darshan-logutils.c 33.1 KB
Newer Older
1
2
3
4
5
/*
 *  (C) 2009 by Argonne National Laboratory.
 *      See COPYRIGHT in top-level directory.
 */

Philip Carns's avatar
Philip Carns committed
6
#include "darshan-config.h"
7
8
#include <stdio.h>
#include <string.h>
9
#include <assert.h>
10
#include <stdlib.h>
11
#include <inttypes.h>
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
#include "darshan-logutils.h"

/* isn't there a clever c way to avoid this? */
char *darshan_names[] = {
    "CP_INDEP_OPENS",
    "CP_COLL_OPENS",               /* count of MPI collective opens */
    "CP_INDEP_READS",              /* count of independent MPI reads */
    "CP_INDEP_WRITES",             /* count of independent MPI writes */
    "CP_COLL_READS",               /* count of collective MPI reads */
    "CP_COLL_WRITES",              /* count of collective MPI writes */
    "CP_SPLIT_READS",              /* count of split collective MPI reads */
    "CP_SPLIT_WRITES",             /* count of split collective MPI writes */
    "CP_NB_READS",                 /* count of nonblocking MPI reads */
    "CP_NB_WRITES",                /* count of nonblocking MPI writes */
    "CP_SYNCS",                    /* count of MPI_File_sync */
    "CP_POSIX_READS",              /* count of posix reads */
    "CP_POSIX_WRITES",             /* count of posix writes */
    "CP_POSIX_OPENS",              /* count of posix opens */
    "CP_POSIX_SEEKS",              /* count of posix seeks */
    "CP_POSIX_STATS",              /* count of posix stat/lstat/fstats */
    "CP_POSIX_MMAPS",              /* count of posix mmaps */
    "CP_POSIX_FREADS",
    "CP_POSIX_FWRITES",
    "CP_POSIX_FOPENS",
    "CP_POSIX_FSEEKS",
    "CP_POSIX_FSYNCS",
    "CP_POSIX_FDSYNCS",
39
40
41
    "CP_INDEP_NC_OPENS",
    "CP_COLL_NC_OPENS",
    "CP_HDF5_OPENS",
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
    "CP_COMBINER_NAMED",           /* count of each MPI datatype category */
    "CP_COMBINER_DUP",
    "CP_COMBINER_CONTIGUOUS",
    "CP_COMBINER_VECTOR",
    "CP_COMBINER_HVECTOR_INTEGER",
    "CP_COMBINER_HVECTOR",
    "CP_COMBINER_INDEXED",
    "CP_COMBINER_HINDEXED_INTEGER",
    "CP_COMBINER_HINDEXED",
    "CP_COMBINER_INDEXED_BLOCK",
    "CP_COMBINER_STRUCT_INTEGER",
    "CP_COMBINER_STRUCT",
    "CP_COMBINER_SUBARRAY",
    "CP_COMBINER_DARRAY",
    "CP_COMBINER_F90_REAL",
    "CP_COMBINER_F90_COMPLEX",
    "CP_COMBINER_F90_INTEGER",
    "CP_COMBINER_RESIZED",
    "CP_HINTS",                     /* count of MPI hints used */
    "CP_VIEWS",                     /* count of MPI set view calls */
    "CP_MODE",                      /* mode of file */
    "CP_BYTES_READ",                /* total bytes read */
    "CP_BYTES_WRITTEN",             /* total bytes written */
    "CP_MAX_BYTE_READ",             /* highest offset byte read */
    "CP_MAX_BYTE_WRITTEN",          /* highest offset byte written */
    "CP_CONSEC_READS",              /* count of consecutive reads */
    "CP_CONSEC_WRITES",             /* count of consecutive writes */
    "CP_SEQ_READS",                 /* count of sequential reads */
    "CP_SEQ_WRITES",                /* count of sequential writes */
    "CP_RW_SWITCHES",
    "CP_MEM_NOT_ALIGNED",           /* count of accesses not mem aligned */
    "CP_MEM_ALIGNMENT",             /* mem alignment in bytes */
    "CP_FILE_NOT_ALIGNED",          /* count of accesses not file aligned */
    "CP_FILE_ALIGNMENT",            /* file alignment in bytes */
76
77
    "CP_MAX_READ_TIME_SIZE",
    "CP_MAX_WRITE_TIME_SIZE",
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
    "CP_SIZE_READ_0_100",           /* count of posix read size ranges */
    "CP_SIZE_READ_100_1K",
    "CP_SIZE_READ_1K_10K",
    "CP_SIZE_READ_10K_100K",
    "CP_SIZE_READ_100K_1M",
    "CP_SIZE_READ_1M_4M",
    "CP_SIZE_READ_4M_10M",
    "CP_SIZE_READ_10M_100M",
    "CP_SIZE_READ_100M_1G",
    "CP_SIZE_READ_1G_PLUS",
    "CP_SIZE_WRITE_0_100",          /* count of posix write size ranges */
    "CP_SIZE_WRITE_100_1K",
    "CP_SIZE_WRITE_1K_10K",
    "CP_SIZE_WRITE_10K_100K",
    "CP_SIZE_WRITE_100K_1M",
    "CP_SIZE_WRITE_1M_4M",
    "CP_SIZE_WRITE_4M_10M",
    "CP_SIZE_WRITE_10M_100M",
    "CP_SIZE_WRITE_100M_1G",
    "CP_SIZE_WRITE_1G_PLUS",
    "CP_SIZE_READ_AGG_0_100",       /* count of MPI read size ranges */
    "CP_SIZE_READ_AGG_100_1K",
    "CP_SIZE_READ_AGG_1K_10K",
    "CP_SIZE_READ_AGG_10K_100K",
    "CP_SIZE_READ_AGG_100K_1M",
    "CP_SIZE_READ_AGG_1M_4M",
    "CP_SIZE_READ_AGG_4M_10M",
    "CP_SIZE_READ_AGG_10M_100M",
    "CP_SIZE_READ_AGG_100M_1G",
    "CP_SIZE_READ_AGG_1G_PLUS",
    "CP_SIZE_WRITE_AGG_0_100",      /* count of MPI write size ranges */
    "CP_SIZE_WRITE_AGG_100_1K",
    "CP_SIZE_WRITE_AGG_1K_10K",
    "CP_SIZE_WRITE_AGG_10K_100K",
    "CP_SIZE_WRITE_AGG_100K_1M",
    "CP_SIZE_WRITE_AGG_1M_4M",
    "CP_SIZE_WRITE_AGG_4M_10M",
    "CP_SIZE_WRITE_AGG_10M_100M",
    "CP_SIZE_WRITE_AGG_100M_1G",
    "CP_SIZE_WRITE_AGG_1G_PLUS",
    "CP_EXTENT_READ_0_100",          /* count of MPI read extent ranges */
    "CP_EXTENT_READ_100_1K",
    "CP_EXTENT_READ_1K_10K",
    "CP_EXTENT_READ_10K_100K",
    "CP_EXTENT_READ_100K_1M",
    "CP_EXTENT_READ_1M_4M",
    "CP_EXTENT_READ_4M_10M",
    "CP_EXTENT_READ_10M_100M",
    "CP_EXTENT_READ_100M_1G",
    "CP_EXTENT_READ_1G_PLUS",
    "CP_EXTENT_WRITE_0_100",         /* count of MPI write extent ranges */
    "CP_EXTENT_WRITE_100_1K",
    "CP_EXTENT_WRITE_1K_10K",
    "CP_EXTENT_WRITE_10K_100K",
    "CP_EXTENT_WRITE_100K_1M",
    "CP_EXTENT_WRITE_1M_4M",
    "CP_EXTENT_WRITE_4M_10M",
    "CP_EXTENT_WRITE_10M_100M",
    "CP_EXTENT_WRITE_100M_1G",
    "CP_EXTENT_WRITE_1G_PLUS",
    "CP_STRIDE1_STRIDE",             /* the four most frequently appearing strides */
    "CP_STRIDE2_STRIDE",
    "CP_STRIDE3_STRIDE",
    "CP_STRIDE4_STRIDE",
    "CP_STRIDE1_COUNT",              /* count of each of the most frequent strides */
    "CP_STRIDE2_COUNT",
    "CP_STRIDE3_COUNT",
    "CP_STRIDE4_COUNT",
    "CP_ACCESS1_ACCESS",
    "CP_ACCESS2_ACCESS",
    "CP_ACCESS3_ACCESS",
    "CP_ACCESS4_ACCESS",
    "CP_ACCESS1_COUNT",
    "CP_ACCESS2_COUNT",
    "CP_ACCESS3_COUNT",
    "CP_ACCESS4_COUNT",
154
    "CP_DEVICE",
155
    "CP_SIZE_AT_OPEN",
156
157
158
159
    "CP_FASTEST_RANK",
    "CP_FASTEST_RANK_BYTES",
    "CP_SLOWEST_RANK",
    "CP_SLOWEST_RANK_BYTES",
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177

