exp/raw_mem_bw.c

   1 /*
   2   Copyright 2008 Kristian Nielsen
   3
   4   This file is part of BeeDB.
   5
   6   Foobar is free software: you can redistribute it and/or modify
   7   it under the terms of the GNU General Public License as published by
   8   the Free Software Foundation, either version 2 of the License, or
   9   (at your option) any later version.
  10
  11   Foobar is distributed in the hope that it will be useful,
  12   but WITHOUT ANY WARRANTY; without even the implied warranty of
  13   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  14   GNU General Public License for more details.
  15
  16   You should have received a copy of the GNU General Public License
  17   along with Foobar.  If not, see <http://www.gnu.org/licenses/>.
  18 */
  19
  20 #include <stdint.h>
  21 #include <stdlib.h>
  22 #include <stdio.h>
  23
  24 typedef int v4si __attribute__ ((vector_size (16)));
  25 typedef float v4sf __attribute__ ((vector_size (16)));
  26
  27 #if 0
  28 void
  29 load_loop(float *p, uint64_t count)
  30 {
  31   if (0xf & (unsigned long)p)
  32   {
  33     fprintf(stderr, "load_loop: error: unaligned source\n");
  34     exit(1);
  35   }
  36   while (count > 0)
  37   {
  38     v4sf x0= __builtin_ia32_loadaps(p);
  39     v4sf x1= __builtin_ia32_loadaps(p+4);
  40     v4sf x2= __builtin_ia32_loadaps(p+8);
  41     v4sf x3= __builtin_ia32_loadaps(p+12);
  42   }
  43 }
  44 #endif
  45
  46 void
  47 foo(int *p)
  48 {
  49   v4si x= *(v4si *)p;
  50   v4si y= *(v4si *)(p+4);
  51   *(v4si *)(p+4)= x;
  52   *(v4si *)(p)= y;
  53 }
  54
  55 void
  56 load_loop(v4si *p, uint64_t count)
  57 {
  58   if (0xf & (unsigned long)p)
  59   {
  60     fprintf(stderr, "load_loop: error: unaligned source\n");
  61     exit(1);
  62   }
  63   v4si x;
  64   while (count-- > 0)
  65   {
  66     x+= p[0];
  67     x+= p[1];
  68     x+= p[2];
  69     x+= p[3];
  70     p+= 4;
  71   }
  72   *p= x;
  73 }
  74
  75 int
  76 main(int argc, char *argv)
  77 {
  78   uint64_t count= 1024*1024*16;
  79   uint64_t loops= 10;
  80   v4si *p= malloc((count+1)*64);
  81   while (loops--)
  82   {
  83     load_loop(p, count);
  84   }
  85   printf("%d\n", *(char *)(p+count));
  86 }