mpid_rma_shm.h 23.8 KB
Newer Older
1
2
3
4
5
6
7
8
9
10
11
12
/* -*- Mode: C; c-basic-offset:4 ; indent-tabs-mode:nil ; -*- */
/*
 *  (C) 2001 by Argonne National Laboratory.
 *      See COPYRIGHT in top-level directory.
 */

#if !defined(MPID_RMA_SHM_H_INCLUDED)
#define MPID_RMA_SHM_H_INCLUDED

#include "mpl_utlist.h"
#include "mpid_rma_types.h"

13
14
15
16
17
18
19
/* define ACC stream size as the SRBuf size */
#define MPIDI_CH3U_Acc_stream_size MPIDI_CH3U_SRBuf_size

static inline int do_accumulate_op(void *source_buf, int source_count, MPI_Datatype source_dtp,
                                   void *target_buf, int target_count, MPI_Datatype target_dtp,
                                   MPI_Aint stream_offset, MPI_Op acc_op);

20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
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
75
76
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
153
154
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
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
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
#define ASSIGN_COPY(src, dest, count, type)     \
    {                                           \
        type *src_ = (type *) src;              \
        type *dest_ = (type *) dest;            \
        int i;                                  \
        for (i = 0; i < count; i++)             \
            dest_[i] = src_[i];                 \
        goto fn_exit;                           \
    }

static inline int shm_copy(const void *src, int scount, MPI_Datatype stype,
                           void *dest, int dcount, MPI_Datatype dtype)
{
    int mpi_errno = MPI_SUCCESS;

    /* We use a threshold of operations under which a for loop of assignments is
     * used.  Even though this happens at smaller block lengths, making it
     * potentially inefficient, it can take advantage of some vectorization
     * available on most modern processors. */
#define SHM_OPS_THRESHOLD  (16)

    if (MPIR_DATATYPE_IS_PREDEFINED(stype) && MPIR_DATATYPE_IS_PREDEFINED(dtype) &&
        scount <= SHM_OPS_THRESHOLD) {

        /* FIXME: We currently only optimize a few predefined datatypes, which
         * have a direct C datatype mapping. */

        /* The below list of datatypes is based on those specified in the MPI-3
         * standard on page 665. */
        switch (stype) {
        case MPI_CHAR:
            ASSIGN_COPY(src, dest, scount, char);

        case MPI_SHORT:
            ASSIGN_COPY(src, dest, scount, signed short int);

        case MPI_INT:
            ASSIGN_COPY(src, dest, scount, signed int);

        case MPI_LONG:
            ASSIGN_COPY(src, dest, scount, signed long int);

        case MPI_LONG_LONG_INT:        /* covers MPI_LONG_LONG too */
            ASSIGN_COPY(src, dest, scount, signed long long int);

        case MPI_SIGNED_CHAR:
            ASSIGN_COPY(src, dest, scount, signed char);

        case MPI_UNSIGNED_CHAR:
            ASSIGN_COPY(src, dest, scount, unsigned char);

        case MPI_UNSIGNED_SHORT:
            ASSIGN_COPY(src, dest, scount, unsigned short int);

        case MPI_UNSIGNED:
            ASSIGN_COPY(src, dest, scount, unsigned int);

        case MPI_UNSIGNED_LONG:
            ASSIGN_COPY(src, dest, scount, unsigned long int);

        case MPI_UNSIGNED_LONG_LONG:
            ASSIGN_COPY(src, dest, scount, unsigned long long int);

        case MPI_FLOAT:
            ASSIGN_COPY(src, dest, scount, float);

        case MPI_DOUBLE:
            ASSIGN_COPY(src, dest, scount, double);

        case MPI_LONG_DOUBLE:
            ASSIGN_COPY(src, dest, scount, long double);

#if 0
            /* FIXME: we need a configure check to define HAVE_WCHAR_T before
             * this can be enabled */
        case MPI_WCHAR:
            ASSIGN_COPY(src, dest, scount, wchar_t);
#endif

