darshan-diff.c 15.8 KB
Newer Older
1
/*
2 3 4
 * Copyright (C) 2015 University of Chicago.
 * See COPYRIGHT notice in top-level directory.
 *
5 6
 */

7
#include <stdio.h>
8 9 10 11 12
#include <stdlib.h>
#include <inttypes.h>
#include <sys/types.h>
#include <assert.h>

13
#include "darshan-logutils.h"
14 15 16 17 18
#include "uthash-1.9.2/src/uthash.h"

struct darshan_mod_record_ref
{
    int rank;
19
    char *mod_dat;
20 21 22 23 24 25 26 27 28 29
    struct darshan_mod_record_ref *prev;
    struct darshan_mod_record_ref *next;
};

struct darshan_file_record_ref
{
    darshan_record_id rec_id;
    struct darshan_mod_record_ref *mod_recs[DARSHAN_MAX_MODS];
    UT_hash_handle hlink;
};
30

31 32
static int darshan_build_global_record_hash(
    darshan_fd fd, struct darshan_file_record_ref **rec_hash);
33

34
static void print_str_diff(char *prefix, char *arg1, char *arg2)
35 36 37
{
    printf("- %s %s\n", prefix, arg1);
    printf("+ %s %s\n", prefix, arg2);
38
    return;
39
}
40 41

static void print_int64_diff(char *prefix, int64_t arg1, int64_t arg2)
42
{
Philip Carns's avatar
Philip Carns committed
43 44
    printf("- %s %" PRId64 "\n", prefix, arg1);
    printf("+ %s %" PRId64 "\n", prefix, arg2);
45
    return;
46 47
}

48
int main(int argc, char *argv[])
49
{
50
    char *logfile1, *logfile2;
51 52
    darshan_fd file1, file2;
    struct darshan_job job1, job2;
53
    char exe1[4096], exe2[4096];
54 55
    struct darshan_name_record_ref *name_hash1 = NULL, *name_hash2 = NULL;
    struct darshan_name_record_ref *name_ref1, *name_ref2;
56 57 58 59 60 61 62 63
    struct darshan_file_record_ref *rec_hash1 = NULL, *rec_hash2 = NULL;
    struct darshan_file_record_ref *rec_ref1, *rec_ref2, *rec_tmp;
    struct darshan_mod_record_ref *mod_rec1, *mod_rec2;
    void *mod_buf1, *mod_buf2;
    struct darshan_base_record *base_rec1, *base_rec2;
    char *file_name1, *file_name2;
    int i;
    int ret;
64

65
    if(argc != 3)
66
    {
67
        fprintf(stderr, "Usage: darshan-diff <logfile1> <logfile2>\n");
68 69 70
        return(-1);
    }

71 72 73 74 75 76 77
    logfile1 = argv[1];
    logfile2 = argv[2];

    file1 = darshan_log_open(logfile1);
    if(!file1)
    {
        fprintf(stderr, "Error: unable to open darshan log file %s.\n", logfile1);
78 79
        return(-1);
    }
80 81 82 83 84 85

    file2 = darshan_log_open(logfile2);
    if(!file2)
    {
        darshan_log_close(file1);
        fprintf(stderr, "Error: unable to open darshan log file %s.\n", logfile2);
86 87 88
        return(-1);
    }

89
    /* get job data for each log file */
90
    ret = darshan_log_get_job(file1, &job1);
91
    if(ret < 0)
92
    {
93
        darshan_log_close(file1);
94 95
        darshan_log_close(file2);
        fprintf(stderr, "Error: unable to read job info for darshan log file %s.\n", logfile1);
96 97
        return(-1);
    }
98

99
    ret = darshan_log_get_job(file2, &job2);
100
    if(ret < 0)
101
    {
102
        darshan_log_close(file1);
103
        darshan_log_close(file2);
104
        fprintf(stderr, "Error: unable to read job info for darshan log file %s.\n", logfile2);
105 106 107
        return(-1);
    }

108
    /* get exe string for each log file */
109
    ret = darshan_log_get_exe(file1, exe1);
110
    if(ret < 0)
111
    {
112
        darshan_log_close(file1);
113 114
        darshan_log_close(file2);
        fprintf(stderr, "Error: unable to read exe for darshan log file %s.\n", logfile1);
115 116
        return(-1);
    }
117

