ceph/src/isa-l/erasure_code/gf_vect_dot_prod_avx_perf.c

   1 /**********************************************************************
   2   Copyright(c) 2011-2015 Intel Corporation All rights reserved.
   3
   4   Redistribution and use in source and binary forms, with or without
   5   modification, are permitted provided that the following conditions
   6   are met:
   7     * Redistributions of source code must retain the above copyright
   8       notice, this list of conditions and the following disclaimer.
   9     * Redistributions in binary form must reproduce the above copyright
  10       notice, this list of conditions and the following disclaimer in
  11       the documentation and/or other materials provided with the
  12       distribution.
  13     * Neither the name of Intel Corporation nor the names of its
  14       contributors may be used to endorse or promote products derived
  15       from this software without specific prior written permission.
  16
  17   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
  18   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
  19   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
  20   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
  21   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
  22   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
  23   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
  24   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
  25   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
  26   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
  27   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  28 **********************************************************************/
  29
  30 #include <stdio.h>
  31 #include <stdlib.h>
  32 #include <string.h>             // for memset, memcmp
  33 #include "erasure_code.h"
  34 #include "test.h"
  35
  36 #ifndef FUNCTION_UNDER_TEST
  37 # define FUNCTION_UNDER_TEST gf_vect_dot_prod_avx
  38 #endif
  39
  40 #define str(s) #s
  41 #define xstr(s) str(s)
  42
  43 //#define CACHED_TEST
  44 #ifdef CACHED_TEST
  45 // Cached test, loop many times over small dataset
  46 # define TEST_SOURCES 10
  47 # define TEST_LEN     8*1024
  48 # define TEST_LOOPS   40000
  49 # define TEST_TYPE_STR "_warm"
  50 #else
  51 # ifndef TEST_CUSTOM
  52 // Uncached test.  Pull from large mem base.
  53 #  define TEST_SOURCES 10
  54 #  define GT_L3_CACHE  32*1024*1024     /* some number > last level cache */
  55 #  define TEST_LEN     ((GT_L3_CACHE / TEST_SOURCES) & ~(64-1))
  56 #  define TEST_LOOPS   100
  57 #  define TEST_TYPE_STR "_cold"
  58 # else
  59 #  define TEST_TYPE_STR "_cus"
  60 #  ifndef TEST_LOOPS
  61 #    define TEST_LOOPS 1000
  62 #  endif
  63 # endif
  64 #endif
  65
  66 typedef unsigned char u8;
  67
  68 void dump(unsigned char *buf, int len)
  69 {
  70         int i;
  71         for (i = 0; i < len;) {
  72                 printf(" %2x", 0xff & buf[i++]);
  73                 if (i % 32 == 0)
  74                         printf("\n");
  75         }
  76         printf("\n");
  77 }
  78
  79 void dump_matrix(unsigned char **s, int k, int m)
  80 {
  81         int i, j;
  82         for (i = 0; i < k; i++) {
  83                 for (j = 0; j < m; j++) {
  84                         printf(" %2x", s[i][j]);
  85                 }
  86                 printf("\n");
  87         }
  88         printf("\n");
  89 }
  90
  91 int main(int argc, char *argv[])
  92 {
  93         int i, j;
  94         void *buf;
  95         u8 g[TEST_SOURCES], g_tbls[TEST_SOURCES * 32], *dest, *dest_ref;
  96         u8 *temp_buff, *buffs[TEST_SOURCES];
  97         struct perf start, stop;
  98
  99         printf(xstr(FUNCTION_UNDER_TEST) ": %dx%d\n", TEST_SOURCES, TEST_LEN);
 100
 101         // Allocate the arrays
 102         for (i = 0; i < TEST_SOURCES; i++) {
 103                 if (posix_memalign(&buf, 64, TEST_LEN)) {
 104                         printf("alloc error: Fail");
 105                         return -1;
 106                 }
 107                 buffs[i] = buf;
 108         }
 109
 110         if (posix_memalign(&buf, 64, TEST_LEN)) {
 111                 printf("alloc error: Fail");
 112                 return -1;
 113         }
 114         dest = buf;
 115
 116         if (posix_memalign(&buf, 64, TEST_LEN)) {
 117                 printf("alloc error: Fail");
 118                 return -1;
 119         }
 120         dest_ref = buf;
 121
 122         if (posix_memalign(&buf, 64, TEST_LEN)) {
 123                 printf("alloc error: Fail");
 124                 return -1;
 125         }
 126         temp_buff = buf;
 127
 128         // Performance test
 129         for (i = 0; i < TEST_SOURCES; i++)
 130                 for (j = 0; j < TEST_LEN; j++)
 131                         buffs[i][j] = rand();
 132
 133         memset(dest, 0, TEST_LEN);
 134         memset(temp_buff, 0, TEST_LEN);
 135         memset(dest_ref, 0, TEST_LEN);
 136         memset(g, 0, TEST_SOURCES);
 137
 138         for (i = 0; i < TEST_SOURCES; i++)
 139                 g[i] = rand();
 140
 141         for (j = 0; j < TEST_SOURCES; j++)
 142                 gf_vect_mul_init(g[j], &g_tbls[j * 32]);
 143
 144         gf_vect_dot_prod_base(TEST_LEN, TEST_SOURCES, &g_tbls[0], buffs, dest_ref);
 145
 146 #ifdef DO_REF_PERF
 147         perf_start(&start);
 148         for (i = 0; i < TEST_LOOPS; i++) {
 149                 for (j = 0; j < TEST_SOURCES; j++)
 150                         gf_vect_mul_init(g[j], &g_tbls[j * 32]);
 151
 152                 gf_vect_dot_prod_base(TEST_LEN, TEST_SOURCES, &g_tbls[0], buffs, dest_ref);
 153         }
 154         perf_stop(&stop);
 155         printf("gf_vect_dot_prod_base" TEST_TYPE_STR ": ");
 156         perf_print(stop, start, (long long)TEST_LEN * (TEST_SOURCES + 1) * i);
 157 #endif
 158
 159         FUNCTION_UNDER_TEST(TEST_LEN, TEST_SOURCES, g_tbls, buffs, dest);
 160
 161         perf_start(&start);
 162         for (i = 0; i < TEST_LOOPS; i++) {
 163                 for (j = 0; j < TEST_SOURCES; j++)
 164                         gf_vect_mul_init(g[j], &g_tbls[j * 32]);
 165
 166                 FUNCTION_UNDER_TEST(TEST_LEN, TEST_SOURCES, g_tbls, buffs, dest);
 167         }
 168         perf_stop(&stop);
 169         printf(xstr(FUNCTION_UNDER_TEST) TEST_TYPE_STR ": ");
 170         perf_print(stop, start, (long long)TEST_LEN * (TEST_SOURCES + 1) * i);
 171
 172         if (0 != memcmp(dest_ref, dest, TEST_LEN)) {
 173                 printf("Fail zero " xstr(FUNCTION_UNDER_TEST) " test\n");
 174                 dump_matrix(buffs, 5, TEST_SOURCES);
 175                 printf("dprod_base:");
 176                 dump(dest_ref, 25);
 177                 printf("dprod:");
 178                 dump(dest, 25);
 179                 return -1;
 180         }
 181
 182         printf("pass perf check\n");
 183         return 0;
 184 }