#if 0
            /* FIXME: we need a configure check to define HAVE_C_BOOL before
             * this can be enabled */
        case MPI_C_BOOL:
            ASSIGN_COPY(src, dest, scount, _Bool);
#endif

#if HAVE_INT8_T
        case MPI_INT8_T:
            ASSIGN_COPY(src, dest, scount, int8_t);
#endif /* HAVE_INT8_T */

#if HAVE_INT16_T
        case MPI_INT16_T:
            ASSIGN_COPY(src, dest, scount, int16_t);
#endif /* HAVE_INT16_T */

#if HAVE_INT32_T
        case MPI_INT32_T:
            ASSIGN_COPY(src, dest, scount, int32_t);
#endif /* HAVE_INT32_T */

#if HAVE_INT64_T
        case MPI_INT64_T:
            ASSIGN_COPY(src, dest, scount, int64_t);
#endif /* HAVE_INT64_T */

#if HAVE_UINT8_T
        case MPI_UINT8_T:
            ASSIGN_COPY(src, dest, scount, uint8_t);
#endif /* HAVE_UINT8_T */

#if HAVE_UINT16_T
        case MPI_UINT16_T:
            ASSIGN_COPY(src, dest, scount, uint16_t);
#endif /* HAVE_UINT16_T */

#if HAVE_UINT32_T
        case MPI_UINT32_T:
            ASSIGN_COPY(src, dest, scount, uint32_t);
#endif /* HAVE_UINT32_T */

#if HAVE_UINT64_T
        case MPI_UINT64_T:
            ASSIGN_COPY(src, dest, scount, uint64_t);
#endif /* HAVE_UINT64_T */

        case MPI_AINT:
            ASSIGN_COPY(src, dest, scount, MPI_Aint);

        case MPI_COUNT:
            ASSIGN_COPY(src, dest, scount, MPI_Count);

        case MPI_OFFSET:
            ASSIGN_COPY(src, dest, scount, MPI_Offset);

#if 0
            /* FIXME: we need a configure check to define HAVE_C_COMPLEX before
             * this can be enabled */
        case MPI_C_COMPLEX:    /* covers MPI_C_FLOAT_COMPLEX as well */
            ASSIGN_COPY(src, dest, scount, float _Complex);
#endif

#if 0
            /* FIXME: we need a configure check to define HAVE_C_DOUPLE_COMPLEX
             * before this can be enabled */
        case MPI_C_DOUBLE_COMPLEX:
            ASSIGN_COPY(src, dest, scount, double _Complex);
#endif

#if 0
            /* FIXME: we need a configure check to define
             * HAVE_C_LONG_DOUPLE_COMPLEX before this can be enabled */
        case MPI_C_LONG_DOUBLE_COMPLEX:
            ASSIGN_COPY(src, dest, scount, long double _Complex);
#endif

#if 0
            /* Types that don't have a direct equivalent */
        case MPI_BYTE:
        case MPI_PACKED:
#endif

#if 0   /* Fortran types */
        case MPI_INTEGER:
        case MPI_REAL:
        case MPI_DOUBLE_PRECISION:
        case MPI_COMPLEX:
        case MPI_LOGICAL:
        case MPI_CHARACTER:
#endif

#if 0   /* C++ types */
        case MPI_CXX_BOOL:
        case MPI_CXX_FLOAT_COMPLEX:
        case MPI_CXX_DOUBLE_COMPLEX:
        case MPI_CXX_LONG_DOUBLE_COMPLEX:
#endif

#if 0   /* Optional Fortran types */
        case MPI_DOUBLE_COMPLEX:
        case MPI_INTEGER1:
        case MPI_INTEGER2:
        case MPI_INTEGER4:
        case MPI_INTEGER8:
        case MPI_INTEGER16:
        case MPI_REAL2:
        case MPI_REAL4:
        case MPI_REAL8:
        case MPI_REAL16:
        case MPI_COMPLEX4:
        case MPI_COMPLEX8:
        case MPI_COMPLEX16:
        case MPI_COMPLEX32:
#endif

