time_serial.c (5939B)
1 /* 2 Copyright 2010-2011, D. E. Shaw Research. 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 * Redistributions of source code must retain the above copyright 10 notice, this list of conditions, and the following disclaimer. 11 12 * Redistributions in binary form must reproduce the above copyright 13 notice, this list of conditions, and the following disclaimer in the 14 documentation and/or other materials provided with the distribution. 15 16 * Neither the name of D. E. Shaw Research nor the names of its 17 contributors may be used to endorse or promote products derived from 18 this software without specific prior written permission. 19 20 THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS 21 "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT 22 LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR 23 A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT 24 OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, 25 SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT 26 LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, 27 DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY 28 THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 29 (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE 30 OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 31 */ 32 /* 33 * Single-core CPU test and timing harness for Random123 RNGs. 34 * Uses macros and util_expandtpl.h to "templatize" over all the 35 * different permutations of RNGs and NxW and R. 36 */ 37 #include "util.h" 38 39 #include "Random123/philox.h" 40 #include "Random123/threefry.h" 41 #include "Random123/ars.h" 42 #include "Random123/aes.h" 43 44 #include "time_misc.h" 45 #include "util_print.h" 46 47 #include "util.h" 48 #define KERNEL R123_STATIC_INLINE 49 #define get_global_id(i) (i) 50 #include "time_random123.h" 51 52 static double cpu_hz = -1.; 53 54 #define TEST_TPL(NAME, N, W, R) \ 55 void NAME##N##x##W##_##R(NAME##N##x##W##_ctr_t ctr, NAME##N##x##W##_ukey_t ukey, NAME##N##x##W##_ctr_t kactr, unsigned count, void* unused) \ 56 { \ 57 const char *kernelname = #NAME #N "x" #W "_" #R; \ 58 double cur_time; \ 59 int n, niterations = numtrials; /* we make niterations + 2 (warmup, overhead) calls to the kernel */ \ 60 unsigned kcount = 0; \ 61 double basetime = 0., dt = 0., mindt = 0.; \ 62 NAME##N##x##W##_ctr_t C, *hC = &C; \ 63 (void)unused; /* suppress warning */ \ 64 \ 65 for (n = -2; n < niterations; n++) { \ 66 if (n == -2) { \ 67 if (count == 0) { \ 68 /* try to set a good guess for count */ \ 69 count = 1000000; \ 70 dprintf(("starting with count = %u\n", count)); \ 71 } \ 72 kcount = count; \ 73 } else if (n == -1) { \ 74 /* use first iteration time to calibrate count to get approximately sec_per_trial */ \ 75 if (count > 1) { \ 76 count = (unsigned)(count * sec_per_trial / dt); \ 77 dprintf(("scaled count = %u\n", count)); \ 78 } \ 79 /* second iteration is to calculate overhead after warmup */ \ 80 kcount = 1; \ 81 } else if (n == 0) { \ 82 int xj; \ 83 /* Check that we got the expected value */ \ 84 for (xj = 0; xj < N; xj++) { \ 85 if (kactr.v[xj] != hC[0].v[xj]) { \ 86 printf("%s mismatch: xj = %d, expected\n", kernelname, xj); \ 87 printline_##NAME##N##x##W##_##R(ukey, ctr, &kactr, 1); \ 88 printf(" but got\n"); \ 89 printline_##NAME##N##x##W##_##R(ukey, ctr, hC, 1); \ 90 if(!debug) exit(1); \ 91 else break; \ 92 } else { \ 93 dprintf(("%s matched word %d\n", kernelname, xj)); \ 94 } \ 95 } \ 96 basetime = dt; \ 97 if (debug||verbose) { \ 98 dprintf(("%s %.3f secs\n", kernelname, basetime)); \ 99 printline_##NAME##N##x##W##_##R(ukey, ctr, hC, 1); \ 100 } \ 101 kcount = count + 1; \ 102 } \ 103 /* calling timer *before* dprintf avoids an ARMv7 gcc 4.8.3 -O3 compiler bug! */ \ 104 (void)timer(&cur_time); \ 105 dprintf(("call function %s\n", kernelname)); \ 106 test_##NAME##N##x##W##_##R(kcount, ctr, ukey, hC); \ 107 dt = timer(&cur_time); \ 108 dprintf(("iteration %d took %.3f secs\n", n, dt)); \ 109 ALLZEROS(hC, 1, N); \ 110 if (n == 0 || dt < mindt) mindt = dt; \ 111 } \ 112 if (count > 1) { \ 113 double tpB = (mindt - basetime) / ( (kcount - 1.) * (N * W / 8.) ); \ 114 if(cpu_hz > 0.) \ 115 printf("%-17s %#5.3g cpB %#5.3g GB/s %u B granularity (best %u in %.3f s - %.6f s)\n", \ 116 kernelname, tpB*cpu_hz, 1e-9/tpB, \ 117 (unsigned)(N*W/8), kcount, mindt, basetime ); \ 118 else \ 119 printf("%-17s %#5.3g GB/s %u B granularity (best %u in %.3f s - %.6f s)\n", \ 120 kernelname, 1e-9/tpB, \ 121 (unsigned)(N*W/8), kcount, mindt, basetime ); \ 122 fflush(stdout); \ 123 } \ 124 } 125 126 #include "util_expandtpl.h" 127 128 int main(int argc, char **argv) 129 { 130 char *cp; 131 unsigned count = 0; 132 int keyctroffset = 0; 133 void* infop = NULL; 134 135 progname = argv[0]; 136 if (argc > 3|| (argv[1] && argv[1][0] == '-')) { 137 fprintf(stderr, "Usage: %s [COUNT]]\n", progname); 138 exit(1); 139 } 140 if (argc > 1) 141 count = atoi(argv[1]); 142 if ((cp = getenv("TIME_SERIAL_CPU_GHZ")) != NULL) { 143 cpu_hz = 1.e9 * atof(cp); 144 } 145 if ((cp = getenv("TIME_SERIAL_VERBOSE")) != NULL) { 146 verbose = atoi(cp); 147 } 148 if ((cp = getenv("TIME_SERIAL_DEBUG")) != NULL) { 149 debug = atoi(cp); 150 } 151 if ((cp = getenv("TIME_SERIAL_OFFSET")) != NULL) { 152 keyctroffset = atoi(cp); 153 } 154 if ((cp = getenv("TIME_SERIAL_NUMTRIALS")) != NULL) { 155 numtrials = atoi(cp); 156 } 157 if ((cp = getenv("TIME_SERIAL_SEC_PER_TRIAL")) != NULL) { 158 sec_per_trial = atof(cp); 159 } 160 # include "time_initkeyctr.h" 161 return 0; 162 }