[753] | 1 | /* |
---|
| 2 | Author : Shay Gal-On, EEMBC |
---|
| 3 | |
---|
| 4 | This file is part of EEMBC(R) and CoreMark(TM), which are Copyright (C) 2009 |
---|
| 5 | All rights reserved. |
---|
| 6 | |
---|
| 7 | EEMBC CoreMark Software is a product of EEMBC and is provided under the terms of the |
---|
| 8 | CoreMark License that is distributed with the official EEMBC COREMARK Software release. |
---|
| 9 | If you received this EEMBC CoreMark Software without the accompanying CoreMark License, |
---|
| 10 | you must discontinue use and download the official release from www.coremark.org. |
---|
| 11 | |
---|
| 12 | Also, if you are publicly displaying scores generated from the EEMBC CoreMark software, |
---|
| 13 | make sure that you are in compliance with Run and Reporting rules specified in the accompanying readme.txt file. |
---|
| 14 | |
---|
| 15 | EEMBC |
---|
| 16 | 4354 Town Center Blvd. Suite 114-200 |
---|
| 17 | El Dorado Hills, CA, 95762 |
---|
| 18 | */ |
---|
| 19 | /* File: core_main.c |
---|
| 20 | This file contains the framework to acquire a block of memory, seed initial parameters, tun t he benchmark and report the results. |
---|
| 21 | */ |
---|
| 22 | #include "coremark.h" |
---|
| 23 | |
---|
| 24 | /* Function: iterate |
---|
| 25 | Run the benchmark for a specified number of iterations. |
---|
| 26 | |
---|
| 27 | Operation: |
---|
| 28 | For each type of benchmarked algorithm: |
---|
| 29 | a - Initialize the data block for the algorithm. |
---|
| 30 | b - Execute the algorithm N times. |
---|
| 31 | |
---|
| 32 | Returns: |
---|
| 33 | NULL. |
---|
| 34 | */ |
---|
| 35 | static ee_u16 list_known_crc[] = {(ee_u16)0xd4b0,(ee_u16)0x3340,(ee_u16)0x6a79,(ee_u16)0xe714,(ee_u16)0xe3c1}; |
---|
| 36 | static ee_u16 matrix_known_crc[] = {(ee_u16)0xbe52,(ee_u16)0x1199,(ee_u16)0x5608,(ee_u16)0x1fd7,(ee_u16)0x0747}; |
---|
| 37 | static ee_u16 state_known_crc[] = {(ee_u16)0x5e47,(ee_u16)0x39bf,(ee_u16)0xe5a4,(ee_u16)0x8e3a,(ee_u16)0x8d84}; |
---|
| 38 | void *iterate(void *pres) { |
---|
| 39 | ee_u32 i; |
---|
| 40 | ee_u16 crc; |
---|
| 41 | core_results *res=(core_results *)pres; |
---|
| 42 | ee_u32 iterations=res->iterations; |
---|
| 43 | res->crc=0; |
---|
| 44 | res->crclist=0; |
---|
| 45 | res->crcmatrix=0; |
---|
| 46 | res->crcstate=0; |
---|
| 47 | |
---|
| 48 | for (i=0; i<iterations; i++) { |
---|
| 49 | crc=core_bench_list(res,1); |
---|
| 50 | res->crc=crcu16(crc,res->crc); |
---|
| 51 | crc=core_bench_list(res,-1); |
---|
| 52 | res->crc=crcu16(crc,res->crc); |
---|
| 53 | if (i==0) res->crclist=res->crc; |
---|
| 54 | } |
---|
| 55 | return NULL; |
---|
| 56 | } |
---|
| 57 | |
---|
| 58 | #if (SEED_METHOD==SEED_ARG) |
---|
| 59 | ee_s32 get_seed_args(int i, int argc, char *argv[]); |
---|
| 60 | #define get_seed(x) (ee_s16)get_seed_args(x,argc,argv) |
---|