#if 0   /* C datatypes for reduction functions */
        case MPI_FLOAT_INT:
        case MPI_DOUBLE_INT:
        case MPI_LONG_INT:
        case MPI_2INT:
        case MPI_LONG_DOUBLE_INT:
#endif

#if 0   /* Fortran datatypes for reduction functions */
        case MPI_2REAL:
        case MPI_2DOUBLE_PRECISION:
        case MPI_2INTEGER:
#endif

#if 0   /* Random types not present in the standard */
        case MPI_2COMPLEX:
        case MPI_2DOUBLE_COMPLEX:
#endif

        default:
            /* Just to make sure the switch statement is not empty */
            ;
        }
    }

    mpi_errno = MPIR_Localcopy(src, scount, stype, dest, dcount, dtype);
    if (mpi_errno) {
        MPIU_ERR_POP(mpi_errno);
    }

  fn_exit:
    return mpi_errno;
    /* --BEGIN ERROR HANDLING-- */
  fn_fail:
    goto fn_exit;
    /* --END ERROR HANDLING-- */
}

#undef FUNCNAME
#define FUNCNAME MPIDI_CH3I_Shm_put_op
#undef FCNAME
#define FCNAME MPIDI_QUOTE(FUNCNAME)
static inline int MPIDI_CH3I_Shm_put_op(const void *origin_addr, int origin_count, MPI_Datatype
                                        origin_datatype, int target_rank, MPI_Aint target_disp,
                                        int target_count, MPI_Datatype target_datatype,
                                        MPID_Win * win_ptr)
{
    int mpi_errno = MPI_SUCCESS;
    void *base = NULL;
    int disp_unit;
    MPIDI_STATE_DECL(MPID_STATE_MPIDI_CH3I_SHM_PUT_OP);

    MPIDI_RMA_FUNC_ENTER(MPID_STATE_MPIDI_CH3I_SHM_PUT_OP);

    if (win_ptr->shm_allocated == TRUE) {
270
271
272
        int local_target_rank = win_ptr->comm_ptr->intranode_table[target_rank];
        MPIU_Assert(local_target_rank >= 0);
        base = win_ptr->shm_base_addrs[local_target_rank];
273
        disp_unit = win_ptr->basic_info_table[target_rank].disp_unit;
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
    }
    else {
        base = win_ptr->base;
        disp_unit = win_ptr->disp_unit;
    }

    mpi_errno = shm_copy(origin_addr, origin_count, origin_datatype,
                         (char *) base + disp_unit * target_disp, target_count, target_datatype);
    if (mpi_errno) {
        MPIU_ERR_POP(mpi_errno);
    }

  fn_exit:
    MPIDI_RMA_FUNC_EXIT(MPID_STATE_MPIDI_CH3I_SHM_PUT_OP);
    return mpi_errno;
    /* --BEGIN ERROR HANDLING-- */
  fn_fail:
    goto fn_exit;
    /* --END ERROR HANDLING-- */
}


#undef FUNCNAME
#define FUNCNAME MPIDI_CH3I_Shm_acc_op
#undef FCNAME
#define FCNAME MPIDI_QUOTE(FUNCNAME)
static inline int MPIDI_CH3I_Shm_acc_op(const void *origin_addr, int origin_count, MPI_Datatype
                                        origin_datatype, int target_rank, MPI_Aint target_disp,
                                        int target_count, MPI_Datatype target_datatype, MPI_Op op,
                                        MPID_Win * win_ptr)
{
    void *base = NULL;
    int disp_unit, shm_op = 0;
    int mpi_errno = MPI_SUCCESS;
308
309
310
311
312
313
314
    int i;
    MPI_Datatype predefined_type;
    MPI_Aint stream_elem_count, stream_unit_count;
    MPI_Aint predefined_dtp_size, predefined_dtp_extent, predefined_dtp_count;
    MPI_Aint total_len, rest_len;
    MPI_Aint origin_dtp_size;
    MPID_Datatype *origin_dtp_ptr = NULL;
315
316
317
318
319
    MPIDI_STATE_DECL(MPID_STATE_MPIDI_CH3I_SHM_ACC_OP);

