1 // Copyright 2015 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 // eight_bit_int_gemm.h: exposes the standard EightBitIntGemm interface. 16 17 #ifndef GEMMLOWP_EIGHT_BIT_INT_GEMM_EIGHT_BIT_INT_GEMM_H_ 18 #define GEMMLOWP_EIGHT_BIT_INT_GEMM_EIGHT_BIT_INT_GEMM_H_ 19 20 #ifndef GEMMLOWP_USE_STLPORT 21 #include <cstdint> 22 #else 23 #include <stdint.h> 24 namespace std { 25 using ::uint8_t; 26 using ::int32_t; 27 using ::int64_t; 28 using ::uint64_t; 29 } 30 #endif 31 32 namespace gemmlowp { 33 34 namespace eight_bit_int_gemm { 35 36 // Concurrency / reentrancy notice 37 // =============================== 38 // 39 // This eight_bit_int_gemm has global singleton persistent state. 40 // A global lock ensures serialization of calls, so this library 41 // is fully reentrant but only one calling thread gets to actually run 42 // at a time, while other calling threads would wait. So it is safe 43 // albeit potentially slow to call the functions exposed here on 44 // multiple threads concurrently. 45 // 46 // Users who prefer a state-less, singleton-less interface, 47 // should use the main gemmlowp interface (public/gemmlowp.h) instead. 48 49 // The main entry point to compute a Gemm. This is the standard 50 enum class BitDepthSetting { 51 A8B8, // 8-bit a, 8-bit b 52 A5B7 // 5-bit a, 7-bit b 53 }; 54 55 // The main entry point to compute a Gemm. This is the standard 56 // EightBitIntGemm interface. 57 void EightBitIntGemm(bool transpose_a, bool transpose_b, bool transpose_c, 58 int m, int n, int k, const std::uint8_t *a, 59 std::int32_t a_offset, int lda, const std::uint8_t *b, 60 std::int32_t b_offset, int ldb, std::uint8_t *c, 61 std::int32_t c_offset, std::int32_t c_mult_int, 62 std::int32_t c_shift, int ldc, BitDepthSetting bit_depth); 63 64 void EightBitIntGemm(bool transpose_a, bool transpose_b, bool transpose_c, 65 int m, int n, int k, const std::uint8_t *a, 66 std::int32_t a_offset, int lda, const std::uint8_t *b, 67 std::int32_t b_offset, int ldb, float *c, float c_offset, 68 int ldc, BitDepthSetting bit_depth); 69 70 // Frees any persistent resources 71 // (threads, thread pools, allocators, buffers, ...) 72 // that gemmlowp might hold. This is called automatically 73 // on thread exit, but one may also call it earlier, at any time. 74 void FreePersistentResources(); 75 76 // Allows specifying the number of hardware threads, as a hint as to 77 // how many worker threads to use for sufficiently large Gemm's. 78 // We will never use more threads than that, but may use fewer, 79 // for instance on Gemm's that are too small to benefit from all 80 // available threads. The value 0 lets the implementation query 81 // the system to determine the number of hardware threads. 82 // Default value: 0. 83 void SetMaxNumThreads(int n); 84 85 } // namespace eight_bit_int_gemm 86 87 } // namespace gemmlowp 88 89 #endif // GEMMLOWP_EIGHT_BIT_INT_GEMM_EIGHT_BIT_INT_GEMM_H_ 90
