| 1 | //                               -*- Mode: C -*- 
 | 
|---|
| 2 | // 
 | 
|---|
| 3 | // Cforall Version 1.0.0 Copyright (C) 2021 University of Waterloo
 | 
|---|
| 4 | // 
 | 
|---|
| 5 | // PRNG.c -- 
 | 
|---|
| 6 | // 
 | 
|---|
| 7 | // Author           : Peter A. Buhr
 | 
|---|
| 8 | // Created On       : Wed Dec 29 09:38:12 2021
 | 
|---|
| 9 | // Last Modified By : Peter A. Buhr
 | 
|---|
| 10 | // Last Modified On : Sat Feb 12 12:23:57 2022
 | 
|---|
| 11 | // Update Count     : 342
 | 
|---|
| 12 | // 
 | 
|---|
| 13 | 
 | 
|---|
| 14 | #include <fstream.hfa>                                                                  // sout
 | 
|---|
| 15 | #include <stdlib.hfa>                                                                   // PRNG
 | 
|---|
| 16 | #include <clock.hfa>
 | 
|---|
| 17 | #include <thread.hfa>
 | 
|---|
| 18 | #include <limits.hfa>                                                                   // MAX
 | 
|---|
| 19 | #include <math.hfa>                                                                             // sqrt
 | 
|---|
| 20 | #include <malloc.h>                                                                             // malloc_stats
 | 
|---|
| 21 | #include <locale.h>                                                                             // setlocale
 | 
|---|
| 22 | #include <mutex_stmt.hfa>
 | 
|---|
| 23 | 
 | 
|---|
| 24 | // FIX ME: spurious characters appear in output
 | 
|---|
| 25 | Duration default_preemption() { return 0; }
 | 
|---|
| 26 | 
 | 
|---|
| 27 | #ifdef TIME                                                                                             // use -O2 -nodebug
 | 
|---|
| 28 | #define STARTTIME start = timeHiRes()
 | 
|---|
| 29 | #define ENDTIME( extra ) sout | wd(0,1, (timeHiRes() - start)`ms / 1000.) | extra "seconds"
 | 
|---|
| 30 | enum { BUCKETS = 100_000, TRIALS = 1_000_000_000 };
 | 
|---|
| 31 | #else
 | 
|---|
| 32 | #define STARTTIME
 | 
|---|
| 33 | #define ENDTIME( extra )
 | 
|---|
| 34 | enum { BUCKETS = 100_000, TRIALS = 100_000_000 };
 | 
|---|
| 35 | #endif // TIME
 | 
|---|
| 36 | 
 | 
|---|
| 37 | void avgstd( unsigned int buckets[] ) {
 | 
|---|
| 38 |         unsigned int min = MAX, max = 0;
 | 
|---|
| 39 |         double sum = 0.0, diff;
 | 
|---|
| 40 |         for ( i; BUCKETS ) {
 | 
|---|
| 41 |                 if ( buckets[i] < min ) min = buckets[i];
 | 
|---|
| 42 |                 if ( buckets[i] > max ) max = buckets[i];
 | 
|---|
| 43 |                 sum += buckets[i];
 | 
|---|
| 44 |         } // for
 | 
|---|
| 45 | 
 | 
|---|
| 46 |         double avg = sum / BUCKETS;                                                     // average
 | 
|---|
| 47 |         sum = 0.0;
 | 
|---|
| 48 |         for ( i; BUCKETS ) {                                                            // sum squared differences from average
 | 
|---|
| 49 |                 diff = buckets[i] - avg;
 | 
|---|
| 50 |                 sum += diff * diff;
 | 
|---|
| 51 |         } // for
 | 
|---|
| 52 |         double std = sqrt( sum / BUCKETS );
 | 
|---|
| 53 |         mutex( sout ) sout | "trials"  | TRIALS | "buckets" | BUCKETS
 | 
|---|
| 54 |                 | "min" | min | "max" | max
 | 
|---|
| 55 |                 | "avg" | wd(0,1, avg) | "std" | wd(0,1, std) | "rstd" | wd(0,1, (avg == 0 ? 0.0 : std / avg * 100)) | "%";
 | 
|---|
| 56 | } // avgstd
 | 
|---|
| 57 | 
 | 
|---|
| 58 | 
 | 
|---|
| 59 | uint32_t seed = 1009;
 | 
|---|
| 60 | 
 | 
|---|
| 61 | thread T1 {};
 | 
|---|
| 62 | void main( T1 & ) {
 | 
|---|