    "CP_NUM_INDICES"
};

/* isn't there a clever c way to avoid this? */
char *darshan_f_names[] = {
    "CP_F_OPEN_TIMESTAMP",        /* timestamp of first open */
    "CP_F_READ_START_TIMESTAMP",  /* timestamp of first read */
    "CP_F_WRITE_START_TIMESTAMP", /* timestamp of first write */
    "CP_F_CLOSE_TIMESTAMP",       /* timestamp of last close */
    "CP_F_READ_END_TIMESTAMP",    /* timestamp of last read */
    "CP_F_WRITE_END_TIMESTAMP",   /* timestamp of last write */
    "CP_F_POSIX_READ_TIME",       /* cumulative posix read time */
    "CP_F_POSIX_WRITE_TIME",      /* cumulative posix write time */
    "CP_F_POSIX_META_TIME",       /* cumulative posix meta time */
    "CP_F_MPI_META_TIME",         /* cumulative mpi-io metadata time */
    "CP_F_MPI_READ_TIME",         /* cumulative mpi-io read time */
    "CP_F_MPI_WRITE_TIME",        /* cumulative mpi-io write time */
178
179
    "CP_F_MAX_READ_TIME",
    "CP_F_MAX_WRITE_TIME",
180
181
182
183
184
185
    "CP_F_FASTEST_RANK_TIME",
    "CP_F_SLOWEST_RANK_TIME",
    "CP_F_VARIANCE_RANK_TIME",
    "CP_F_VARIANCE_RANK_BYTES",

    "CP_F_NUM_INDICES"
186
187
};

188
189
190
191
192
193
194
/* function pointers so that we can switch functions depending on what file
 * version is detected
 */
int (*getjob_internal)(darshan_fd file, struct darshan_job *job);
int (*getfile_internal)(darshan_fd fd, 
    struct darshan_job *job, 
    struct darshan_file *file);
195
#define JOB_SIZE_124 28
196

