Line data    Source code

       1             : /*************************************************************************
       2             :  *
       3             :  *  Project
       4             :  *                         _____ _____  __  __ _____
       5             :  *                        / ____|  __ \|  \/  |  __ \
       6             :  *  ___  _ __   ___ _ __ | |  __| |__) | \  / | |__) |
       7             :  * / _ \| '_ \ / _ \ '_ \| | |_ |  ___/| |\/| |  ___/
       8             :  *| (_) | |_) |  __/ | | | |__| | |    | |  | | |
       9             :  * \___/| .__/ \___|_| |_|\_____|_|    |_|  |_|_|
      10             :  *      | |
      11             :  *      |_|
      12             :  *
      13             :  * Copyright (C) Akiel Aries, <akiel@akiel.org>, et al.
      14             :  *
      15             :  * This software is licensed as described in the file LICENSE, which
      16             :  * you should have received as part of this distribution. The terms
      17             :  * among other details are referenced in the official documentation
      18             :  * seen here : https://akielaries.github.io/openGPMP/ along with
      19             :  * important files seen in this project.
      20             :  *
      21             :  * You may opt to use, copy, modify, merge, publish, distribute
      22             :  * and/or sell copies of the Software, and permit persons to whom
      23             :  * the Software is furnished to do so, under the terms of the
      24             :  * LICENSE file. As this is an Open Source effort, all implementations
      25             :  * must be of the same methodology.
      26             :  *
      27             :  *
      28             :  *
      29             :  * This software is distributed on an AS IS basis, WITHOUT
      30             :  * WARRANTY OF ANY KIND, either express or implied.
      31             :  *
      32             :  ************************************************************************/
      33             : #include <cassert>
      34             : #include <cstddef>
      35             : #include <cstdint>
      36             : #include <iostream>
      37             : #include <openGPMP/linalg/mtx.hpp>
      38             : #include <vector>
      39             : 
      40             : #if defined(__x86_64__) || defined(__amd64__) || defined(__amd64)
      41             : 
      42             : /************************************************************************
      43             :  *
      44             :  * Matrix Operations for AVX ISA
      45             :  *
      46             :  ************************************************************************/
      47             : #if defined(__AVX2__)
      48             : 
      49             : // AVX family intrinsics
      50             : #include <immintrin.h>
      51             : 
      52             : /************************************************************************
      53             :  *
      54             :  * Matrix Operations on Arrays
      55             :  *
      56             :  ************************************************************************/
      57             : 
      58             : // matrix addition using Intel intrinsics, accepts float arrays as matrices
      59           0 : void gpmp::linalg::Mtx::mtx_add(const float *A,
      60             :                                 const float *B,
      61             :                                 float *C,
      62             :                                 int rows,
      63             :                                 int cols) {
      64           0 :     if (rows > 16) {
      65           0 :         for (int i = 0; i < rows; ++i) {
      66           0 :             int j = 0;
      67             :             // requires at least size 4x4 size matrices
      68           0 :             for (; j < cols - 7; j += 8) {
      69             :                 // load 4 elements from A, B, and C matrices using SIMD
      70           0 :                 __m256 a = _mm256_loadu_ps(&A[i * cols + j]);
      71           0 :                 __m256 b = _mm256_loadu_ps(&B[i * cols + j]);
      72           0 :                 __m256 c = _mm256_loadu_ps(&C[i * cols + j]);
      73             :                 // perform vectorized addition and accumulate the result
      74           0 :                 c = _mm256_add_ps(a, b);
      75             : 
      76             :                 // store the result back to the C matrix
      77           0 :                 _mm256_storeu_ps(&C[i * cols + j], c);
      78             :             }
      79             : 
      80             :             // handle the remaining elements that are not multiples of 8
      81           0 :             for (; j < cols; ++j) {
      82           0 :                 C[i * cols + j] = A[i * cols + j] + B[i * cols + j];
      83             :             }
      84             :         }
      85             :     } else {
      86             :         // use standard matrix addition
      87           0 :         std_mtx_add(A, B, C, rows, cols);
      88             :     }
      89           0 : }
      90             : 
      91             : #endif
      92             : 
      93             : // x86
      94             : #endif