darshan-logutils.c 30.4 KB
Newer Older
1
2
3
4
5
/*
 *  (C) 2009 by Argonne National Laboratory.
 *      See COPYRIGHT in top-level directory.
 */

6
7
#include <stdio.h>
#include <string.h>
8
#include <assert.h>
9
#include <stdlib.h>
10
#include "darshan-logutils.h"
11
#include "darshan-config.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

/* 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",
38
39
40
    "CP_INDEP_NC_OPENS",
    "CP_COLL_NC_OPENS",
    "CP_HDF5_OPENS",
41
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
    "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 */
75
76
    "CP_MAX_READ_TIME_SIZE",
    "CP_MAX_WRITE_TIME_SIZE",
77
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
    "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",
153
    "CP_DEVICE",
154
    "CP_SIZE_AT_OPEN",
155
156
157
158
    "CP_FASTEST_RANK",
    "CP_FASTEST_RANK_BYTES",
    "CP_SLOWEST_RANK",
    "CP_SLOWEST_RANK_BYTES",
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176

    "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 */
177
178
    "CP_F_MAX_READ_TIME",
    "CP_F_MAX_WRITE_TIME",
179
180
181
182
183
184
    "CP_F_FASTEST_RANK_TIME",
    "CP_F_SLOWEST_RANK_TIME",
    "CP_F_VARIANCE_RANK_TIME",
    "CP_F_VARIANCE_RANK_BYTES",

    "CP_F_NUM_INDICES"
185
186
};

187
188
189
190
191
192
193
/* 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);
194
#define JOB_SIZE_124 28
195

196
/* internal routines for parsing different file versions */
197
198
199
200
201
202
203
204
205
206
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);
207
208
static int getfile_internal_1x(darshan_fd fd, struct darshan_job *job, 
    struct darshan_file *file, int n_counters, int n_fcounters);
209
static void shift_missing_1_24(struct darshan_file* file);
210
211
static void shift_missing_1_22(struct darshan_file* file);
static void shift_missing_1_21(struct darshan_file* file);
212

213
/* a rather crude API for accessing raw binary darshan files */
214
darshan_fd darshan_log_open(const char *name)
215
{
216
217
218
219
220
221
222
223
224
225
226
227
228
    darshan_fd tmp_fd = malloc(sizeof(*tmp_fd));
    if(!tmp_fd)
        return(NULL);

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

    tmp_fd->gzf = gzopen(name, "r");
    if(!tmp_fd->gzf)
    {
        free(tmp_fd);
        tmp_fd = NULL;
    }
    return tmp_fd;
229
230
}

231
/* darshan_log_getjob()
232
233
234
 *
 * returns 0 on success, -1 on failure
 */
235
int darshan_log_getjob(darshan_fd file, struct darshan_job *job)
236
237
{
    int ret;
238

239
    gzseek(file->gzf, 0, SEEK_SET);
240

241
242
243
244
245
    /* 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)
246
    {
247
        fprintf(stderr, "Error: invalid log file (failed to read version).\n");
248
249
250
        return(-1);
    }

251
    if(strcmp(file->version, "2.00") == 0)
252
    {
253
254
        getjob_internal = getjob_internal_200;
        getfile_internal = getfile_internal_200;
255
        file->job_struct_size = sizeof(*job);
256
    }
257
    else if(strcmp(file->version, "1.24") == 0)
258
    {
259
260
        getjob_internal = getjob_internal_124;
        getfile_internal = getfile_internal_124;
261
        file->job_struct_size = JOB_SIZE_124;
262
    }
263
    else if(strcmp(file->version, "1.23") == 0)
264
    {
265
266
267
        /* same as 1.24, except that mnt points may be incorrect */
        getjob_internal = getjob_internal_124;
        getfile_internal = getfile_internal_124;
268
        file->job_struct_size = JOB_SIZE_124;
269
    }
270
    else if(strcmp(file->version, "1.22") == 0)
271
    {
272
273
        getjob_internal = getjob_internal_124;
        getfile_internal = getfile_internal_122;
274
        file->job_struct_size = JOB_SIZE_124;
275
    }
276
    else if(strcmp(file->version, "1.21") == 0)
277
    {
278
        getjob_internal = getjob_internal_124;
Philip Carns's avatar
Philip Carns committed
279
        getfile_internal = getfile_internal_121;
280
        file->job_struct_size = JOB_SIZE_124;
281
282
    }
    else
283
    {
284
285
286
        fprintf(stderr, "Error: incompatible darshan file.\n");
        fprintf(stderr, "Error: expected version %s, but got %s\n", 
                CP_VERSION, file->version);
287
        return(-1);
288
289
    }