197
/* internal routines for parsing different file versions */
198
199
200
201
202
203
204
205
206
207
static int getjob_internal_200(darshan_fd file, struct darshan_job *job);
static int getfile_internal_200(darshan_fd fd, struct darshan_job *job, 
    struct darshan_file *file);
static int getjob_internal_124(darshan_fd file, struct darshan_job *job);
static int getfile_internal_124(darshan_fd fd, struct darshan_job *job, 
    struct darshan_file *file);
static int getfile_internal_122(darshan_fd fd, struct darshan_job *job, 
    struct darshan_file *file);
static int getfile_internal_121(darshan_fd fd, struct darshan_job *job, 
    struct darshan_file *file);
208
209
static int getfile_internal_1x(darshan_fd fd, struct darshan_job *job, 
    struct darshan_file *file, int n_counters, int n_fcounters);
210
static void shift_missing_1_24(struct darshan_file* file);
211
212
static void shift_missing_1_22(struct darshan_file* file);
static void shift_missing_1_21(struct darshan_file* file);
213

214
/* a rather crude API for accessing raw binary darshan files */
215
darshan_fd darshan_log_open(const char *name, const char* mode)
216
{
217
218
219
220
221
222
    darshan_fd tmp_fd = malloc(sizeof(*tmp_fd));
    if(!tmp_fd)
        return(NULL);

    memset(tmp_fd, 0, sizeof(*tmp_fd));

223
    tmp_fd->gzf = gzopen(name, mode);
224
225
226
227
228
229
    if(!tmp_fd->gzf)
    {
        free(tmp_fd);
        tmp_fd = NULL;
    }
    return tmp_fd;
230
231
}

232
/* darshan_log_getjob()
233
234
235
 *
 * returns 0 on success, -1 on failure
 */
236
int darshan_log_getjob(darshan_fd file, struct darshan_job *job)
237
238
{
    int ret;
239

240
    gzseek(file->gzf, 0, SEEK_SET);
241

242
243
244
245
246
    /* read version number first so we know how to digest the rest of the
     * file
     */
    ret = gzread(file->gzf, file->version, 10);
    if(ret < 10)
247
    {
248
        fprintf(stderr, "Error: invalid log file (failed to read version).\n");
249
250
251
        return(-1);
    }

252
    if(strcmp(file->version, "2.00") == 0)
253
    {
254
255
        getjob_internal = getjob_internal_200;
        getfile_internal = getfile_internal_200;
256
        file->job_struct_size = sizeof(*job);
257
    }
258
    else if(strcmp(file->version, "1.24") == 0)
259
    {
260
261
        getjob_internal = getjob_internal_124;
        getfile_internal = getfile_internal_124;
262
        file->job_struct_size = JOB_SIZE_124;
263
    }
264
    else if(strcmp(file->version, "1.23") == 0)
265
    {
266
267
268
        /* same as 1.24, except that mnt points may be incorrect */
        getjob_internal = getjob_internal_124;
        getfile_internal = getfile_internal_124;
269
        file->job_struct_size = JOB_SIZE_124;
270
    }
271
    else if(strcmp(file->version, "1.22") == 0)
272
    {
273
274
        getjob_internal = getjob_internal_124;
        getfile_internal = getfile_internal_122;
275
        file->job_struct_size = JOB_SIZE_124;
276
    }
277
    else if(strcmp(file->version, "1.21") == 0)
278
    {
279
        getjob_internal = getjob_internal_124;
Philip Carns's avatar
Philip Carns committed
280
        getfile_internal = getfile_internal_121;
281
        file->job_struct_size = JOB_SIZE_124;
282
283
    }
    else
284
    {
285
286
287
        fprintf(stderr, "Error: incompatible darshan file.\n");
        fprintf(stderr, "Error: expected version %s, but got %s\n", 
                CP_VERSION, file->version);
288
        return(-1);
289
290
    }

291
292
293
    ret = getjob_internal(file, job);
    return(ret);
}
294

295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
/* darshan_putjob()
 * write job header in gzfile
 *
 * return 0 on success, -1 on failure.
 */
int darshan_log_putjob(darshan_fd file, struct darshan_job *job)
{
    struct darshan_job job_copy;
    z_off_t off;
    int     ret;

    off = gzseek(file->gzf, 0, SEEK_SET);
    if (off != 0)
    {
        ret = -1;
        fprintf(stderr, "Error: couldn't seek to beginning of file.\n");
        return(ret);
    }

    job_copy = *job;
    memset(job_copy.version_string, 0, sizeof(job_copy.version_string));
    strcpy(job_copy.version_string, CP_VERSION);
    job_copy.magic_nr = CP_MAGIC_NR;

    ret = gzwrite(file->gzf, &job_copy, sizeof(job_copy));
    if (ret != sizeof(job_copy))
    {
        fprintf(stderr, "Error: failed to write job header: %d\n", ret);
        return(-1);
    }

    return(0);
}

329
330
331
332
333
334
335
/* darshan_log_getfile()
 *
 * return 1 if file record found, 0 on eof, and -1 on error
 */
int darshan_log_getfile(darshan_fd fd, struct darshan_job *job, struct darshan_file *file)
{
    int ret;
336

337
338
339
    ret = getfile_internal(fd, job, file);

    return(ret);
340
341
}

342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
/* darshan_log_putfile()
 *
 * return 0 if file record written, -1 on error.
 */
int darshan_log_putfile(darshan_fd fd, struct darshan_job *job, struct darshan_file *file)
{
    z_off_t off;
    int     ret;

    if(gztell(fd->gzf) < CP_JOB_RECORD_SIZE)
    {
        off = gzseek(fd->gzf, CP_JOB_RECORD_SIZE, SEEK_SET);
        if (off == -1)
        {
            fprintf(stderr, "Error: Failed to set start of file records.\n");
            return(-1);
        }
    }

    ret = gzwrite(fd->gzf, file, sizeof(*file));
    if (ret != sizeof(*file))
    {
        fprintf(stderr, "Error: writing file record failed: %d\n", ret);
        return(-1);
    }

    return(0);
}

371
372
373
374
375
376
/* darshan_log_getmounts()
 * 
 * retrieves mount table information from the log.  Note that devs, mnt_pts,
 * and fs_types are arrays that will be allocated by the function and must
 * be freed by the caller.  count will indicate the size of the arrays
 */
377
int darshan_log_getmounts(darshan_fd fd, int64_t** devs, char*** mnt_pts, char***
378
379
380
381
382
383
384
    fs_types, int* count, int *flag)
{
    int ret;
    char* pos;
    int array_index = 0;
    char buf[CP_EXE_LEN+1];