    MPIDI_RMA_FUNC_ENTER(MPID_STATE_MPIDI_CH3I_SHM_ACC_OP);

    if (win_ptr->shm_allocated == TRUE) {
320
321
        int local_target_rank = win_ptr->comm_ptr->intranode_table[target_rank];
        MPIU_Assert(local_target_rank >= 0);
322
        shm_op = 1;
323
        base = win_ptr->shm_base_addrs[local_target_rank];
324
        disp_unit = win_ptr->basic_info_table[target_rank].disp_unit;
325
326
327
328
329
330
    }
    else {
        base = win_ptr->base;
        disp_unit = win_ptr->disp_unit;
    }

331
332
    if (MPIR_DATATYPE_IS_PREDEFINED(origin_datatype)) {
        if (shm_op) {
333
            MPIDI_CH3I_SHM_MUTEX_LOCK(win_ptr);
334
335
336
337
338
        }
        mpi_errno = do_accumulate_op((void*)origin_addr, origin_count, origin_datatype,
                                     (void*)((char *)base+disp_unit*target_disp), target_count, target_datatype,
                                     0, op);
        if (shm_op) {
339
340
341
            MPIDI_CH3I_SHM_MUTEX_UNLOCK(win_ptr);
        }

342
        if (mpi_errno != MPI_SUCCESS) MPIU_ERR_POP(mpi_errno);
343

344
        goto fn_exit;
345
346
    }

347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
    /* Get total length of origin data */
    MPID_Datatype_get_size_macro(origin_datatype, origin_dtp_size);
    total_len = origin_dtp_size * origin_count;

    MPID_Datatype_get_ptr(origin_datatype, origin_dtp_ptr);
    MPIU_Assert(origin_dtp_ptr != NULL && origin_dtp_ptr->eltype != MPI_DATATYPE_NULL);
    predefined_type = origin_dtp_ptr->eltype;
    MPID_Datatype_get_size_macro(predefined_type, predefined_dtp_size);
    predefined_dtp_count = total_len / predefined_dtp_size;
    MPID_Datatype_get_extent_macro(predefined_type, predefined_dtp_extent);
    MPIU_Assert(predefined_dtp_count > 0 && predefined_dtp_size > 0 && predefined_dtp_extent > 0);

    stream_elem_count = MPIDI_CH3U_Acc_stream_size / predefined_dtp_extent;
    stream_unit_count = (predefined_dtp_count - 1) / stream_elem_count + 1;
    MPIU_Assert(stream_elem_count > 0 && stream_unit_count > 0);

    rest_len = total_len;
    for (i = 0; i < stream_unit_count; i++) {
        MPID_Segment *seg = NULL;
        void *packed_buf = NULL;
367
        MPI_Aint first, last;
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
        int is_predef_contig;
        MPI_Aint stream_offset, stream_size, stream_count;

        stream_offset = i * stream_elem_count * predefined_dtp_size;
        stream_size = MPIR_MIN(stream_elem_count * predefined_dtp_size, rest_len);
        stream_count = stream_size / predefined_dtp_size;
        rest_len -= stream_size;

        first = stream_offset;
        last = stream_offset + stream_size;

        packed_buf = MPIU_Malloc(stream_size);

        seg = MPID_Segment_alloc();
        MPIU_ERR_CHKANDJUMP1(seg == NULL, mpi_errno, MPI_ERR_OTHER, "**nomem", "**nomem %s", "MPID_Segment");
        MPID_Segment_init(origin_addr, origin_count, origin_datatype, seg, 0);
        MPID_Segment_pack(seg, first, &last, packed_buf);
        MPID_Segment_free(seg);

