/src/highwayhash/highwayhash/instruction_sets.h
Line | Count | Source |
1 | | // Copyright 2017 Google Inc. All Rights Reserved. |
2 | | // |
3 | | // Licensed under the Apache License, Version 2.0 (the "License"); |
4 | | // you may not use this file except in compliance with the License. |
5 | | // You may obtain a copy of the License at |
6 | | // |
7 | | // http://www.apache.org/licenses/LICENSE-2.0 |
8 | | // |
9 | | // Unless required by applicable law or agreed to in writing, software |
10 | | // distributed under the License is distributed on an "AS IS" BASIS, |
11 | | // WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. |
12 | | // See the License for the specific language governing permissions and |
13 | | // limitations under the License. |
14 | | |
15 | | #ifndef HIGHWAYHASH_INSTRUCTION_SETS_H_ |
16 | | #define HIGHWAYHASH_INSTRUCTION_SETS_H_ |
17 | | |
18 | | // Calls the best specialization of a template supported by the current CPU. |
19 | | // |
20 | | // Usage: for each dispatch site, declare a Functor template with a 'Target' |
21 | | // argument, add a source file defining its operator() and instantiating |
22 | | // Functor<HH_TARGET>, add a cc_library_for_targets rule for that source file, |
23 | | // and call InstructionSets::Run<Functor>(/*args*/). |
24 | | |
25 | | #include <utility> // std::forward |
26 | | |
27 | | #include "highwayhash/arch_specific.h" // HH_TARGET_* |
28 | | #include "highwayhash/compiler_specific.h" |
29 | | |
30 | | namespace highwayhash { |
31 | | |
32 | | // Detects TargetBits and calls specializations of a user-defined functor. |
33 | | class InstructionSets { |
34 | | public: |
35 | | // Returns bit array of HH_TARGET_* supported by the current CPU. |
36 | | // The HH_TARGET_Portable bit is guaranteed to be set. |
37 | | #if HH_ARCH_X64 |
38 | | static TargetBits Supported(); |
39 | | #elif HH_ARCH_PPC |
40 | | static HH_INLINE TargetBits Supported() { |
41 | | return HH_TARGET_VSX | HH_TARGET_Portable; |
42 | | } |
43 | | #elif HH_ARCH_NEON |
44 | | static HH_INLINE TargetBits Supported() { |
45 | | return HH_TARGET_NEON | HH_TARGET_Portable; |
46 | | } |
47 | | #else |
48 | | static HH_INLINE TargetBits Supported() { return HH_TARGET_Portable; } |
49 | | #endif |
50 | | |
51 | | // Chooses the best available "Target" for the current CPU, runs the |
52 | | // corresponding Func<Target>::operator()(args) and returns that Target |
53 | | // (a single bit). The overhead of dispatching is low, about 4 cycles, but |
54 | | // this should only be called infrequently (e.g. hoisting it out of loops). |
55 | | template <template <TargetBits> class Func, typename... Args> |
56 | 49 | static HH_INLINE TargetBits Run(Args&&... args) { |
57 | 49 | #if HH_ARCH_X64 |
58 | 49 | const TargetBits supported = Supported(); |
59 | 49 | if (supported & HH_TARGET_AVX2) { |
60 | 49 | Func<HH_TARGET_AVX2>()(std::forward<Args>(args)...); |
61 | 49 | return HH_TARGET_AVX2; |
62 | 49 | } |
63 | 0 | if (supported & HH_TARGET_SSE41) { |
64 | 0 | Func<HH_TARGET_SSE41>()(std::forward<Args>(args)...); |
65 | 0 | return HH_TARGET_SSE41; |
66 | 0 | } |
67 | | #elif HH_ARCH_PPC |
68 | | const TargetBits supported = Supported(); |
69 | | if (supported & HH_TARGET_VSX) { |
70 | | Func<HH_TARGET_VSX>()(std::forward<Args>(args)...); |
71 | | return HH_TARGET_VSX; |
72 | | } |
73 | | #elif HH_ARCH_NEON |
74 | | const TargetBits supported = Supported(); |
75 | | if (supported & HH_TARGET_NEON) { |
76 | | Func<HH_TARGET_NEON>()(std::forward<Args>(args)...); |
77 | | return HH_TARGET_NEON; |
78 | | } |
79 | | #endif |
80 | | |
81 | | // No matching HH_ARCH or no supported HH_TARGET: |
82 | 0 | Func<HH_TARGET_Portable>()(std::forward<Args>(args)...); |
83 | 0 | return HH_TARGET_Portable; |
84 | 0 | } |
85 | | |
86 | | // Calls Func<Target>::operator()(args) for all Target supported by the |
87 | | // current CPU, and returns their HH_TARGET_* bits. |
88 | | template <template <TargetBits> class Func, typename... Args> |
89 | | static HH_INLINE TargetBits RunAll(Args&&... args) { |
90 | | const TargetBits supported = Supported(); |
91 | | |
92 | | #if HH_ARCH_X64 |
93 | | if (supported & HH_TARGET_AVX2) { |
94 | | Func<HH_TARGET_AVX2>()(std::forward<Args>(args)...); |
95 | | } |
96 | | if (supported & HH_TARGET_SSE41) { |
97 | | Func<HH_TARGET_SSE41>()(std::forward<Args>(args)...); |
98 | | } |
99 | | #elif HH_ARCH_PPC |
100 | | if (supported & HH_TARGET_VSX) { |
101 | | Func<HH_TARGET_VSX>()(std::forward<Args>(args)...); |
102 | | } |
103 | | |
104 | | #elif HH_ARCH_NEON |
105 | | if (supported & HH_TARGET_NEON) { |
106 | | Func<HH_TARGET_NEON>()(std::forward<Args>(args)...); |
107 | | } |
108 | | #endif |
109 | | |
110 | | Func<HH_TARGET_Portable>()(std::forward<Args>(args)...); |
111 | | |
112 | | return supported; // i.e. all that were run |
113 | | } |
114 | | }; |
115 | | |
116 | | } // namespace highwayhash |
117 | | |
118 | | #endif // HIGHWAYHASH_INSTRUCTION_SETS_H_ |