290
291
292
    ret = getjob_internal(file, job);
    return(ret);
}
293

294
295
296
297
298
299
300
/* 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;
301

302
303
304
    ret = getfile_internal(fd, job, file);

    return(ret);
305
306
}

307
308
309
310
311
312
/* 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
 */
313
int darshan_log_getmounts(darshan_fd fd, int64_t** devs, char*** mnt_pts, char***
314
315
316
317
318
319
320
    fs_types, int* count, int *flag)
{
    int ret;
    char* pos;
    int array_index = 0;
    char buf[CP_EXE_LEN+1];

321
    gzseek(fd->gzf, fd->job_struct_size, SEEK_SET);
322

323
    ret = gzread(fd->gzf, buf, (CP_EXE_LEN + 1));
324
325
326
327
328
    if (ret < (CP_EXE_LEN + 1))
    {
        perror("gzread");
        return(-1);
    }
329
    if (gzeof(fd->gzf))
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
        *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 */
350
    *devs = malloc((*count)*sizeof(int64_t));
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
    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]);
        
368
369
#if SIZEOF_LONG_INT == 4
        ret = sscanf(++pos, "%lld\t%s\t%s", &(*devs)[array_index],
370
            (*fs_types)[array_index], (*mnt_pts)[array_index]);
371
372
373
374
375
376
377
#elif SIZEOF_LONG_INT == 8
        ret = sscanf(++pos, "%ld\t%s\t%s", &(*devs)[array_index],
            (*fs_types)[array_index], (*mnt_pts)[array_index]);
#else
#  error Unexpected sizeof(long int)
#endif

378
379
380
381
382
383
384
385
386
387
388
389
390
391
        if(ret != 3)
        {
            fprintf(stderr, "Error: poorly formatted mount table in log file.\n");
            return(-1);
        }
        pos--;
        *pos = '\0';
        array_index++;
    }
   
    return (0);
}


392
int darshan_log_getexe(darshan_fd fd, char *buf, int *flag)
393
394
{
    int ret;
395
    char* newline;
396

397
    gzseek(fd->gzf, fd->job_struct_size, SEEK_SET);
398

399
    ret = gzread(fd->gzf, buf, (CP_EXE_LEN + 1));
400
401
402
403
404
    if (ret < (CP_EXE_LEN + 1))
    {
        perror("gzread");
        return(-1);
    }
405
    if (gzeof(fd->gzf))
406
407
408
409
        *flag = 1;
    else
        *flag = 0;

410
411
412
413
414
415
416
417
    /* 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';

418
419
420
    return (0);
}

421
void darshan_log_close(darshan_fd file)
422
{
423
424
    gzclose(file->gzf);
    free(file);
425
426
}

427
428
429
430
431
432
/* 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)
{
433
434
435
436
437
438
    if(strcmp(job->version_string, "2.00") == 0)
    {
        /* current version */
        return;
    }
 
439
    if(strcmp(job->version_string, "1.24") == 0)
440
    {
441
442
443
444
445
446
447
448
449
450
        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");
451
452
        return;
    }
453
    
454
455
    if(strcmp(job->version_string, "1.23") == 0)
    {
456
457
458
459
460
461
462
463
464
465
        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");
466
        printf("# It also does not store the job id in the log file.\n");
467
        return;
468
469
    }