385
    gzseek(fd->gzf, fd->job_struct_size, SEEK_SET);
386

387
    ret = gzread(fd->gzf, buf, (CP_EXE_LEN + 1));
388
389
390
391
392
    if (ret < (CP_EXE_LEN + 1))
    {
        perror("gzread");
        return(-1);
    }
393
    if (gzeof(fd->gzf))
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
        *flag = 1;
    else
        *flag = 0;

    /* count entries */
    *count = 0;
    pos = buf;
    while((pos = strchr(pos, '\n')) != NULL)
    {
        pos++;
        (*count)++;
    }

    if(*count == 0)
    {
        /* no mount entries present */
        return(0);
    }

    /* allocate output arrays */
414
    *devs = malloc((*count)*sizeof(int64_t));
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
    assert(*devs);
    *mnt_pts = malloc((*count)*sizeof(char*));
    assert(*mnt_pts);
    *fs_types = malloc((*count)*sizeof(char*));
    assert(*fs_types);
    
    /* work backwards through the table and parse each line (except for
     * first, which holds command line information)
     */
    while((pos = strrchr(buf, '\n')) != NULL)
    {
        /* overestimate string lengths */
        (*mnt_pts)[array_index] = malloc(CP_EXE_LEN);
        assert((*mnt_pts)[array_index]);
        (*fs_types)[array_index] = malloc(CP_EXE_LEN);
        assert((*fs_types)[array_index]);
        
Philip Carns's avatar
Philip Carns committed
432
        ret = sscanf(++pos, "%" PRId64 "\t%s\t%s", &(*devs)[array_index],
433
434
            (*fs_types)[array_index], (*mnt_pts)[array_index]);

435
436
437
438
439
440
441
442
443
        if(ret != 3)
        {
            fprintf(stderr, "Error: poorly formatted mount table in log file.\n");
            return(-1);
        }
        pos--;
        *pos = '\0';
        array_index++;
    }
444

445
446
447
    return (0);
}

448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
/* darshan_log_putmounts
 *
 * encode mount information back into mtab format.
 *
 * returns 0 on success, -1 on failure.
 */
int darshan_log_putmounts(darshan_fd fd, int64_t* devs, char** mnt_pts, char** fs_types, int count)
{
    z_off_t off;
    int     ret;
    char    line[1024];
    int     i;

    for(i=count-1; i>=0; i--)
    {
        sprintf(line, "\n%" PRId64 "\t%s\t%s",
                devs[i], fs_types[i], mnt_pts[i]);
        ret = gzwrite(fd->gzf, line, strlen(line));
        if (ret != strlen(line))
        {
            fprintf(stderr, "Error: failed to write mount entry: %d\n", ret);
            return(-1);
        }
    }
    
    return(0);
}
475

476
int darshan_log_getexe(darshan_fd fd, char *buf, int *flag)
477
478
{
    int ret;
479
    char* newline;
480

481
    gzseek(fd->gzf, fd->job_struct_size, SEEK_SET);
482

483
    ret = gzread(fd->gzf, buf, (CP_EXE_LEN + 1));
484
485
486
487
488
    if (ret < (CP_EXE_LEN + 1))
    {
        perror("gzread");
        return(-1);
    }
489
    if (gzeof(fd->gzf))
490
491
492
493
        *flag = 1;
    else
        *flag = 0;

494
495
496
497
498
499
500
501
    /* this call is only supposed to return the exe string, but starting in
     * log format 1.23 there could be a table of mount entry information
     * after the exe.  Look for newline character and truncate there.
     */
    newline = strchr(buf, '\n');
    if(newline)
        *newline = '\0';

502
503
504
    return (0);
}

505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
/* darshan_log_putexe()
 *
 * Write the exe string to the log.
 *
 * return 0 on success, -1 on failure.
 */
int darshan_log_putexe(darshan_fd fd, char *buf)
{
    z_off_t off;
    int     ret;
    int     len;

    off = gzseek(fd->gzf, sizeof(struct darshan_job), SEEK_SET);
    if (off != sizeof(struct darshan_job))
    {
        ret = -1;
        fprintf(stderr, "Error: failed to set position to exe section.\n");
        return(ret);
    }

    len = strlen(buf);

    ret = gzwrite(fd->gzf, buf, len);
    if (ret != len)
    {
        fprintf(stderr, "Error: failed to write exe info: %d\n", ret);
        ret = -1;
    }

    return(ret);
}

537
void darshan_log_close(darshan_fd file)
538
{
539
540
    gzclose(file->gzf);
    free(file);
541
542
}

543
544
545
546
547
548
/* darshan_log_print_version_warnings()
 *
 * Print summary of any problems with the detected log format
 */
void darshan_log_print_version_warnings(struct darshan_job *job)
{
549
550
551
552
553
554
    if(strcmp(job->version_string, "2.00") == 0)
    {
        /* current version */
        return;
    }
 