118
    ret = darshan_log_get_exe(file2, exe2);
119
    if(ret < 0)
120
    {
121
        darshan_log_close(file1);
122
        darshan_log_close(file2);
123
        fprintf(stderr, "Error: unable to read exe for darshan log file %s.\n", logfile2);
124 125 126
        return(-1);
    }

127 128 129
    /* print diff of exe and job data */
    if (strcmp(exe1, exe2))
        print_str_diff("# exe: ", exe1, exe2);
130 131

    if (job1.uid != job2.uid)
132
        print_int64_diff("# uid:", job1.uid, job2.uid);
133
    if (job1.start_time != job2.start_time)
134 135 136 137 138
        print_int64_diff("# start_time:", job1.start_time, job2.start_time);
    if (job1.end_time != job2.end_time)
        print_int64_diff("# end_time:", job1.end_time, job2.end_time);
    if (job1.nprocs != job2.nprocs)
        print_int64_diff("# nprocs:", job1.nprocs, job2.nprocs);
139
    if ((job1.end_time-job1.start_time) != (job2.end_time - job2.start_time))
140 141
        print_int64_diff("# run time:",
                (int64_t)(job1.end_time - job1.start_time + 1),
142 143
                (int64_t)(job2.end_time - job2.start_time + 1));

144
    /* get hash of record ids to file names for each log */
145
    ret = darshan_log_get_namehash(file1, &name_hash1);
146 147 148 149 150 151 152 153
    if(ret < 0)
    {
        darshan_log_close(file1);
        darshan_log_close(file2);
        fprintf(stderr, "Error: unable to read record hash for darshan log file %s.\n", logfile1);
        return(-1);
    }

154
    ret = darshan_log_get_namehash(file2, &name_hash2);
155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201
    if(ret < 0)
    {
        darshan_log_close(file1);
        darshan_log_close(file2);
        fprintf(stderr, "Error: unable to read record hash for darshan log file %s.\n", logfile2);
        return(-1);
    }

    /* build hash tables of all records opened by all modules for each darshan log file */
    ret = darshan_build_global_record_hash(file1, &rec_hash1);
    if(ret < 0)
    {
        fprintf(stderr, "Error: unable to build record hash for darshan log file %s.\n", logfile1);
        darshan_log_close(file1);
        darshan_log_close(file2);
        return(-1);
    }

    ret = darshan_build_global_record_hash(file2, &rec_hash2);
    if(ret < 0)
    {
        fprintf(stderr, "Error: unable to build record hash for darshan log file %s.\n", logfile2);
        darshan_log_close(file1);
        darshan_log_close(file2);
        return(-1);
    }

    /* iterate records for the first log file and correlate/diff with records from
     * the second log file
     */
    HASH_ITER(hlink, rec_hash1, rec_ref1, rec_tmp)
    {
        printf("\n");

        /* search hash2 for this record */
        HASH_FIND(hlink, rec_hash2, &(rec_ref1->rec_id), sizeof(darshan_record_id), rec_ref2);

        for(i = 0; i < DARSHAN_MAX_MODS; i++)
        {
            /* TODO: skip modules that don't have the same format version, for now */
            if(rec_ref1->mod_recs[i] && rec_ref2 && rec_ref2->mod_recs[i] &&
                (file1->mod_ver[i] != file2->mod_ver[i]))
            {
                fprintf(stderr, "Warning: skipping %s module data due to incompatible"
                    "version numbers (file1=%d, file2=%d).\n",
                    darshan_module_names[i], file1->mod_ver[i], file2->mod_ver[i]);
                continue;
202
            }
203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243

            while(1)
            {
                mod_rec1 = rec_ref1->mod_recs[i];
                if(rec_ref2)
                    mod_rec2 = rec_ref2->mod_recs[i];
                else
                    mod_rec2 = NULL;

                if(mod_rec1 == NULL)
                    mod_buf1 = NULL;
                else
                    mod_buf1 = mod_rec1->mod_dat;
                if(mod_rec2 == NULL)
                    mod_buf2 = NULL;
                else
                    mod_buf2 = mod_rec2->mod_dat;

                base_rec1 = (struct darshan_base_record *)mod_buf1;
                base_rec2 = (struct darshan_base_record *)mod_buf2;
                if(!base_rec1 && !base_rec2)
                {
                    /* break out if there are no more records for this module */
                    break;
                }
                else if(base_rec1 && base_rec2)
                {
                    /* make sure if we have module records for this darshan record
                     * from both log files, that we are performing the diff rank-by-rank
                     * (i.e., we diff rank 1 records together, rank 2 records together,
                     * and so on)
                     */
                    if(base_rec1->rank < base_rec2->rank)
                        mod_buf2 = NULL;
                    else if(base_rec1->rank > base_rec2->rank)
                        mod_buf1 = NULL;
                }