        MPID_Datatype_is_contig(predefined_type, &is_predef_contig);

        if (!is_predef_contig) {
            void *tmpbuf = MPIU_Malloc(stream_count * predefined_dtp_extent);
            mpi_errno = MPIR_Localcopy(tmpbuf, stream_count, predefined_type,
                                       packed_buf, stream_size, MPI_BYTE);
            if (mpi_errno != MPI_SUCCESS) MPIU_ERR_POP(mpi_errno);
            MPIU_Free(packed_buf);
            packed_buf = tmpbuf;
396
397
        }

398
399
        if (shm_op) {
            MPIDI_CH3I_SHM_MUTEX_LOCK(win_ptr);
400
        }
401
402
403
404
405
406
407

        mpi_errno = do_accumulate_op((void*)packed_buf, stream_count, predefined_type,
                                     (void*)((char*)base+disp_unit*target_disp), target_count, target_datatype,
                                     stream_offset, op);

        if (shm_op) {
            MPIDI_CH3I_SHM_MUTEX_UNLOCK(win_ptr);
408
        }
409
410
411
412

        if (mpi_errno != MPI_SUCCESS) MPIU_ERR_POP(mpi_errno);

        MPIU_Free(packed_buf);
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
    }

  fn_exit:
    MPIDI_RMA_FUNC_EXIT(MPID_STATE_MPIDI_CH3I_SHM_ACC_OP);
    return mpi_errno;
    /* --BEGIN ERROR HANDLING-- */
  fn_fail:
    goto fn_exit;
    /* --END ERROR HANDLING-- */
}


#undef FUNCNAME
#define FUNCNAME MPIDI_CH3I_Shm_get_acc_op
#undef FCNAME
#define FCNAME MPIDI_QUOTE(FUNCNAME)
static inline int MPIDI_CH3I_Shm_get_acc_op(const void *origin_addr, int origin_count, MPI_Datatype
                                            origin_datatype, void *result_addr, int result_count,
                                            MPI_Datatype result_datatype, int target_rank, MPI_Aint
                                            target_disp, int target_count,
                                            MPI_Datatype target_datatype, MPI_Op op,
                                            MPID_Win * win_ptr)
{
    int disp_unit, shm_locked = 0;
    void *base = NULL;
438
439
440
441
442
443
444
    int i;
    MPI_Datatype predefined_type;
    MPI_Aint stream_elem_count, stream_unit_count;
    MPI_Aint predefined_dtp_size, predefined_dtp_extent, predefined_dtp_count;
    MPI_Aint total_len, rest_len;
    MPI_Aint origin_dtp_size;
    MPID_Datatype *origin_dtp_ptr = NULL;
445
446
447
448
449
450
    int mpi_errno = MPI_SUCCESS;
    MPIDI_STATE_DECL(MPID_STATE_MPIDI_CH3I_SHM_GET_ACC_OP);

    MPIDI_RMA_FUNC_ENTER(MPID_STATE_MPIDI_CH3I_SHM_GET_ACC_OP);

    if (win_ptr->shm_allocated == TRUE) {
451
452
453
        int local_target_rank = win_ptr->comm_ptr->intranode_table[target_rank];
        MPIU_Assert(local_target_rank >= 0);
        base = win_ptr->shm_base_addrs[local_target_rank];
454
        disp_unit = win_ptr->basic_info_table[target_rank].disp_unit;
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
        MPIDI_CH3I_SHM_MUTEX_LOCK(win_ptr);
        shm_locked = 1;
    }
    else {
        base = win_ptr->base;
        disp_unit = win_ptr->disp_unit;
    }