555
    if(strcmp(job->version_string, "1.24") == 0)
556
    {
557
558
559
560
561
562
563
564
565
566
        printf("# WARNING: version 1.24 log format does not support the following parameters:\n");
        printf("#   CP_FASTEST_RANK\n");
        printf("#   CP_FASTEST_RANK_BYTES\n");
        printf("#   CP_SLOWEST_RANK\n");
        printf("#   CP_SLOWEST_RANK_BYTES\n");
        printf("#   CP_F_FASTEST_RANK_TIME\n");
        printf("#   CP_F_SLOWEST_RANK_TIME\n");
        printf("#   CP_F_VARIANCE_RANK_TIME\n");
        printf("#   CP_F_VARIANCE_RANK_BYTES\n");
        printf("# It also does not store the job id in the log file.\n");
567
568
        return;
    }
569
    
570
571
    if(strcmp(job->version_string, "1.23") == 0)
    {
572
573
574
575
576
577
578
579
580
581
        printf("# WARNING: version 1.23 log format does not support the following parameters:\n");
        printf("#   CP_FASTEST_RANK\n");
        printf("#   CP_FASTEST_RANK_BYTES\n");
        printf("#   CP_SLOWEST_RANK\n");
        printf("#   CP_SLOWEST_RANK_BYTES\n");
        printf("#   CP_F_FASTEST_RANK_TIME\n");
        printf("#   CP_F_SLOWEST_RANK_TIME\n");
        printf("#   CP_F_VARIANCE_RANK_TIME\n");
        printf("#   CP_F_VARIANCE_RANK_BYTES\n");
        printf("# It also may have incorrect mount point mappings for files with rank > 0\n");
582
        printf("# It also does not store the job id in the log file.\n");
583
        return;
584
585
    }

586
587
588
589
    if(strcmp(job->version_string, "1.22") == 0)
    {
        printf("# WARNING: version 1.22 log format does not support the following parameters:\n");
        printf("#   CP_DEVICE\n");
590
        printf("#   CP_SIZE_AT_OPEN\n");
591
592
593
594
595
596
597
598
        printf("#   CP_FASTEST_RANK\n");
        printf("#   CP_FASTEST_RANK_BYTES\n");
        printf("#   CP_SLOWEST_RANK\n");
        printf("#   CP_SLOWEST_RANK_BYTES\n");
        printf("#   CP_F_FASTEST_RANK_TIME\n");
        printf("#   CP_F_SLOWEST_RANK_TIME\n");
        printf("#   CP_F_VARIANCE_RANK_TIME\n");
        printf("#   CP_F_VARIANCE_RANK_BYTES\n");
599
        printf("# It does not record mounted file systems, mount points, or fs types.\n");
600
601
        printf("# It attributes syncs to cumulative metadata time, rather than cumulative write time.\n");
        printf("# It also does not store the job id in the log file.\n");
602
603
        return;
    }
604
605
606
607
608
609
610
611
612

    if(strcmp(job->version_string, "1.21") == 0)
    {
        printf("# WARNING: version 1.21 log format does not support the following parameters:\n");
        printf("#   CP_INDEP_NC_OPENS\n");
        printf("#   CP_COLL_NC_OPENS\n");
        printf("#   CP_HDF5_OPENS\n");
        printf("#   CP_MAX_READ_TIME_SIZE\n");
        printf("#   CP_MAX_WRITE_TIME_SIZE\n");
613
        printf("#   CP_DEVICE\n");
614
        printf("#   CP_SIZE_AT_OPEN\n");
615
616
        printf("#   CP_F_MAX_READ_TIME\n");
        printf("#   CP_F_MAX_WRITE_TIME\n");
617
618
619
620
621
622
623
624
        printf("#   CP_FASTEST_RANK\n");
        printf("#   CP_FASTEST_RANK_BYTES\n");
        printf("#   CP_SLOWEST_RANK\n");
        printf("#   CP_SLOWEST_RANK_BYTES\n");
        printf("#   CP_F_FASTEST_RANK_TIME\n");
        printf("#   CP_F_SLOWEST_RANK_TIME\n");
        printf("#   CP_F_VARIANCE_RANK_TIME\n");
        printf("#   CP_F_VARIANCE_RANK_BYTES\n");
625
626
        printf("# It does not record mounted file systems, mount points, or fs types.\n");
        printf("# It attributes syncs to cumulative metadata time, rather than cumulative write time.\n");
627
        printf("#\n");
628
        printf("# It also does not store the job id in the file.\n");
629
630
631
632
633
634
635
636
        return;
    }

    fprintf(stderr, "Error: version %s not supported by parser.\n",
        job->version_string);
    return;
}