| 61 | #define get_seed_32(x) get_seed_args(x,argc,argv) |
---|
| 62 | #else /* via function or volatile */ |
---|
| 63 | ee_s32 get_seed_32(int i); |
---|
| 64 | #define get_seed(x) (ee_s16)get_seed_32(x) |
---|
| 65 | #endif |
---|
| 66 | |
---|
| 67 | #if (MEM_METHOD==MEM_STATIC) |
---|
| 68 | ee_u8 static_memblk[TOTAL_DATA_SIZE]; |
---|
| 69 | #endif |
---|
| 70 | char *mem_name[3] = {"Static","Heap","Stack"}; |
---|
| 71 | /* Function: main |
---|
| 72 | Main entry routine for the benchmark. |
---|
| 73 | This function is responsible for the following steps: |
---|
| 74 | |
---|
| 75 | 1 - Initialize input seeds from a source that cannot be determined at compile time. |
---|
| 76 | 2 - Initialize memory block for use. |
---|
| 77 | 3 - Run and time the benchmark. |
---|
| 78 | 4 - Report results, testing the validity of the output if the seeds are known. |
---|
| 79 | |
---|
| 80 | Arguments: |
---|
| 81 | 1 - first seed : Any value |
---|
| 82 | 2 - second seed : Must be identical to first for iterations to be identical |
---|
| 83 | 3 - third seed : Any value, should be at least an order of magnitude less then the input size, but bigger then 32. |
---|
| 84 | 4 - Iterations : Special, if set to 0, iterations will be automatically determined such that the benchmark will run between 10 to 100 secs |
---|
| 85 | |
---|
| 86 | */ |
---|
| 87 | |
---|
| 88 | #if MAIN_HAS_NOARGC |
---|
| 89 | MAIN_RETURN_TYPE main(void) { |
---|
| 90 | int argc=0; |
---|
| 91 | char *argv[1]; |
---|
| 92 | #else |
---|
| 93 | MAIN_RETURN_TYPE main(int argc, char *argv[]) { |
---|
| 94 | #endif |
---|
| 95 | ee_u16 i,j=0,num_algorithms=0; |
---|
| 96 | ee_s16 known_id=-1,total_errors=0; |
---|
| 97 | ee_u16 seedcrc=0; |
---|
| 98 | CORE_TICKS total_time; |
---|
| 99 | core_results results[MULTITHREAD]; |
---|
| 100 | #if (MEM_METHOD==MEM_STACK) |
---|
| 101 | ee_u8 stack_memblock[TOTAL_DATA_SIZE*MULTITHREAD]; |
---|
| 102 | #endif |
---|
| 103 | /* first call any initializations needed */ |
---|
| 104 | portable_init(&(results[0].port), &argc, argv); |
---|
| 105 | /* First some checks to make sure benchmark will run ok */ |
---|
| 106 | if (sizeof(struct list_head_s)>128) { |
---|
| 107 | ee_printf("list_head structure too big for comparable data!\n"); |
---|
| 108 | return MAIN_RETURN_VAL; |
---|
| 109 | } |
---|
| 110 | results[0].seed1=get_seed(1); |
---|
| 111 | results[0].seed2=get_seed(2); |
---|
| 112 | results[0].seed3=get_seed(3); |
---|
| 113 | results[0].iterations=get_seed_32(4); |
---|
| 114 | #if CORE_DEBUG |
---|
| 115 | results[0].iterations=1; |
---|
| 116 | #endif |
---|
| 117 | results[0].execs=get_seed_32(5); |
---|
| 118 | if (results[0].execs==0) { /* if not supplied, execute all algorithms */ |
---|
| 119 | results[0].execs=ALL_ALGORITHMS_MASK; |
---|
| 120 | } |
---|