470
471
472
473
    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");
474
        printf("#   CP_SIZE_AT_OPEN\n");
475
476
477
478
479
480
481
482
        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");
483
        printf("# It does not record mounted file systems, mount points, or fs types.\n");
484
485
        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");
486
487
        return;
    }
488
489
490
491
492
493
494
495
496

    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");
497
        printf("#   CP_DEVICE\n");
498
        printf("#   CP_SIZE_AT_OPEN\n");
499
500
        printf("#   CP_F_MAX_READ_TIME\n");
        printf("#   CP_F_MAX_WRITE_TIME\n");
501
502
503
504
505
506
507
508
        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");
509
510
        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");
511
        printf("#\n");
512
        printf("# It also does not store the job id in the file.\n");
513
514
515
516
517
518
519
520
        return;
    }

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

521
522
523
524
/* shift_missing_1_21()
 *
 * translates indices to account for counters that weren't present in log
 * format 1.21
525
 */
526
/*******************************
Philip Carns's avatar
Philip Carns committed
527
 * version 1.21 to 2.00 differences 
528
529
530
531
532
533
534
535
 * - 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
536
537
538
539
 *   - CP_FASTEST_RANK
 *   - CP_FASTEST_RANK_BYTES
 *   - CP_SLOWEST_RANK
 *   - CP_SLOWEST_RANK_BYTES
540
541
 *   - CP_F_MAX_READ_TIME
 *   - CP_F_MAX_WRITE_TIME
542
543
544
545
 *   - CP_F_FASTEST_RANK_TIME
 *   - CP_F_SLOWEST_RANK_TIME
 *   - CP_F_VARIANCE_RANK_TIME
 *   - CP_F_VARIANCE_RANK_BYTES
546
 * - changed params:
547
548
549
 *   - CP_FILE_RECORD_SIZE: 1184 to 1328
 *   - CP_NUM_INDICES: 133 to 144
 *   - CP_F_NUM_INDICES: 12 to 18
550
 */
551
552
553
554
555
556
557
558
559
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,
560
        CP_DEVICE,
561
        CP_SIZE_AT_OPEN,
562
563
564
565
        CP_FASTEST_RANK,
        CP_FASTEST_RANK_BYTES,
        CP_SLOWEST_RANK,
        CP_SLOWEST_RANK_BYTES,
566
567
568
569
        -1};
    int missing_f_counters[] = {
        CP_F_MAX_READ_TIME,
        CP_F_MAX_WRITE_TIME,
570
571
572
573
        CP_F_FASTEST_RANK_TIME,
        CP_F_SLOWEST_RANK_TIME,
        CP_F_VARIANCE_RANK_TIME,
        CP_F_VARIANCE_RANK_BYTES,
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
        -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;
}
610
611
612
613
614
615

/* shift_missing_1_22()
 *
 * translates indices to account for counters that weren't present in log
 * format 1.22
 */
616
/*******************************
Philip Carns's avatar
Philip Carns committed
617
 * version 1.22 to 2.00 differences
618
619
620
621
 *
 * - added:
 *   - CP_DEVICE
 *   - CP_SIZE_AT_OPEN
622
623
624
625
626
627
628
629
 *   - 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
630
 * - changed params:
631
632
633
 *   - CP_FILE_RECORD_SIZE: 1240 to 1328
 *   - CP_NUM_INDICES: 138 to 144
 *   - CP_F_NUM_INDICES: 14 to 18
634
 */
635
636
637
638
639
static void shift_missing_1_22(struct darshan_file* file)
{
    int c_index = 0;
    int missing_counters[] = {
        CP_DEVICE,
640
        CP_SIZE_AT_OPEN,
641
642
643
644
645
646
647
648
649
650
        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,
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
        -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;
    }

669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
    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;
    }

685
686
    return;
}
687

688
689
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
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
/* 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:
 *   - 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;
}


758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
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);
791
        DARSHAN_BSWAP64(&job->jobid);
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
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
        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);
}