| 63 |         unsigned int * buckets = calloc( BUCKETS );                     // too big for task stack
 | 
|---|
| 64 |         for ( TRIALS / 100 ) {
 | 
|---|
| 65 |                 buckets[rand() % BUCKETS] += 1;                                 // concurrent
 | 
|---|
| 66 |         } // for
 | 
|---|
| 67 |         avgstd( buckets );
 | 
|---|
| 68 |         free( buckets );
 | 
|---|
| 69 | } // main
 | 
|---|
| 70 | 
 | 
|---|
| 71 | thread T2 {};
 | 
|---|
| 72 | void main( T2 & ) {
 | 
|---|
| 73 |         PRNG prng;
 | 
|---|
| 74 |         if ( seed != 0 ) set_seed( prng, seed );
 | 
|---|
| 75 |         unsigned int * buckets = calloc( BUCKETS );                     // too big for task stack
 | 
|---|
| 76 |         for ( TRIALS ) {
 | 
|---|
| 77 |                 buckets[prng( prng ) % BUCKETS] += 1;                   // concurrent
 | 
|---|
| 78 |         } // for
 | 
|---|
| 79 |         avgstd( buckets );
 | 
|---|
| 80 |         free( buckets );
 | 
|---|
| 81 | } // main
 | 
|---|
| 82 | 
 | 
|---|
| 83 | thread T3 {};
 | 
|---|
| 84 | void main( T3 & th ) {
 | 
|---|
| 85 |         unsigned int * buckets = calloc( BUCKETS );                     // too big for task stack
 | 
|---|
| 86 |         for ( TRIALS ) {
 | 
|---|
| 87 |                 buckets[prng() % BUCKETS] += 1;                                 // concurrent
 | 
|---|
| 88 |         } // for
 | 
|---|
| 89 |         avgstd( buckets );
 | 
|---|
| 90 |         free( buckets );
 | 
|---|
| 91 | } // main
 | 
|---|
| 92 | 
 | 
|---|
| 93 | thread T4 {};
 | 
|---|
| 94 | void main( T4 & th ) {
 | 
|---|
| 95 |         unsigned int * buckets = calloc( BUCKETS );                     // too big for task stack
 | 
|---|
| 96 |         for ( TRIALS ) {
 | 
|---|
| 97 |                 buckets[prng( th ) % BUCKETS] += 1;     // concurrent
 | 
|---|
| 98 |         } // for
 | 
|---|
| 99 |         avgstd( buckets );
 | 
|---|
| 100 |         free( buckets );
 | 
|---|
| 101 | } // main
 | 
|---|
| 102 | 
 | 
|---|
| 103 | // Compiler bug requires hiding declaration of th from the bucket access, otherwise the compiler thinks th is aliased
 | 
|---|
| 104 | // and continually reloads it from memory, which doubles the cost.
 | 
|---|
| 105 | static void dummy( thread$ & th ) __attribute__(( noinline ));
 | 
|---|
| 106 | static void dummy( thread$ & th ) {
 | 
|---|
| 107 |         unsigned int * buckets = (unsigned int *)calloc( BUCKETS, sizeof(unsigned int) ); // too big for task stack
 | 
|---|
| 108 |         for ( unsigned int i = 0; i < TRIALS; i += 1 ) {
 | 
|---|
| 109 |                 buckets[prng( th ) % BUCKETS] += 1;                             // sequential
 | 
|---|
| 110 |         } // for
 | 
|---|
| 111 |         avgstd( buckets );
 | 
|---|
| 112 |         free( buckets );
 | 
|---|
| 113 | } // dummy
 | 
|---|
| 114 | 
 | 
|---|
| 115 | int main() {
 | 
|---|
| 116 |         // causes leaked storage message
 | 
|---|
| 117 | //      setlocale( LC_NUMERIC, getenv( "LANG" ) );                      // print digit separator
 | 
|---|
| 118 | 
 | 
|---|
| 119 |         enum { TASKS = 4 };
 | 
|---|
| 120 |         Time start;
 | 
|---|
| 121 | #ifdef TIME                                                                                             // too slow for test and generates non-repeatable results
 | 
|---|
| 122 | #if 1
 | 
|---|
| 123 |         unsigned int rseed;
 | 
|---|
| 124 |         if ( seed != 0 ) rseed = seed;
 | 
|---|
| 125 |         else rseed = rdtscl();
 | 
|---|
| 126 |         srand( rseed );
 | 
|---|
| 127 | 
 | 
|---|
| 128 |         sout | sepDisable;
 | 
|---|