637
638
639
640
/* shift_missing_1_21()
 *
 * translates indices to account for counters that weren't present in log
 * format 1.21
641
 */
642
/*******************************
Philip Carns's avatar
Philip Carns committed
643
 * version 1.21 to 2.00 differences 
644
645
646
647
648
649
650
651
 * - added:
 *   - CP_INDEP_NC_OPENS
 *   - CP_COLL_NC_OPENS
 *   - CP_HDF5_OPENS
 *   - CP_MAX_READ_TIME_SIZE
 *   - CP_MAX_WRITE_TIME_SIZE
 *   - CP_DEVICE
 *   - CP_SIZE_AT_OPEN
652
653
654
655
 *   - CP_FASTEST_RANK
 *   - CP_FASTEST_RANK_BYTES
 *   - CP_SLOWEST_RANK
 *   - CP_SLOWEST_RANK_BYTES
656
657
 *   - CP_F_MAX_READ_TIME
 *   - CP_F_MAX_WRITE_TIME
658
659
660
661
 *   - CP_F_FASTEST_RANK_TIME
 *   - CP_F_SLOWEST_RANK_TIME
 *   - CP_F_VARIANCE_RANK_TIME
 *   - CP_F_VARIANCE_RANK_BYTES
662
 * - changed params:
663
664
665
 *   - CP_FILE_RECORD_SIZE: 1184 to 1328
 *   - CP_NUM_INDICES: 133 to 144
 *   - CP_F_NUM_INDICES: 12 to 18
666
 */
667
668
669
670
671
672
673
674
675
static void shift_missing_1_21(struct darshan_file* file)
{
    int c_index = 0;
    int missing_counters[] = {
        CP_INDEP_NC_OPENS,
        CP_COLL_NC_OPENS,
        CP_HDF5_OPENS,
        CP_MAX_READ_TIME_SIZE,
        CP_MAX_WRITE_TIME_SIZE,
676
        CP_DEVICE,
677
        CP_SIZE_AT_OPEN,
678
679
680
681
        CP_FASTEST_RANK,
        CP_FASTEST_RANK_BYTES,
        CP_SLOWEST_RANK,
        CP_SLOWEST_RANK_BYTES,
682
683
684
685
        -1};
    int missing_f_counters[] = {
        CP_F_MAX_READ_TIME,
        CP_F_MAX_WRITE_TIME,
686
687
688
689
        CP_F_FASTEST_RANK_TIME,
        CP_F_SLOWEST_RANK_TIME,
        CP_F_VARIANCE_RANK_TIME,
        CP_F_VARIANCE_RANK_BYTES,
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
        -1};

    c_index = 0;
    while(missing_counters[c_index] != -1)
    {
        int missing_counter = missing_counters[c_index];
        c_index++;
        if(missing_counter < (CP_NUM_INDICES - 1))
        {
            /* shift down */
            memmove(&file->counters[missing_counter+1],
                &file->counters[missing_counter],
                (CP_NUM_INDICES-missing_counter-1)*sizeof(int64_t));
        }
        /* zero out missing counter */
        file->counters[missing_counter] = 0;
    }

    c_index = 0;
    while(missing_f_counters[c_index] != -1)
    {
        int missing_counter = missing_f_counters[c_index];
        c_index++;
        if(missing_counter < (CP_F_NUM_INDICES - 1))
        {
            /* shift down */
            memmove(&file->fcounters[missing_counter+1],
                &file->fcounters[missing_counter],
                (CP_F_NUM_INDICES-missing_counter-1)*sizeof(double));
        }
        /* zero out missing counter */
        file->fcounters[missing_counter] = 0;
    }

    return;
}
726
727
728
729
730
731

/* shift_missing_1_22()
 *
 * translates indices to account for counters that weren't present in log
 * format 1.22
 */
732
/*******************************
Philip Carns's avatar
Philip Carns committed
733
 * version 1.22 to 2.00 differences
734
735
736
737
 *
 * - added:
 *   - CP_DEVICE
 *   - CP_SIZE_AT_OPEN
738
739
740
741
742
743
744
745
 *   - CP_FASTEST_RANK
 *   - CP_FASTEST_RANK_BYTES
 *   - CP_SLOWEST_RANK
 *   - CP_SLOWEST_RANK_BYTES
 *   - CP_F_FASTEST_RANK_TIME
 *   - CP_F_SLOWEST_RANK_TIME
 *   - CP_F_VARIANCE_RANK_TIME
 *   - CP_F_VARIANCE_RANK_BYTES
746
 * - changed params:
747
748
749
 *   - CP_FILE_RECORD_SIZE: 1240 to 1328
 *   - CP_NUM_INDICES: 138 to 144
 *   - CP_F_NUM_INDICES: 14 to 18
750
 */
751
752
753
754
755
static void shift_missing_1_22(struct darshan_file* file)
{
    int c_index = 0;
    int missing_counters[] = {
        CP_DEVICE,
756
        CP_SIZE_AT_OPEN,
757
758
759
760
761
762
763
764
765
766
        CP_FASTEST_RANK,
        CP_FASTEST_RANK_BYTES,
        CP_SLOWEST_RANK,
        CP_SLOWEST_RANK_BYTES,
        -1};
    int missing_f_counters[] = {
        CP_F_FASTEST_RANK_TIME,
        CP_F_SLOWEST_RANK_TIME,
        CP_F_VARIANCE_RANK_TIME,
        CP_F_VARIANCE_RANK_BYTES,
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
        -1};

    c_index = 0;
    while(missing_counters[c_index] != -1)
    {
        int missing_counter = missing_counters[c_index];
        c_index++;
        if(missing_counter < (CP_NUM_INDICES - 1))
        {
            /* shift down */
            memmove(&file->counters[missing_counter+1],
                &file->counters[missing_counter],
                (CP_NUM_INDICES-missing_counter-1)*sizeof(int64_t));
        }
        /* zero out missing counter */
        file->counters[missing_counter] = 0;
    }

785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
    c_index = 0;
    while(missing_f_counters[c_index] != -1)
    {
        int missing_counter = missing_f_counters[c_index];
        c_index++;
        if(missing_counter < (CP_F_NUM_INDICES - 1))
        {
            /* shift down */
            memmove(&file->fcounters[missing_counter+1],
                &file->fcounters[missing_counter],
                (CP_F_NUM_INDICES-missing_counter-1)*sizeof(double));
        }
        /* zero out missing counter */
        file->fcounters[missing_counter] = 0;
    }