| 121 | /* put in some default values based on one seed only for easy testing */ |
---|
| 122 | if ((results[0].seed1==0) && (results[0].seed2==0) && (results[0].seed3==0)) { /* validation run */ |
---|
| 123 | results[0].seed1=0; |
---|
| 124 | results[0].seed2=0; |
---|
| 125 | results[0].seed3=0x66; |
---|
| 126 | } |
---|
| 127 | if ((results[0].seed1==1) && (results[0].seed2==0) && (results[0].seed3==0)) { /* perfromance run */ |
---|
| 128 | results[0].seed1=0x3415; |
---|
| 129 | results[0].seed2=0x3415; |
---|
| 130 | results[0].seed3=0x66; |
---|
| 131 | } |
---|
| 132 | #if (MEM_METHOD==MEM_STATIC) |
---|
| 133 | results[0].memblock[0]=(void *)static_memblk; |
---|
| 134 | results[0].size=TOTAL_DATA_SIZE; |
---|
| 135 | results[0].err=0; |
---|
| 136 | #if (MULTITHREAD>1) |
---|
| 137 | #error "Cannot use a static data area with multiple contexts!" |
---|
| 138 | #endif |
---|
| 139 | #elif (MEM_METHOD==MEM_MALLOC) |
---|
| 140 | for (i=0 ; i<MULTITHREAD; i++) { |
---|
| 141 | ee_s32 malloc_override=get_seed(7); |
---|
| 142 | if (malloc_override != 0) |
---|
| 143 | results[i].size=malloc_override; |
---|
| 144 | else |
---|
| 145 | results[i].size=TOTAL_DATA_SIZE; |
---|
| 146 | results[i].memblock[0]=portable_malloc(results[i].size); |
---|
| 147 | results[i].seed1=results[0].seed1; |
---|
| 148 | results[i].seed2=results[0].seed2; |
---|
| 149 | results[i].seed3=results[0].seed3; |
---|
| 150 | results[i].err=0; |
---|
| 151 | results[i].execs=results[0].execs; |
---|
| 152 | } |
---|
| 153 | #elif (MEM_METHOD==MEM_STACK) |
---|
| 154 | for (i=0 ; i<MULTITHREAD; i++) { |
---|
| 155 | results[i].memblock[0]=stack_memblock+i*TOTAL_DATA_SIZE; |
---|
| 156 | results[i].size=TOTAL_DATA_SIZE; |
---|
| 157 | results[i].seed1=results[0].seed1; |
---|
| 158 | results[i].seed2=results[0].seed2; |
---|
| 159 | results[i].seed3=results[0].seed3; |
---|
| 160 | results[i].err=0; |
---|
| 161 | results[i].execs=results[0].execs; |
---|
| 162 | } |
---|
| 163 | #else |
---|
| 164 | #error "Please define a way to initialize a memory block." |
---|
| 165 | #endif |
---|
| 166 | /* Data init */ |
---|
| 167 | /* Find out how space much we have based on number of algorithms */ |
---|
| 168 | for (i=0; i<NUM_ALGORITHMS; i++) { |
---|
| 169 | if ((1<<(ee_u32)i) & results[0].execs) |
---|
| 170 | num_algorithms++; |
---|
| 171 | } |
---|
| 172 | for (i=0 ; i<MULTITHREAD; i++) |
---|
| 173 | results[i].size=results[i].size/num_algorithms; |
---|
| 174 | /* Assign pointers */ |
---|
| 175 | for (i=0; i<NUM_ALGORITHMS; i++) { |
---|
| 176 | ee_u32 ctx; |
---|
| 177 | if ((1<<(ee_u32)i) & results[0].execs) { |
---|
| 178 | for (ctx=0 ; ctx<MULTITHREAD; ctx++) |
---|
| 179 | results[ctx].memblock[i+1]=(char *)(results[ctx].memblock[0])+results[0].size*j; |
---|
| 180 | j++; |
---|
| 181 | } |
---|
| 182 | } |
---|
| 183 | /* call inits */ |
---|