                /* get corresponding file name for each record */
                if(mod_buf1)
                {
Shane Snyder's avatar
Shane Snyder committed
244
                    HASH_FIND(hlink, name_hash1, &(base_rec1->id),
245 246
                        sizeof(darshan_record_id), name_ref1);
                    assert(name_ref1);
247
                    file_name1 = name_ref1->name_record->name;
248 249 250
                }
                if(mod_buf2)
                {
Shane Snyder's avatar
Shane Snyder committed
251
                    HASH_FIND(hlink, name_hash2, &(base_rec2->id),
252 253
                        sizeof(darshan_record_id), name_ref2);
                    assert(name_ref2);
254
                    file_name2 = name_ref2->name_record->name;
255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271
                }

                mod_logutils[i]->log_print_diff(mod_buf1, file_name1, mod_buf2, file_name2);

                /* remove records which we have diffed already */
                if(mod_buf1)
                {
                    if(mod_rec1->next == mod_rec1)
                    {
                        rec_ref1->mod_recs[i] = NULL;
                    }
                    else
                    {
                        mod_rec1->prev->next = mod_rec1->next;
                        mod_rec1->next->prev = mod_rec1->prev;
                        rec_ref1->mod_recs[i] = mod_rec1->next;
                    }
272
                    free(mod_rec1->mod_dat);
273 274 275 276 277 278 279 280 281 282 283 284 285 286
                    free(mod_rec1);
                }
                if(mod_buf2)
                {
                    if(mod_rec2->next == mod_rec2)
                    {
                        rec_ref2->mod_recs[i] = NULL;
                    }
                    else
                    {
                        mod_rec2->prev->next = mod_rec2->next;
                        mod_rec2->next->prev = mod_rec2->prev;
                        rec_ref2->mod_recs[i] = mod_rec2->next;
                    }
287
                    free(mod_rec2->mod_dat);
288 289
                    free(mod_rec2);
                }
290 291 292
            }
        }

293 294 295 296 297 298 299 300
        HASH_DELETE(hlink, rec_hash1, rec_ref1);
        free(rec_ref1);
        if(rec_ref2)
        {
            HASH_DELETE(hlink, rec_hash2, rec_ref2);
            free(rec_ref2);
        }
    }
301

302 303 304 305 306 307 308 309 310 311 312 313
    /* iterate any remaning records from the 2nd log file and print the diff output --
     * NOTE: that these records do not have corresponding records in the first log file
     */
    HASH_ITER(hlink, rec_hash2, rec_ref2, rec_tmp)
    {
        for(i = 0; i < DARSHAN_MAX_MODS; i++)
        {
            while(rec_ref2->mod_recs[i])
            {
                mod_rec2 = rec_ref2->mod_recs[i];
                base_rec2 = (struct darshan_base_record *)mod_rec2->mod_dat;

Shane Snyder's avatar
Shane Snyder committed
314
                HASH_FIND(hlink, name_hash2, &(base_rec2->id),
315 316
                    sizeof(darshan_record_id), name_ref2);
                assert(name_ref2);
317
                file_name2 = name_ref2->name_record->name;
318

319 320 321 322 323 324 325 326 327 328 329 330 331
                mod_logutils[i]->log_print_diff(NULL, NULL, mod_rec2->mod_dat, file_name2);
                
                /* remove the record we just diffed */
                if(mod_rec2->next == mod_rec2)
                {
                    rec_ref2->mod_recs[i] = NULL;
                }
                else
                {
                    mod_rec2->prev->next = mod_rec2->next;
                    mod_rec2->next->prev = mod_rec2->prev;
                    rec_ref2->mod_recs[i] = mod_rec2->next;
                }
332
                free(mod_rec2->mod_dat);
333 334 335 336 337 338 339 340 341 342 343
                free(mod_rec2);
            }
        }

        HASH_DELETE(hlink, rec_hash2, rec_ref2);
        free(rec_ref2);
    }

    HASH_ITER(hlink, name_hash1, name_ref1, name_ref2)
    {
        HASH_DELETE(hlink, name_hash1, name_ref1);
344
        free(name_ref1->name_record);
345 346 347 348 349
        free(name_ref1);
    }
    HASH_ITER(hlink, name_hash2, name_ref2, name_ref1)
    {
        HASH_DELETE(hlink, name_hash2, name_ref2);
350
        free(name_ref2->name_record);
351 352
        free(name_ref2);
    }
353