    /* Perform the local get first, then the accumulate */
    mpi_errno = shm_copy((char *) base + disp_unit * target_disp, target_count, target_datatype,
                         result_addr, result_count, result_datatype);
    if (mpi_errno) {
        MPIU_ERR_POP(mpi_errno);
    }

    /* NO_OP: Don't perform the accumulate */
    if (op == MPI_NO_OP) {
        if (shm_locked) {
            MPIDI_CH3I_SHM_MUTEX_UNLOCK(win_ptr);
            shm_locked = 0;
        }

        goto fn_exit;
    }

480
    if (MPIR_DATATYPE_IS_PREDEFINED(origin_datatype)) {
481

482
483
484
        mpi_errno = do_accumulate_op((void*)origin_addr, origin_count, origin_datatype,
                                     (void*)((char *)base+disp_unit*target_disp), target_count, target_datatype,
                                     0, op);
485
486
487
488
        if (shm_locked) {
            MPIDI_CH3I_SHM_MUTEX_UNLOCK(win_ptr);
        }

489
        if (mpi_errno != MPI_SUCCESS) MPIU_ERR_POP(mpi_errno);
490

491
        goto fn_exit;
492
493
    }

494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
    /* Get total length of origin data */
    MPID_Datatype_get_size_macro(origin_datatype, origin_dtp_size);
    total_len = origin_dtp_size * origin_count;

    MPID_Datatype_get_ptr(origin_datatype, origin_dtp_ptr);
    MPIU_Assert(origin_dtp_ptr != NULL && origin_dtp_ptr->eltype != MPI_DATATYPE_NULL);
    predefined_type = origin_dtp_ptr->eltype;
    MPID_Datatype_get_size_macro(predefined_type, predefined_dtp_size);
    predefined_dtp_count = total_len / predefined_dtp_size;
    MPID_Datatype_get_extent_macro(predefined_type, predefined_dtp_extent);
    MPIU_Assert(predefined_dtp_count > 0 && predefined_dtp_size > 0 && predefined_dtp_extent > 0);

    stream_elem_count = MPIDI_CH3U_Acc_stream_size / predefined_dtp_extent;
    stream_unit_count = (predefined_dtp_count - 1) / stream_elem_count + 1;
    MPIU_Assert(stream_elem_count > 0 && stream_unit_count > 0);

    rest_len = total_len;
    for (i = 0; i < stream_unit_count; i++) {
        MPID_Segment *seg = NULL;
        void *packed_buf = NULL;
514
        MPI_Aint first, last;
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
        int is_predef_contig;
        MPI_Aint stream_offset, stream_size, stream_count;

        stream_offset = i * stream_elem_count * predefined_dtp_size;
        stream_size = MPIR_MIN(stream_elem_count * predefined_dtp_size, rest_len);
        stream_count = stream_size / predefined_dtp_size;
        rest_len -= stream_size;

        first = stream_offset;
        last = stream_offset + stream_size;

        packed_buf = MPIU_Malloc(stream_size);

        seg = MPID_Segment_alloc();
        MPIU_ERR_CHKANDJUMP1(seg == NULL, mpi_errno, MPI_ERR_OTHER, "**nomem", "**nomem %s", "MPID_Segment");
        MPID_Segment_init(origin_addr, origin_count, origin_datatype, seg, 0);
        MPID_Segment_pack(seg, first, &last, packed_buf);
        MPID_Segment_free(seg);

        MPID_Datatype_is_contig(predefined_type, &is_predef_contig);

        if (!is_predef_contig) {
            void *tmpbuf = MPIU_Malloc(stream_count * predefined_dtp_extent);
            mpi_errno = MPIR_Localcopy(tmpbuf, stream_count, predefined_type,
                                       packed_buf, stream_size, MPI_BYTE);
            if (mpi_errno != MPI_SUCCESS) MPIU_ERR_POP(mpi_errno);
            MPIU_Free(packed_buf);
            packed_buf = tmpbuf;
543
544
        }