| 129 |         sout | wd(13, "rand()" ) | wd(10, "rand(5)") | wd(13, "rand(0,5)" );
 | 
|---|
| 130 |         for ( 20 ) {
 | 
|---|
| 131 |                 sout | wd(13, rand()) | nonl;
 | 
|---|
| 132 |                 sout | wd(10, rand() % 5) | nonl;
 | 
|---|
| 133 |                 sout | wd(13, rand() % (5 - 0 + 1) + 0);
 | 
|---|
| 134 |         } // for
 | 
|---|
| 135 |         sout | sepEnable;
 | 
|---|
| 136 |         sout | "seed" | rseed;
 | 
|---|
| 137 | 
 | 
|---|
| 138 |         sout | nl | "Sequential";
 | 
|---|
| 139 |         STARTTIME;
 | 
|---|
| 140 |         {
 | 
|---|
| 141 |                 unsigned int * buckets = calloc( BUCKETS );             // too big for task stack
 | 
|---|
| 142 |                 for ( i; TRIALS / 10 ) {
 | 
|---|
| 143 |                         buckets[rand() % BUCKETS] += 1;                         // sequential
 | 
|---|
| 144 |                 } // for
 | 
|---|
| 145 |                 avgstd( buckets );
 | 
|---|
| 146 |                 free( buckets );
 | 
|---|
| 147 |         }
 | 
|---|
| 148 |         ENDTIME( " x 10 " );
 | 
|---|
| 149 | 
 | 
|---|
| 150 |         sout | nl | "Concurrent";
 | 
|---|
| 151 |         STARTTIME;
 | 
|---|
| 152 |         {
 | 
|---|
| 153 |                 processor p[TASKS - 1];                                                 // already 1 processor
 | 
|---|
| 154 |                 {
 | 
|---|
| 155 |                         T1 t[TASKS];
 | 
|---|
| 156 |                 } // wait for threads to complete
 | 
|---|
| 157 |         }
 | 
|---|
| 158 |         ENDTIME( " x 100 " );
 | 
|---|
| 159 | #endif // 0
 | 
|---|
| 160 | #endif // TIME
 | 
|---|
| 161 | #if 1
 | 
|---|
| 162 |         PRNG prng;
 | 
|---|
| 163 |         if ( seed != 0 ) set_seed( prng, seed );
 | 
|---|
| 164 | 
 | 
|---|
| 165 |         sout | sepDisable;
 | 
|---|
| 166 |         sout | nl | wd(13, "PRNG()" ) | wd(10, "PRNG(5)") | wd(13, "PRNG(0,5)" );
 | 
|---|
| 167 |         for ( 20 ) {
 | 
|---|
| 168 |                 sout | wd(13, prng( prng )) | nonl;                             // cascading => side-effect functions called in arbitary order
 | 
|---|
| 169 |                 sout | wd(10, prng( prng, 5 )) | nonl;
 | 
|---|
| 170 |                 sout | wd(13, prng( prng, 0, 5 ));
 | 
|---|
| 171 |         } // for
 | 
|---|
| 172 |         sout | sepEnable;
 | 
|---|
| 173 |         sout | "seed" | get_seed( prng );
 | 
|---|
| 174 | 
 | 
|---|
| 175 |         sout | nl | "Sequential";
 | 
|---|
| 176 |         STARTTIME;
 | 
|---|
| 177 |         {
 | 
|---|
| 178 |                 unsigned int * buckets = calloc( BUCKETS );             // too big for task stack
 | 
|---|
| 179 |                 for ( TRIALS ) {
 | 
|---|
| 180 |                         buckets[prng( prng ) % BUCKETS] += 1;           // sequential
 | 
|---|
| 181 |                 } // for
 | 
|---|
| 182 |                 avgstd( buckets );
 | 
|---|
| 183 |                 free( buckets );
 | 
|---|
| 184 |         }
 | 
|---|
| 185 |         ENDTIME();
 | 
|---|
| 186 | 
 | 
|---|
| 187 |         sout | nl | "Concurrent";
 | 
|---|
| 188 |         STARTTIME;
 | 
|---|
| 189 |         {
 | 
|---|
| 190 |                 processor p[TASKS - 1];                                                 // already 1 processor
 | 
|---|
| 191 |                 {
 | 
|---|
| 192 |                         T2 t[TASKS];
 | 
|---|
| 193 |                 } // wait for threads to complete
 | 
|---|
| 194 |         }
 | 
|---|
| 195 |         ENDTIME();
 | 
|---|
| 196 | #endif // 0
 | 
|---|
| 197 | #if 1
 | 
|---|