354 355
    darshan_log_close(file1);
    darshan_log_close(file2);
356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382

    return(0);
}

static int darshan_build_global_record_hash(
    darshan_fd fd, struct darshan_file_record_ref **rec_hash)
{
    struct darshan_mod_record_ref *mod_rec;
    struct darshan_file_record_ref *file_rec;
    darshan_record_id tmp_rec_id;
    struct darshan_base_record *base_rec;
    int i;
    int ret;

    /* iterate over all modules in each log file, adding records to the
     * appropriate hash table
     */
    for(i = 0; i < DARSHAN_MAX_MODS; i++)
    {
        while(1)
        {
            if(!mod_logutils[i]) break;

            mod_rec = malloc(sizeof(struct darshan_mod_record_ref));
            assert(mod_rec);
            memset(mod_rec, 0, sizeof(struct darshan_mod_record_ref));

383
            ret = mod_logutils[i]->log_get_record(fd, (void **)&(mod_rec->mod_dat));
384 385 386 387
            if(ret < 0)
            {
                fprintf(stderr, "Error: unable to read module %s data from log file.\n",
                    darshan_module_names[i]);
388
                free(mod_rec->mod_dat);
389 390 391 392 393
                free(mod_rec);
                return(-1);
            }
            else if(ret == 0)
            {
394
                free(mod_rec->mod_dat);
395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467
                free(mod_rec);
                break;
            }
            else
            {
                base_rec = (struct darshan_base_record *)mod_rec->mod_dat;
                mod_rec->rank = base_rec->rank;

                HASH_FIND(hlink, *rec_hash, &tmp_rec_id, sizeof(darshan_record_id), file_rec);
                if(!file_rec)
                {
                    /* there is no entry in the global hash table of darshan records
                     * for this log file, so create one and add it.
                     */
                    file_rec = malloc(sizeof(struct darshan_file_record_ref));
                    assert(file_rec);

                    memset(file_rec, 0, sizeof(struct darshan_file_record_ref));
                    file_rec->rec_id = tmp_rec_id;
                    HASH_ADD(hlink, *rec_hash, rec_id, sizeof(darshan_record_id), file_rec);

                }

                /* add new record into the linked list of this module's records */
                if(file_rec->mod_recs[i])
                {
                    /* there is already an initialized linked list for this module */

                    /* we start at the end of the list and work backwards to insert this
                     * record (the list is sorted according to increasing ranks, and in
                     * general, darshan log records are sorted according to increasing
                     * ranks, as well)
                     */
                    struct darshan_mod_record_ref *tmp_mod_rec = file_rec->mod_recs[i]->prev;
                    while(1)
                    {
                        if(mod_rec->rank > tmp_mod_rec->rank)
                        {
                            /* insert new module record after this record */
                            mod_rec->prev = tmp_mod_rec;
                            mod_rec->next = tmp_mod_rec->next;
                            tmp_mod_rec->next->prev = mod_rec;
                            tmp_mod_rec->next = mod_rec;
                            break;
                        }
                        else if(mod_rec->rank < tmp_mod_rec->rank)
                        {
                            /* insert new module record before this record */
                            mod_rec->prev = tmp_mod_rec->prev;
                            mod_rec->next = tmp_mod_rec;
                            tmp_mod_rec->prev->next = mod_rec;
                            tmp_mod_rec->prev = mod_rec;
                            if(file_rec->mod_recs[i] == mod_rec->next)
                                file_rec->mod_recs[i] = mod_rec;
                            break;
                        }

                        tmp_mod_rec = tmp_mod_rec->prev;
                        assert(tmp_mod_rec != file_rec->mod_recs[i]);
                    }
                }
                else
                {
                    /* there are currently no records for this module, so just
                     * initialize a new linked list
                     */
                    mod_rec->prev = mod_rec->next = mod_rec;
                    file_rec->mod_recs[i] = mod_rec;
                }
            }
        }
    }

468 469 470 471 472 473 474 475 476 477 478
    return(0);
}

/*
 * Local variables:
 *  c-indent-level: 4
 *  c-basic-offset: 4
 * End:
 *
 * vim: ts=8 sts=4 sw=4 expandtab
 */