801
802
    return;
}
803

804
805
806
807
808
809
810
811
812
/* shift_missing_1_24()
 *
 * translates indices to account for counters that weren't present in log
 * format 1.24
 */
/*******************************
 * version 1.24 to 2.00 differences
 *
 * - added:
813
814
815
816
 *   - CP_FASTEST_RANK
 *   - CP_FASTEST_RANK_BYTES
 *   - CP_SLOWEST_RANK
 *   - CP_SLOWEST_RANK_BYTES
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
 *   - CP_F_FASTEST_RANK_TIME
 *   - CP_F_SLOWEST_RANK_TIME
 *   - CP_F_VARIANCE_RANK_TIME
 *   - CP_F_VARIANCE_RANK_BYTES
 * - changed params:
 *   - CP_FILE_RECORD_SIZE: ? to 1328
 *   - CP_NUM_INDICES: 140 to 144
 *   - CP_F_NUM_INDICES: 14 to 18
 */
static void shift_missing_1_24(struct darshan_file* file)
{
    int c_index = 0;
    int missing_counters[] = {
        CP_FASTEST_RANK,
        CP_FASTEST_RANK_BYTES,
        CP_SLOWEST_RANK,
        CP_SLOWEST_RANK_BYTES,
        -1};
    int missing_f_counters[] = {
        CP_F_FASTEST_RANK_TIME,
        CP_F_SLOWEST_RANK_TIME,
        CP_F_VARIANCE_RANK_TIME,
        CP_F_VARIANCE_RANK_BYTES,
        -1};

    c_index = 0;
    while(missing_counters[c_index] != -1)
    {
        int missing_counter = missing_counters[c_index];
        c_index++;
        if(missing_counter < (CP_NUM_INDICES - 1))
        {
            /* shift down */
            memmove(&file->counters[missing_counter+1],
                &file->counters[missing_counter],
                (CP_NUM_INDICES-missing_counter-1)*sizeof(int64_t));
        }
        /* zero out missing counter */
        file->counters[missing_counter] = 0;
    }

    c_index = 0;
    while(missing_f_counters[c_index] != -1)
    {
        int missing_counter = missing_f_counters[c_index];
        c_index++;
        if(missing_counter < (CP_F_NUM_INDICES - 1))
        {
            /* shift down */
            memmove(&file->fcounters[missing_counter+1],
                &file->fcounters[missing_counter],
                (CP_F_NUM_INDICES-missing_counter-1)*sizeof(double));
        }
        /* zero out missing counter */
        file->fcounters[missing_counter] = 0;
    }

    return;
}


878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
static int getjob_internal_200(darshan_fd file, struct darshan_job *job)
{
    int ret;

    gzseek(file->gzf, 0, SEEK_SET);

    ret = gzread(file->gzf, job, sizeof(*job));
    if (ret < sizeof(*job))
    {
        if(gzeof(file->gzf))
        {
            fprintf(stderr, "Error: invalid log file (too short).\n");
        }
        perror("darshan_job_init");
        return(-1);
    }

    if(job->magic_nr == CP_MAGIC_NR)
    {
        /* no byte swapping needed, this file is in host format already */
        file->swap_flag = 0;
        return(0);
    }

    /* try byte swapping */
    DARSHAN_BSWAP64(&job->magic_nr);
    if(job->magic_nr == CP_MAGIC_NR)
    {
        file->swap_flag = 1;
        DARSHAN_BSWAP64(&job->uid);
        DARSHAN_BSWAP64(&job->start_time);
        DARSHAN_BSWAP64(&job->end_time);
        DARSHAN_BSWAP64(&job->nprocs);
911
        DARSHAN_BSWAP64(&job->jobid);
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
        return(0);
    }

    /* otherwise this file is just broken */
    fprintf(stderr, "Error: bad magic number in darshan file.\n");
    return(-1);
}

static int getfile_internal_200(darshan_fd fd, struct darshan_job *job, 
    struct darshan_file *file)
{
    int ret;
    const char* err_string;
    int i;
    
    if(gztell(fd->gzf) < CP_JOB_RECORD_SIZE)
        gzseek(fd->gzf, CP_JOB_RECORD_SIZE, SEEK_SET);

    /* reset file record, so that diff compares against a zero'd out record
     * if file is missing
     */
    memset(file, 0, sizeof(&file));

    ret = gzread(fd->gzf, file, sizeof(*file));
    if(ret == sizeof(*file))
    {
        /* got exactly one, correct size record */
        if(fd->swap_flag)
        {
            /* swap bytes if necessary */
            DARSHAN_BSWAP64(&file->hash);
            DARSHAN_BSWAP64(&file->rank);
            for(i=0; i<CP_NUM_INDICES; i++)
                DARSHAN_BSWAP64(&file->counters[i]);
            for(i=0; i<CP_F_NUM_INDICES; i++)
                DARSHAN_BSWAP64(&file->fcounters[i]);
        }
        return(1);
    }


    if(ret > 0)
    {
        /* got a short read */
        fprintf(stderr, "Error: invalid file record (too small)\n");
        return(-1);
    }

    if(ret == 0 && gzeof(fd->gzf))
    {
        /* hit end of file */
        return(0);
    }

    /* all other errors */
    err_string = gzerror(fd->gzf, &ret);
    fprintf(stderr, "Error: %s\n", err_string);
    return(-1);
}

972
973
974
/* If we see version 1.24, assume that it is stored in big endian 32 bit
 * format.  Convert up to current format.
 */
975
976
static int getjob_internal_124(darshan_fd fd, struct darshan_job *job)
{
977
978
979
980
981
982
983
    char* buffer;
    int ret;
    uint32_t uid;
    int32_t start_time;
    int32_t end_time;
    int32_t nprocs;

984
985
986
987
988
989
#ifdef WORDS_BIGENDIAN
    fd->swap_flag = 0;
#else
    fd->swap_flag = 1;
#endif