545
546
547
        mpi_errno = do_accumulate_op((void*)packed_buf, stream_count, predefined_type,
                                     (void*)((char*)base+disp_unit*target_disp), target_count, target_datatype,
                                     stream_offset, op);
548

549
550
551
        if (mpi_errno != MPI_SUCCESS) MPIU_ERR_POP(mpi_errno);

        MPIU_Free(packed_buf);
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
    }

    if (shm_locked) {
        MPIDI_CH3I_SHM_MUTEX_UNLOCK(win_ptr);
        shm_locked = 0;
    }

  fn_exit:
    MPIDI_RMA_FUNC_EXIT(MPID_STATE_MPIDI_CH3I_SHM_GET_ACC_OP);
    return mpi_errno;
    /* --BEGIN ERROR HANDLING-- */
  fn_fail:
    if (shm_locked) {
        MPIDI_CH3I_SHM_MUTEX_UNLOCK(win_ptr);
    }
    goto fn_exit;
    /* --END ERROR HANDLING-- */
}


#undef FUNCNAME
#define FUNCNAME MPIDI_CH3I_Shm_get_op
#undef FCNAME
#define FCNAME MPIDI_QUOTE(FUNCNAME)
static inline int MPIDI_CH3I_Shm_get_op(void *origin_addr, int origin_count,
                                        MPI_Datatype origin_datatype, int target_rank,
                                        MPI_Aint target_disp, int target_count,
                                        MPI_Datatype target_datatype, MPID_Win * win_ptr)
{
    void *base = NULL;
    int disp_unit;
    int mpi_errno = MPI_SUCCESS;
    MPIDI_STATE_DECL(MPID_STATE_MPIDI_CH3I_SHM_GET_OP);

    MPIDI_RMA_FUNC_ENTER(MPID_STATE_MPIDI_CH3I_SHM_GET_OP);

    if (win_ptr->shm_allocated == TRUE) {
589
590
591
        int local_target_rank = win_ptr->comm_ptr->intranode_table[target_rank];
        MPIU_Assert(local_target_rank >= 0);
        base = win_ptr->shm_base_addrs[local_target_rank];
592
        disp_unit = win_ptr->basic_info_table[target_rank].disp_unit;
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
    }
    else {
        base = win_ptr->base;
        disp_unit = win_ptr->disp_unit;
    }

    mpi_errno = shm_copy((char *) base + disp_unit * target_disp, target_count, target_datatype,
                         origin_addr, origin_count, origin_datatype);
    if (mpi_errno) {
        MPIU_ERR_POP(mpi_errno);
    }

  fn_exit:
    MPIDI_RMA_FUNC_EXIT(MPID_STATE_MPIDI_CH3I_SHM_GET_OP);
    return mpi_errno;
    /* --BEGIN ERROR HANDLING-- */
  fn_fail:
    goto fn_exit;
    /* --END ERROR HANDLING-- */
}


#undef FUNCNAME
#define FUNCNAME MPIDI_CH3I_Shm_cas_op
#undef FCNAME
#define FCNAME MPIDI_QUOTE(FUNCNAME)
static inline int MPIDI_CH3I_Shm_cas_op(const void *origin_addr, const void *compare_addr,
                                        void *result_addr, MPI_Datatype datatype, int target_rank,
                                        MPI_Aint target_disp, MPID_Win * win_ptr)
{
    void *base = NULL, *dest_addr = NULL;
    int disp_unit;
    MPI_Aint len;
    int shm_locked = 0;
    int mpi_errno = MPI_SUCCESS;
    MPIDI_STATE_DECL(MPID_STATE_MPIDI_CH3I_SHM_CAS_OP);

    MPIDI_RMA_FUNC_ENTER(MPID_STATE_MPIDI_CH3I_SHM_CAS_OP);

    if (win_ptr->shm_allocated == TRUE) {
633
634
635
        int local_target_rank = win_ptr->comm_ptr->intranode_table[target_rank];
        MPIU_Assert(local_target_rank >= 0);
        base = win_ptr->shm_base_addrs[local_target_rank];
636
        disp_unit = win_ptr->basic_info_table[target_rank].disp_unit;
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691