| 184 | for (i=0 ; i<MULTITHREAD; i++) { |
---|
| 185 | if (results[i].execs & ID_LIST) { |
---|
| 186 | results[i].list=core_list_init(results[0].size,results[i].memblock[1],results[i].seed1); |
---|
| 187 | } |
---|
| 188 | if (results[i].execs & ID_MATRIX) { |
---|
| 189 | core_init_matrix(results[0].size, results[i].memblock[2], (ee_s32)results[i].seed1 | (((ee_s32)results[i].seed2) << 16), &(results[i].mat) ); |
---|
| 190 | } |
---|
| 191 | if (results[i].execs & ID_STATE) { |
---|
| 192 | core_init_state(results[0].size,results[i].seed1,results[i].memblock[3]); |
---|
| 193 | } |
---|
| 194 | } |
---|
| 195 | |
---|
| 196 | /* automatically determine number of iterations if not set */ |
---|
| 197 | if (results[0].iterations==0) { |
---|
| 198 | secs_ret secs_passed=0; |
---|
| 199 | ee_u32 divisor; |
---|
| 200 | results[0].iterations=1; |
---|
| 201 | while (secs_passed < (secs_ret)1) { |
---|
| 202 | results[0].iterations*=10; |
---|
| 203 | start_time(); |
---|
| 204 | iterate(&results[0]); |
---|
| 205 | stop_time(); |
---|
| 206 | secs_passed=time_in_secs(get_time()); |
---|
| 207 | } |
---|
| 208 | /* now we know it executes for at least 1 sec, set actual run time at about 10 secs */ |
---|
| 209 | divisor=(ee_u32)secs_passed; |
---|
| 210 | if (divisor==0) /* some machines cast float to int as 0 since this conversion is not defined by ANSI, but we know at least one second passed */ |
---|
| 211 | divisor=1; |
---|
| 212 | results[0].iterations*=1+10/divisor; |
---|
| 213 | } |
---|
| 214 | /* perform actual benchmark */ |
---|
| 215 | start_time(); |
---|
| 216 | #if (MULTITHREAD>1) |
---|
| 217 | if (default_num_contexts>MULTITHREAD) { |
---|
| 218 | default_num_contexts=MULTITHREAD; |
---|
| 219 | } |
---|
| 220 | for (i=0 ; i<default_num_contexts; i++) { |
---|
| 221 | results[i].iterations=results[0].iterations; |
---|
| 222 | results[i].execs=results[0].execs; |
---|
| 223 | core_start_parallel(&results[i]); |
---|
| 224 | } |
---|
| 225 | for (i=0 ; i<default_num_contexts; i++) { |
---|
| 226 | core_stop_parallel(&results[i]); |
---|
| 227 | } |
---|
| 228 | #else |
---|
| 229 | iterate(&results[0]); |
---|
| 230 | #endif |
---|
| 231 | stop_time(); |
---|
| 232 | total_time=get_time(); |
---|
| 233 | /* get a function of the input to report */ |
---|
| 234 | seedcrc=crc16(results[0].seed1,seedcrc); |
---|
| 235 | seedcrc=crc16(results[0].seed2,seedcrc); |
---|
| 236 | seedcrc=crc16(results[0].seed3,seedcrc); |
---|
| 237 | seedcrc=crc16(results[0].size,seedcrc); |
---|
| 238 | |
---|
| 239 | switch (seedcrc) { /* test known output for common seeds */ |
---|
| 240 | case 0x8a02: /* seed1=0, seed2=0, seed3=0x66, size 2000 per algorithm */ |
---|
| 241 | known_id=0; |
---|
| 242 | ee_printf("6k performance run parameters for coremark.\n"); |
---|
| 243 | break; |
---|
