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
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
|
/*
* Copyright (c) 2017-2018 ARM Limited.
*
* SPDX-License-Identifier: MIT
*
* Permission is hereby granted, free of charge, to any person obtaining a copy
* of this software and associated documentation files (the "Software"), to
* deal in the Software without restriction, including without limitation the
* rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
* sell copies of the Software, and to permit persons to whom the Software is
* furnished to do so, subject to the following conditions:
*
* The above copyright notice and this permission notice shall be included in all
* copies or substantial portions of the Software.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
* AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
* OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
* SOFTWARE.
*/
// This can only be built if the target/compiler supports FP16 arguments.
#ifdef __ARM_FP16_ARGS
#include "arm_gemm.hpp"
#include "gemm_common.hpp"
#include "gemm_implementation.hpp"
#include "gemm_interleaved.hpp"
#include "kernels/a64_hgemm_24x8.hpp"
#include "kernels/a64_sgemm_12x8.hpp"
#include "kernels/a32_sgemm_8x6.hpp"
#include "kernels/sve_interleaved_fp16_mla_3VLx8.hpp"
namespace arm_gemm {
#ifdef __ARM_FEATURE_SVE
class GemmImpl_gemm_fp16_interleaved_fp16 : public GemmImplementation<__fp16, __fp16> {
public:
UniqueGemmCommon<__fp16, __fp16> instantiate(const GemmArgs<__fp16> &args) override {
return UniqueGemmCommon<__fp16, __fp16>(new GemmInterleaved<interleaved_fp16_mla_3VLx8, __fp16, __fp16>(args));
}
GemmImpl_gemm_fp16_interleaved_fp16() : GemmImplementation<__fp16, __fp16>(GemmMethod::GEMM_INTERLEAVED_FP16) { }
};
#elif defined(__aarch64__)
#if defined(__ARM_FEATURE_FP16_VECTOR_ARITHMETIC) || defined(FP16_KERNELS)
class GemmImpl_gemm_fp16_interleaved_fp16 : public GemmImplementation<__fp16, __fp16> {
public:
#ifndef __ARM_FEATURE_FP16_VECTOR_ARITHMETIC
bool is_supported(const GemmArgs<__fp16> &args) override {
return args._ci->has_fp16();
}
#endif
UniqueGemmCommon<__fp16, __fp16> instantiate(const GemmArgs<__fp16> &args) override {
return UniqueGemmCommon<__fp16, __fp16>(new GemmInterleaved<hgemm_24x8, __fp16, __fp16>(args));
}
GemmImpl_gemm_fp16_interleaved_fp16() : GemmImplementation<__fp16, __fp16>(GemmMethod::GEMM_INTERLEAVED_FP16) { }
};
#endif
#endif // __aarch64__
class GemmImpl_gemm_fp16_interleaved : public GemmImplementation<__fp16, __fp16> {
public:
UniqueGemmCommon<__fp16, __fp16> instantiate(const GemmArgs<__fp16> &args) override {
#ifdef __aarch64__
return UniqueGemmCommon<__fp16, __fp16>(new GemmInterleaved<sgemm_12x8, __fp16, __fp16>(args));
#elif defined(__arm__)
return UniqueGemmCommon<__fp16, __fp16>(new GemmInterleaved<sgemm_8x6, __fp16, __fp16>(args));
#else
# error Unknown Architecture
#endif
}
GemmImpl_gemm_fp16_interleaved() : GemmImplementation<__fp16, __fp16>(GemmMethod::GEMM_INTERLEAVED) { }
};
#if defined(__aarch64__) && (defined(__ARM_FEATURE_VECTOR_ARITHMETIC) || defined(FP16_KERNELS) || defined(__ARM_FEATURE_SVE))
static GemmImpl_gemm_fp16_interleaved_fp16 gemm_fp16_interleaved_fp16_impl{};
#endif
static GemmImpl_gemm_fp16_interleaved gemm_fp16_interleaved_impl{};
static std::vector<GemmImplementation<__fp16, __fp16> *> gemm_fp16_methods = {
#if defined(__aarch64__) && (defined(__ARM_FEATURE_VECTOR_ARITHMETIC) || defined(FP16_KERNELS) || defined(__ARM_FEATURE_SVE))
&gemm_fp16_interleaved_fp16_impl,
#endif
&gemm_fp16_interleaved_impl
};
template<>
std::vector<GemmImplementation<__fp16, __fp16> *> &gemm_implementation_list<__fp16, __fp16>() {
return gemm_fp16_methods;
}
/* Explicitly instantiate the external functions for these types. */
template UniqueGemmCommon<__fp16, __fp16> gemm<__fp16, __fp16>(GemmArgs<__fp16> &args, GemmConfig *cfg);
template GemmMethod get_gemm_method<__fp16, __fp16>(GemmArgs<__fp16> &args);
template bool method_is_compatible<__fp16, __fp16>(GemmMethod method, GemmArgs<__fp16> &args);
} // namespace arm_gemm
#endif // __ARM_FP16_ARGS
|