990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
    memset(job, 0, sizeof(*job));

    buffer = (char*)malloc(JOB_SIZE_124);
    if(!buffer)
    {
        return(-1);
    }

    gzseek(fd->gzf, 0, SEEK_SET);

    ret = gzread(fd->gzf, buffer, JOB_SIZE_124);
    if (ret < JOB_SIZE_124)
    {
        fprintf(stderr, "Error: invalid log file (could not read file record).\n");
        free(buffer);
        return(-1);
    }

    /* pull job header information out of specific bytes in case struct
     * padding is off
     */
    strncpy(job->version_string, buffer, 8);
    uid = *((uint32_t*)&buffer[12]);
    start_time = *((int32_t*)&buffer[16]);
    end_time = *((int32_t*)&buffer[20]);
    nprocs = *((int32_t*)&buffer[24]);

    free(buffer);

1019
1020
1021
1022
1023
1024
1025
1026
    if(fd->swap_flag)
    {
        /* byte swap */
        DARSHAN_BSWAP32(&uid);
        DARSHAN_BSWAP32(&start_time);
        DARSHAN_BSWAP32(&end_time);
        DARSHAN_BSWAP32(&nprocs);
    }
1027
1028
1029
1030
1031

    job->uid += uid;
    job->start_time += start_time;
    job->end_time += end_time;
    job->nprocs += nprocs;
1032
    job->jobid = 0; /* old log versions did not have this field */
1033
1034
1035
1036
1037
    
    /* set magic number */
    job->magic_nr = CP_MAGIC_NR;

    return(0);
1038
1039
1040
1041
}

static int getfile_internal_124(darshan_fd fd, struct darshan_job *job, 
    struct darshan_file *file)
1042
{
1043
1044
    int ret;

1045
    ret = getfile_internal_1x(fd, job, file, 140, 14);
1046
1047
1048
1049
1050
1051
    if(ret <= 0)
        return(ret);

    shift_missing_1_24(file);

    return(1);
1052
1053
1054
1055
1056
1057
1058
}

static int getfile_internal_122(darshan_fd fd, struct darshan_job *job, 
    struct darshan_file *file)
{
    int ret;

1059
1060
    ret = getfile_internal_1x(fd, job, file, 138, 14);
    if(ret <= 0)
1061
1062
1063
1064
        return(ret);

    shift_missing_1_22(file);

1065
    return(1);
1066
1067
1068
1069
1070
1071
1072
}

static int getfile_internal_121(darshan_fd fd, struct darshan_job *job, 
    struct darshan_file *file)
{
    int ret;

1073
1074
    ret = getfile_internal_1x(fd, job, file, 133, 12);
    if(ret <= 0)
1075
1076
1077
1078
        return(ret);

    shift_missing_1_21(file);

1079
    return(1);
1080
1081
1082
1083
}

static int getfile_internal_1x(darshan_fd fd, struct darshan_job *job, 
    struct darshan_file *file, int n_counters, int n_fcounters)
1084
{
1085
1086
1087
1088
1089
1090
1091
1092
1093
    char* buffer;
    int ret;
    const char* err_string;
    int i;
    uint64_t hash;
    int32_t rank;
    int64_t* counters;
    double* fcounters;
    char* name_suffix;
1094
    int FILE_SIZE_1x = (32 + n_counters*8 + n_fcounters*8);
1095

1096
1097
1098
1099
1100
1101
1102
1103
1104
    memset(file, 0, sizeof(*file));

    /* set file pointer if this is the first file record; otherwise pick up
     * where we left off last time
     */
    if(gztell(fd->gzf) < CP_JOB_RECORD_SIZE)
        gzseek(fd->gzf, CP_JOB_RECORD_SIZE, SEEK_SET);

    /* space for file struct, int64 array, and double array */
1105
    buffer = (char*)malloc(FILE_SIZE_1x);
1106
1107
1108
1109
1110
    if(!buffer)
    {
        return(-1);
    }

1111
    ret = gzread(fd->gzf, buffer, FILE_SIZE_1x);
1112

1113
    if(ret > 0 && ret < FILE_SIZE_1x)
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
    {
        /* got a short read */
        fprintf(stderr, "Error: invalid file record (too small)\n");
        free(buffer);
        return(-1);
    }
    else if(ret == 0 && gzeof(fd->gzf))
    {
        /* hit end of file */
        free(buffer);
        return(0);
    }
    else if(ret <= 0)
    {
        /* all other errors */
        err_string = gzerror(fd->gzf, &ret);
        fprintf(stderr, "Error: %s\n", err_string);
        free(buffer);
        return(-1);
    }

    /* got exactly one, correct size record */
    hash = *((int64_t*)&buffer[0]);
    rank = *((int32_t*)&buffer[8]);
    counters = ((int64_t*)&buffer[16]);
1139
1140
    fcounters = ((double*)&buffer[16 + n_counters*8]);
    name_suffix = &buffer[16 + n_counters*8 + n_fcounters*8];
1141
1142
1143
1144
1145
1146
1147


    if(fd->swap_flag)
    {
        /* swap bytes if necessary */
        DARSHAN_BSWAP64(&hash);
        DARSHAN_BSWAP32(&rank);
1148
        for(i=0; i<n_counters; i++)
1149
            DARSHAN_BSWAP64(&counters[i]);
1150
        for(i=0; i<n_fcounters; i++)
1151
1152
1153
1154
1155
1156
            DARSHAN_BSWAP64(&fcounters[i]);
    }

    /* assign into new format */
    file->hash = hash;
    file->rank += rank;
1157
1158
    memcpy(file->counters, counters, n_counters*8);
    memcpy(file->fcounters, fcounters, n_fcounters*8);
1159
1160
1161
1162
    memcpy(file->name_suffix, name_suffix, 12);

    free(buffer);
    return(1);
1163
1164
1165
1166
}