| 244 | case 0x7b05: /* seed1=0x3415, seed2=0x3415, seed3=0x66, size 2000 per algorithm */ |
---|
| 245 | known_id=1; |
---|
| 246 | ee_printf("6k validation run parameters for coremark.\n"); |
---|
| 247 | break; |
---|
| 248 | case 0x4eaf: /* seed1=0x8, seed2=0x8, seed3=0x8, size 400 per algorithm */ |
---|
| 249 | known_id=2; |
---|
| 250 | ee_printf("Profile generation run parameters for coremark.\n"); |
---|
| 251 | break; |
---|
| 252 | case 0xe9f5: /* seed1=0, seed2=0, seed3=0x66, size 666 per algorithm */ |
---|
| 253 | known_id=3; |
---|
| 254 | ee_printf("2K performance run parameters for coremark.\n"); |
---|
| 255 | break; |
---|
| 256 | case 0x18f2: /* seed1=0x3415, seed2=0x3415, seed3=0x66, size 666 per algorithm */ |
---|
| 257 | known_id=4; |
---|
| 258 | ee_printf("2K validation run parameters for coremark.\n"); |
---|
| 259 | break; |
---|
| 260 | default: |
---|
| 261 | total_errors=-1; |
---|
| 262 | break; |
---|
| 263 | } |
---|
| 264 | if (known_id>=0) { |
---|
| 265 | for (i=0 ; i<default_num_contexts; i++) { |
---|
| 266 | results[i].err=0; |
---|
| 267 | if ((results[i].execs & ID_LIST) && |
---|
| 268 | (results[i].crclist!=list_known_crc[known_id])) { |
---|
| 269 | ee_printf("[%u]ERROR! list crc 0x%04x - should be 0x%04x\n",i,results[i].crclist,list_known_crc[known_id]); |
---|
| 270 | results[i].err++; |
---|
| 271 | } |
---|
| 272 | if ((results[i].execs & ID_MATRIX) && |
---|
| 273 | (results[i].crcmatrix!=matrix_known_crc[known_id])) { |
---|
| 274 | ee_printf("[%u]ERROR! matrix crc 0x%04x - should be 0x%04x\n",i,results[i].crcmatrix,matrix_known_crc[known_id]); |
---|
| 275 | results[i].err++; |
---|
| 276 | } |
---|
| 277 | if ((results[i].execs & ID_STATE) && |
---|
| 278 | (results[i].crcstate!=state_known_crc[known_id])) { |
---|
| 279 | ee_printf("[%u]ERROR! state crc 0x%04x - should be 0x%04x\n",i,results[i].crcstate,state_known_crc[known_id]); |
---|
| 280 | results[i].err++; |
---|
| 281 | } |
---|
| 282 | total_errors+=results[i].err; |
---|
| 283 | } |
---|
| 284 | } |
---|
| 285 | total_errors+=check_data_types(); |
---|
| 286 | /* and report results */ |
---|
| 287 | ee_printf("CoreMark Size : %lu\n",(ee_u32)results[0].size); |
---|
| 288 | ee_printf("Total ticks : %lu\n",(ee_u32)total_time); |
---|
| 289 | #if HAS_FLOAT |
---|
| 290 | ee_printf("Total time (secs): %f\n",time_in_secs(total_time)); |
---|
| 291 | if (time_in_secs(total_time) > 0) |
---|
| 292 | ee_printf("Iterations/Sec : %f\n",default_num_contexts*results[0].iterations/time_in_secs(total_time)); |
---|
| 293 | #else |
---|
| 294 | ee_printf("Total time (secs): %d\n",time_in_secs(total_time)); |
---|
| 295 | if (time_in_secs(total_time) > 0) |
---|
| 296 | ee_printf("Iterations/Sec : %d\n",default_num_contexts*results[0].iterations/time_in_secs(total_time)); |
---|
| 297 | #endif |
---|