852
853
854
/* If we see version 1.24, assume that it is stored in big endian 32 bit
 * format.  Convert up to current format.
 */
855
856
static int getjob_internal_124(darshan_fd fd, struct darshan_job *job)
{
857
858
859
860
861
862
863
    char* buffer;
    int ret;
    uint32_t uid;
    int32_t start_time;
    int32_t end_time;
    int32_t nprocs;

864
865
866
867
868
869
#ifdef WORDS_BIGENDIAN
    fd->swap_flag = 0;
#else
    fd->swap_flag = 1;
#endif

870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
    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);

899
900
901
902
903
904
905
906
    if(fd->swap_flag)
    {
        /* byte swap */
        DARSHAN_BSWAP32(&uid);
        DARSHAN_BSWAP32(&start_time);
        DARSHAN_BSWAP32(&end_time);
        DARSHAN_BSWAP32(&nprocs);
    }
907
908
909
910
911

    job->uid += uid;
    job->start_time += start_time;
    job->end_time += end_time;
    job->nprocs += nprocs;
912
    job->jobid = 0; /* old log versions did not have this field */
913
914
915
916
917
    
    /* set magic number */
    job->magic_nr = CP_MAGIC_NR;

    return(0);
918
919
920
921
}

static int getfile_internal_124(darshan_fd fd, struct darshan_job *job, 
    struct darshan_file *file)
922
{
923
924
925
926
927
928
929
930
931
    int ret;

    ret = getfile_internal_1x(fd, job, file, 138, 14);
    if(ret <= 0)
        return(ret);

    shift_missing_1_24(file);

    return(1);
932
933
934
935
936
937
938
}

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

939
940
    ret = getfile_internal_1x(fd, job, file, 138, 14);
    if(ret <= 0)
941
942
943
944
        return(ret);

    shift_missing_1_22(file);

945
    return(1);
946
947
948
949
950
951
952
}

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

953
954
    ret = getfile_internal_1x(fd, job, file, 133, 12);
    if(ret <= 0)
955
956
957
958
        return(ret);

    shift_missing_1_21(file);

959
    return(1);
960
961
962
963
}

static int getfile_internal_1x(darshan_fd fd, struct darshan_job *job, 
    struct darshan_file *file, int n_counters, int n_fcounters)
964
{
965
966
967
968
969
970
971
972
973
    char* buffer;
    int ret;
    const char* err_string;
    int i;
    uint64_t hash;
    int32_t rank;
    int64_t* counters;
    double* fcounters;
    char* name_suffix;
974
    int FILE_SIZE_1x = (32 + n_counters*8 + n_fcounters*8);
975

976
977
978
979
980
981
982
983
984
    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 */
985
    buffer = (char*)malloc(FILE_SIZE_1x);
986
987
988
989
990
    if(!buffer)
    {
        return(-1);
    }

991
    ret = gzread(fd->gzf, buffer, FILE_SIZE_1x);
992

993
    if(ret > 0 && ret < FILE_SIZE_1x)
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
    {
        /* 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]);
1019
1020
    fcounters = ((double*)&buffer[16 + n_counters*8]);
    name_suffix = &buffer[16 + n_counters*8 + n_fcounters*8];
1021
1022
1023
1024
1025
1026
1027


    if(fd->swap_flag)
    {
        /* swap bytes if necessary */
        DARSHAN_BSWAP64(&hash);
        DARSHAN_BSWAP32(&rank);
1028
        for(i=0; i<n_counters; i++)
1029
            DARSHAN_BSWAP64(&counters[i]);
1030
        for(i=0; i<n_fcounters; i++)
1031
1032
1033
1034
1035
1036
            DARSHAN_BSWAP64(&fcounters[i]);
    }

    /* assign into new format */
    file->hash = hash;
    file->rank += rank;
1037
1038
    memcpy(file->counters, counters, n_counters*8);
    memcpy(file->fcounters, fcounters, n_fcounters*8);
1039
1040
1041
1042
    memcpy(file->name_suffix, name_suffix, 12);

    free(buffer);
    return(1);
1043
1044
1045
1046
}