        MPIDI_CH3I_SHM_MUTEX_LOCK(win_ptr);
        shm_locked = 1;
    }
    else {
        base = win_ptr->base;
        disp_unit = win_ptr->disp_unit;
    }

    dest_addr = (char *) base + disp_unit * target_disp;

    MPID_Datatype_get_size_macro(datatype, len);
    MPIU_Memcpy(result_addr, dest_addr, len);

    if (MPIR_Compare_equal(compare_addr, dest_addr, datatype)) {
        MPIU_Memcpy(dest_addr, origin_addr, len);
    }

    if (shm_locked) {
        MPIDI_CH3I_SHM_MUTEX_UNLOCK(win_ptr);
        shm_locked = 0;
    }

  fn_exit:
    MPIDI_RMA_FUNC_EXIT(MPID_STATE_MPIDI_CH3I_SHM_CAS_OP);
    return mpi_errno;
    /* --BEGIN ERROR HANDLING-- */
  fn_fail:
    if (shm_locked) {
        MPIDI_CH3I_SHM_MUTEX_UNLOCK(win_ptr);
    }
    goto fn_exit;
    /* --END ERROR HANDLING-- */
}


#undef FUNCNAME
#define FUNCNAME MPIDI_CH3I_Shm_fop_op
#undef FCNAME
#define FCNAME MPIDI_QUOTE(FUNCNAME)
static inline int MPIDI_CH3I_Shm_fop_op(const void *origin_addr, void *result_addr,
                                        MPI_Datatype datatype, int target_rank,
                                        MPI_Aint target_disp, MPI_Op op, MPID_Win * win_ptr)
{
    void *base = NULL, *dest_addr = NULL;
    MPI_User_function *uop = NULL;
    int disp_unit;
    MPI_Aint len;
    int one, shm_locked = 0;
    int mpi_errno = MPI_SUCCESS;
    MPIDI_STATE_DECL(MPID_STATE_MPIDI_CH3I_SHM_FOP_OP);

    MPIDI_RMA_FUNC_ENTER(MPID_STATE_MPIDI_CH3I_SHM_FOP_OP);

    if (win_ptr->shm_allocated == TRUE) {
692
693
694
        int local_target_rank = win_ptr->comm_ptr->intranode_table[target_rank];
        MPIU_Assert(local_target_rank >= 0);
        base = win_ptr->shm_base_addrs[local_target_rank];
695
        disp_unit = win_ptr->basic_info_table[target_rank].disp_unit;
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

        MPIDI_CH3I_SHM_MUTEX_LOCK(win_ptr);
        shm_locked = 1;
    }
    else {
        base = win_ptr->base;
        disp_unit = win_ptr->disp_unit;
    }

    dest_addr = (char *) base + disp_unit * target_disp;

    MPID_Datatype_get_size_macro(datatype, len);
    MPIU_Memcpy(result_addr, dest_addr, len);

    uop = MPIR_OP_HDL_TO_FN(op);
    one = 1;

    (*uop) ((void *) origin_addr, dest_addr, &one, &datatype);

    if (shm_locked) {
        MPIDI_CH3I_SHM_MUTEX_UNLOCK(win_ptr);
        shm_locked = 0;
    }

  fn_exit:
    MPIDI_RMA_FUNC_EXIT(MPID_STATE_MPIDI_CH3I_SHM_FOP_OP);
    return mpi_errno;
    /* --BEGIN ERROR HANDLING-- */
  fn_fail:
    if (shm_locked) {
        MPIDI_CH3I_SHM_MUTEX_UNLOCK(win_ptr);
    }
    goto fn_exit;
    /* --END ERROR HANDLING-- */
}


#endif /* MPID_RMA_SHM_H_INCLUDED */