| 298 | if (time_in_secs(total_time) < 10) { |
---|
| 299 | ee_printf("ERROR! Must execute for at least 10 secs for a valid result!\n"); |
---|
| 300 | total_errors++; |
---|
| 301 | } |
---|
| 302 | |
---|
| 303 | ee_printf("Iterations : %lu\n",(ee_u32)default_num_contexts*results[0].iterations); |
---|
| 304 | ee_printf("Compiler version : %s\n",COMPILER_VERSION); |
---|
| 305 | ee_printf("Compiler flags : %s\n",COMPILER_FLAGS); |
---|
| 306 | #if (MULTITHREAD>1) |
---|
| 307 | ee_printf("Parallel %s : %d\n",PARALLEL_METHOD,default_num_contexts); |
---|
| 308 | #endif |
---|
| 309 | ee_printf("Memory location : %s\n",MEM_LOCATION); |
---|
| 310 | /* output for verification */ |
---|
| 311 | ee_printf("seedcrc : 0x%04x\n",seedcrc); |
---|
| 312 | if (results[0].execs & ID_LIST) |
---|
| 313 | for (i=0 ; i<default_num_contexts; i++) |
---|
| 314 | ee_printf("[%d]crclist : 0x%04x\n",i,results[i].crclist); |
---|
| 315 | if (results[0].execs & ID_MATRIX) |
---|
| 316 | for (i=0 ; i<default_num_contexts; i++) |
---|
| 317 | ee_printf("[%d]crcmatrix : 0x%04x\n",i,results[i].crcmatrix); |
---|
| 318 | if (results[0].execs & ID_STATE) |
---|
| 319 | for (i=0 ; i<default_num_contexts; i++) |
---|
| 320 | ee_printf("[%d]crcstate : 0x%04x\n",i,results[i].crcstate); |
---|
| 321 | for (i=0 ; i<default_num_contexts; i++) |
---|
| 322 | ee_printf("[%d]crcfinal : 0x%04x\n",i,results[i].crc); |
---|
| 323 | if (total_errors==0) { |
---|
| 324 | ee_printf("Correct operation validated. See readme.txt for run and reporting rules.\n"); |
---|
| 325 | #if HAS_FLOAT |
---|
| 326 | if (known_id==3) { |
---|
| 327 | ee_printf("CoreMark 1.0 : %f / %s %s",default_num_contexts*results[0].iterations/time_in_secs(total_time),COMPILER_VERSION,COMPILER_FLAGS); |
---|
| 328 | #if defined(MEM_LOCATION) && !defined(MEM_LOCATION_UNSPEC) |
---|
| 329 | ee_printf(" / %s",MEM_LOCATION); |
---|
| 330 | #else |
---|
| 331 | ee_printf(" / %s",mem_name[MEM_METHOD]); |
---|
| 332 | #endif |
---|
| 333 | |
---|
| 334 | #if (MULTITHREAD>1) |
---|
| 335 | ee_printf(" / %d:%s",default_num_contexts,PARALLEL_METHOD); |
---|
| 336 | #endif |
---|
| 337 | ee_printf("\n"); |
---|
| 338 | } |
---|
| 339 | #endif |
---|
| 340 | } |
---|
| 341 | if (total_errors>0) |
---|
| 342 | ee_printf("Errors detected\n"); |
---|
| 343 | if (total_errors<0) |
---|
| 344 | ee_printf("Cannot validate operation for these seed values, please compare with results on a known platform.\n"); |
---|
| 345 | |
---|
| 346 | #if (MEM_METHOD==MEM_MALLOC) |
---|
| 347 | for (i=0 ; i<MULTITHREAD; i++) |
---|
| 348 | portable_free(results[i].memblock[0]); |
---|
| 349 | #endif |
---|
| 350 | /* And last call any target specific code for finalizing */ |
---|
| 351 | portable_fini(&(results[0].port)); |
---|
| 352 | |
---|
| 353 | return MAIN_RETURN_VAL; |
---|
| 354 | } |
---|
| 355 | |
---|
| 356 | |
---|