| 198 |         if ( seed != 0 ) set_seed( seed );
 | 
|---|
| 199 | 
 | 
|---|
| 200 |         sout | sepDisable;
 | 
|---|
| 201 |         sout | nl | wd(13, "prng()" ) | wd(10, "prng(5)") | wd(13, "prng(0,5)" );
 | 
|---|
| 202 |         for ( 20 ) {
 | 
|---|
| 203 |                 sout | wd(13, prng()) | nonl;                                   // cascading => side-effect functions called in arbitary order
 | 
|---|
| 204 |                 sout | wd(10, prng( 5 )) | nonl;
 | 
|---|
| 205 |                 sout | wd(13, prng( 0, 5 ));
 | 
|---|
| 206 |         } // for
 | 
|---|
| 207 |         sout | sepEnable;
 | 
|---|
| 208 |         sout | "seed" | get_seed( prng );
 | 
|---|
| 209 | 
 | 
|---|
| 210 |         sout | nl | "Sequential";
 | 
|---|
| 211 |         STARTTIME;
 | 
|---|
| 212 |         {
 | 
|---|
| 213 |                 unsigned int * buckets = calloc( BUCKETS );             // too big for task stack
 | 
|---|
| 214 |                 for ( TRIALS ) {
 | 
|---|
| 215 |                         buckets[prng() % BUCKETS] += 1;
 | 
|---|
| 216 |                 } // for
 | 
|---|
| 217 |                 avgstd( buckets );
 | 
|---|
| 218 |                 free( buckets );
 | 
|---|
| 219 |         }
 | 
|---|
| 220 |         ENDTIME();
 | 
|---|
| 221 | 
 | 
|---|
| 222 |         sout | nl | "Concurrent";
 | 
|---|
| 223 |         STARTTIME;
 | 
|---|
| 224 |         {
 | 
|---|
| 225 |                 processor p[TASKS - 1];                                                 // already 1 processor
 | 
|---|
| 226 |                 {
 | 
|---|
| 227 |                         T3 t[TASKS];
 | 
|---|
| 228 |                 } // wait for threads to complete
 | 
|---|
| 229 |         }
 | 
|---|
| 230 |         ENDTIME();
 | 
|---|
| 231 | #endif // 0
 | 
|---|
| 232 | #if 1
 | 
|---|
| 233 |         if ( seed != 0 ) set_seed( seed );
 | 
|---|
| 234 |         thread$ & th = *active_thread();
 | 
|---|
| 235 | 
 | 
|---|
| 236 |         sout | sepDisable;
 | 
|---|
| 237 |         sout | nl | wd(13, "prng(t)" ) | wd(10, "prng(t,5)") | wd(13, "prng(t,0,5)" );
 | 
|---|
| 238 |         for ( 20 ) {
 | 
|---|
| 239 |                 sout | wd(13, prng( th )) | nonl;                               // cascading => side-effect functions called in arbitary order
 | 
|---|
| 240 |                 sout | wd(10, prng( th, 5 )) | nonl;
 | 
|---|
| 241 |                 sout | wd(13, prng( th, 0, 5 ));
 | 
|---|
| 242 |         } // for
 | 
|---|
| 243 |         sout | sepEnable;
 | 
|---|
| 244 |         sout | "seed" | get_seed( prng );
 | 
|---|
| 245 | 
 | 
|---|
| 246 |         sout | nl | "Sequential";
 | 
|---|
| 247 |         STARTTIME;
 | 
|---|
| 248 |         {
 | 
|---|
| 249 |                 dummy( th );
 | 
|---|
| 250 |         }
 | 
|---|
| 251 |         ENDTIME();
 | 
|---|
| 252 | 
 | 
|---|
| 253 |         sout | nl | "Concurrent";
 | 
|---|
| 254 |         STARTTIME;
 | 
|---|
| 255 |         {
 | 
|---|
| 256 |                 processor p[TASKS - 1];                                                 // already 1 processor
 | 
|---|
| 257 |                 {
 | 
|---|
| 258 |                         T4 t[TASKS];
 | 
|---|
| 259 |                 } // wait for threads to complete
 | 
|---|
| 260 |         }
 | 
|---|
| 261 |         ENDTIME();
 | 
|---|
| 262 | #endif // 0
 | 
|---|
| 263 | //      malloc_stats();
 | 
|---|
| 264 | } // main
 | 
|---|
| 265 | 
 | 
|---|
| 266 | 
 | 
|---|
| 267 | // Local Variables: //
 | 
|---|
| 268 | // compile-command: "cfa -DTIME -O2 -nodebug PRNG.cfa" //
 | 
|---|
| 269 | // End: //
 | 
|---|