumulatorsPerThread::kH> FragmentB;
/// The type for C and D.
typedef ScalarC_ ScalarC;
/// The accumulators.
typedef Fragment<ScalarC, AccumulatorsPerThread::kH * AccumulatorsPerThread::kW, 16> Accumulators;
/// Ctor.
CUTLASS_DEVICE ThreadMultiplyAdd() {}
/// Multiply : d = a*b + c.
CUTLASS_DEVICE void multiply_add(FragmentA const& a,
FragmentB const& b,
Accumulators const& c,
Accumulators& d) {
for (int j = 0; j < AccumulatorsPerThread::kH; ++j) {
for (int i = 0; i < AccumulatorsPerThread::kW; ++i) {
d[j * AccumulatorsPerThread::kW + i] = a[i] * b[j] + c[j * AccumulatorsPerThread::kW + i];
}
}
}
};
////////////////////////////////////////////////////////////////////////////////////////////////////
} // namespace gemm
} // namespace cutlass
|
