|
| 1 | +/***************************************************************************** |
| 2 | +Copyright (c) 2020, The OpenBLAS Project |
| 3 | +All rights reserved. |
| 4 | +
|
| 5 | +Redistribution and use in source and binary forms, with or without |
| 6 | +modification, are permitted provided that the following conditions are |
| 7 | +met: |
| 8 | +
|
| 9 | + 1. Redistributions of source code must retain the above copyright |
| 10 | + notice, this list of conditions and the following disclaimer. |
| 11 | +
|
| 12 | + 2. Redistributions in binary form must reproduce the above copyright |
| 13 | + notice, this list of conditions and the following disclaimer in |
| 14 | + the documentation and/or other materials provided with the |
| 15 | + distribution. |
| 16 | + 3. Neither the name of the OpenBLAS project nor the names of |
| 17 | + its contributors may be used to endorse or promote products |
| 18 | + derived from this software without specific prior written |
| 19 | + permission. |
| 20 | +
|
| 21 | +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" |
| 22 | +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE |
| 23 | +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE |
| 24 | +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE |
| 25 | +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL |
| 26 | +DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR |
| 27 | +SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER |
| 28 | +CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, |
| 29 | +OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE |
| 30 | +USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. |
| 31 | +
|
| 32 | +**********************************************************************************/ |
| 33 | + |
| 34 | +#include "common.h" |
| 35 | + |
| 36 | +void openblas_warning(int verbose, const char * msg); |
| 37 | + |
| 38 | +#ifdef SMALL_MATRIX_OPT |
| 39 | +static int inner_small_matrix_thread(blas_arg_t *args, BLASLONG *range_m, BLASLONG *range_n, IFLOAT *sa, IFLOAT *sb, BLASLONG mypos){ |
| 40 | + int routine_mode; |
| 41 | +#ifndef COMPLEX |
| 42 | + int (*gemm_small_kernel)(BLASLONG, BLASLONG, BLASLONG, FLOAT *, BLASLONG, FLOAT ,FLOAT *, BLASLONG, FLOAT, FLOAT *, BLASLONG); |
| 43 | + int (*gemm_small_kernel_b0)(BLASLONG, BLASLONG, BLASLONG, FLOAT *, BLASLONG, FLOAT, FLOAT *, BLASLONG, FLOAT *, BLASLONG); |
| 44 | +#else |
| 45 | + int (*zgemm_small_kernel)(BLASLONG, BLASLONG, BLASLONG, FLOAT *, BLASLONG, FLOAT , FLOAT, FLOAT *, BLASLONG, FLOAT , FLOAT, FLOAT *, BLASLONG); |
| 46 | + int (*zgemm_small_kernel_b0)(BLASLONG, BLASLONG, BLASLONG, FLOAT *, BLASLONG, FLOAT , FLOAT, FLOAT *, BLASLONG, FLOAT *, BLASLONG); |
| 47 | + FLOAT alpha[2], beta[2]; |
| 48 | +#endif |
| 49 | + routine_mode=args.routine_mode; |
| 50 | + if(routine_mode & BLAS_SMALL_B0_OPT){ |
| 51 | +#ifndef COMPLEX |
| 52 | + gemm_small_kernel_b0=args.routine; |
| 53 | + gemm_small_kernel_b0(args.m, args.n, args.k, args.a, args.lda, *(FLOAT *)(args.alpha), args.b, args.ldb, args.c, args.ldc); |
| 54 | +#else |
| 55 | + zgemm_small_kernel_b0=args.routine; |
| 56 | + alpha[0]=(FLOAT *)(args.alpha)[0]; |
| 57 | + alpha[1]=(FLOAT *)(args.alpha)[1]; |
| 58 | + zgemm_small_kernel_b0(args.m, args.n, args.k, args.a, args.lda, alpha[0], alpha[1], args.b, args.ldb, args.c, args.ldc); |
| 59 | +#endif |
| 60 | + }else if(routine_mode & BLAS_SMALL_OPT){ |
| 61 | +#ifndef COMPLEX |
| 62 | + gemm_small_kernel=args.routine; |
| 63 | + gemm_small_kernel(args.m, args.n, args.k, args.a, args.lda, *(FLOAT *)(args.alpha), args.b, args.ldb, *(FLOAT *)(args.beta), args.c, args.ldc); |
| 64 | +#else |
| 65 | + zgemm_small_kernel=args.routine; |
| 66 | + alpha[0]=(FLOAT *)(args.alpha)[0]; |
| 67 | + alpha[1]=(FLOAT *)(args.alpha)[1]; |
| 68 | + beta[0]=(FLOAT *)(args.beta)[0]; |
| 69 | + beta[1]=(FLOAT *)(args.beta)[1]; |
| 70 | + zgemm_small_kernel(args.m, args.n, args.k, args.a, args.lda, alpha[0], alpha[1], args.b, args.ldb, beta[0], beta[1], args.c, args.ldc); |
| 71 | +#endif |
| 72 | + } |
| 73 | +} |
| 74 | +#endif |
| 75 | + |
| 76 | +int CNAME(blas_arg_t * args_array, BLASLONG nums){ |
| 77 | + XFLOAT *buffer; |
| 78 | + XFLOAT *sa, *sb; |
| 79 | + int nthreads=1; |
| 80 | + int (*routine)(blas_arg_t *, void *, void *, double *, double *, BLASLONG); |
| 81 | + int i=0, j, current_nums; |
| 82 | + |
| 83 | +#ifdef SMP |
| 84 | + blas_queue_t * queue=NULL; |
| 85 | +#endif |
| 86 | + |
| 87 | + if(nums <=0 ) return 0; |
| 88 | + |
| 89 | + buffer = (XFLOAT *)blas_memory_alloc(0); |
| 90 | + sa = (XFLOAT *)((BLASLONG)buffer +GEMM_OFFSET_A); |
| 91 | + sb = (XFLOAT *)(((BLASLONG)sa + ((GEMM_P * GEMM_Q * COMPSIZE * SIZE + GEMM_ALIGN) & ~GEMM_ALIGN)) + GEMM_OFFSET_B); |
| 92 | + |
| 93 | +#ifdef SMP |
| 94 | + nthreads=num_cpu_avail(3); |
| 95 | + |
| 96 | + if(nthreads==1){ |
| 97 | +#endif |
| 98 | + //single thread |
| 99 | + for(i=0; i<nums; i++){ |
| 100 | + routine=args_array[i].routine; |
| 101 | +#ifdef SMALL_MATRIX_OPT |
| 102 | + if(args_array[i].routine_mode & BLAS_SMALL_OPT){ |
| 103 | + inner_small_matrix_thread(&args_array[i], NULL, NULL, NULL, NULL, 0); |
| 104 | + }else{ |
| 105 | +#endif |
| 106 | + routine(&args_array[i], NULL, NULL, sa, sb, 0); |
| 107 | +#ifdef SMALL_MATRIX_OPT |
| 108 | + } |
| 109 | +#endif |
| 110 | + } |
| 111 | +#ifdef SMP |
| 112 | + } else { |
| 113 | + //multi thread |
| 114 | + queue=(blas_queue_t *)malloc((nums+1) * sizeof(blas_queue_t)); |
| 115 | + if(queue == NULL){ |
| 116 | + openblas_warning(0, "memory alloc failed!\n"); |
| 117 | + exit(1); |
| 118 | + } |
| 119 | + for(i=0; i<nums; i++){ |
| 120 | + queue[i].args=&args_array[i]; |
| 121 | + queue[i].range_m=NULL; |
| 122 | + queue[i].range_n=NULL; |
| 123 | + queue[i].sa=NULL; |
| 124 | + queue[i].sb=NULL; |
| 125 | + queue[i].next=&queue[i+1]; |
| 126 | + |
| 127 | + queue[i].mode=args_array[i].routine_mode; |
| 128 | + queue[i].routine=args_array[i].routine; |
| 129 | + |
| 130 | +#ifdef SMALL_MATRIX_OPT |
| 131 | + if(args_array[i].routine_mode & BLAS_SMALL_OPT){ |
| 132 | + queue[i].routine=inner_small_matrix_thread; |
| 133 | + } |
| 134 | +#endif |
| 135 | + } |
| 136 | + |
| 137 | + for(i=0; i<nums; i+=nthreads){ |
| 138 | + current_nums=((nums-i)>nthreads)? nthreads: (nums-i); |
| 139 | + |
| 140 | + queue[i].sa=sa; |
| 141 | + queue[i].sb=sb; |
| 142 | + queue[i+current_nums-1].next=NULL; |
| 143 | + |
| 144 | + exec_blas(current_nums, &queue[i]); |
| 145 | + } |
| 146 | + free(queue); |
| 147 | + } |
| 148 | +#endif |
| 149 | + blas_memory_free(buffer); |
| 150 | + return 0; |
| 151 | +} |
0 commit comments