ch3u_rma_ops.c 20.4 KB
Newer Older
1 2 3 4 5 6 7 8 9
/* -*- Mode: C; c-basic-offset:4 ; -*- */
/*
 *  (C) 2001 by Argonne National Laboratory.
 *      See COPYRIGHT in top-level directory.
 */

#include "mpidi_ch3_impl.h"
#include "mpidrma.h"

10 11 12 13 14 15 16 17
static int enableShortACC=1;

MPIU_THREADSAFE_INIT_DECL(initRMAoptions);
#ifdef USE_MPIU_INSTR
MPIU_INSTR_DURATION_DECL(wincreate_allgather);
MPIU_INSTR_DURATION_DECL(winfree_rs);
MPIU_INSTR_DURATION_DECL(winfree_complete);
MPIU_INSTR_DURATION_DECL(rmaqueue_alloc);
18
MPIU_INSTR_DURATION_DECL(rmaqueue_set);
19 20 21
extern void MPIDI_CH3_RMA_InitInstr(void);
#endif

22 23 24
#define MPIDI_PASSIVE_TARGET_DONE_TAG  348297
#define MPIDI_PASSIVE_TARGET_RMA_TAG 563924

25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40
/* 
 * TODO:
 * Explore use of alternate allocation mechanisms for the RMA queue elements
 * (Because profiling has shown that queue element allocation/deallocation
 * can take a significant amount of time in the RMA operations).
 *    1: Current approach (uses perm memory malloc/free)
 *    2: Preallocate and maintain list (use perm memory malloc, but
 *       free onto window; use first; free on window free)
 *    3: Preallocate and maintain list (use separate memory, but free to
 *       thread/process; free in Finalize handler.  Option to use for
 *       single-threaded to avoid thread overheads)
 * Possible interface
 *    int MPIDI_RMAListAlloc(MPIDI_RMA_ops **a,MPID_Win *win)
 *    int MPIDI_RMAListFree(MPIDI_RMA_ops *a, MPID_Win *win)
 *    return value is error code (e.g., allocation failure).
 */
41 42 43 44 45 46

#undef FUNCNAME
#define FUNCNAME MPIDI_Win_create
#undef FCNAME
#define FCNAME MPIDI_QUOTE(FUNCNAME)
int MPIDI_Win_create(void *base, MPI_Aint size, int disp_unit, MPID_Info *info,
47
		     MPID_Comm *comm_ptr, MPID_Win **win_ptr )
48
{
49
    int mpi_errno=MPI_SUCCESS, i, k, comm_size, rank;
50
    MPI_Aint *tmp_buf;
51
    MPID_Comm *win_comm_ptr;
52
    int errflag = FALSE;
53 54 55 56 57 58 59 60 61
    MPIU_CHKPMEM_DECL(4);
    MPIU_CHKLMEM_DECL(1);
    MPIDI_STATE_DECL(MPID_STATE_MPIDI_WIN_CREATE);
    
    MPIDI_RMA_FUNC_ENTER(MPID_STATE_MPIDI_WIN_CREATE);

    /* FIXME: There should be no unreferenced args */
    MPIU_UNREFERENCED_ARG(info);

62 63 64 65
    if(initRMAoptions) {
	int rc;
	MPIU_THREADSAFE_INIT_BLOCK_BEGIN(initRMAoptions);
#ifdef USE_MPIU_INSTR
66
	/* Define all instrumentation handles used in the CH3 RMA here*/
67 68 69 70
	MPIU_INSTR_DURATION_INIT(wincreate_allgather,0,"WIN_CREATE:Allgather");
	MPIU_INSTR_DURATION_INIT(winfree_rs,0,"WIN_FREE:ReduceScatterBlock");
	MPIU_INSTR_DURATION_INIT(winfree_complete,0,"WIN_FREE:Complete");
	MPIU_INSTR_DURATION_INIT(rmaqueue_alloc,0,"Allocate RMA Queue element");
71
	MPIU_INSTR_DURATION_INIT(rmaqueue_set,0,"Set fields in RMA Queue element");
72 73 74 75 76 77 78
	MPIDI_CH3_RMA_InitInstr();

#endif    
	MPIU_THREADSAFE_INIT_CLEAR(initRMAoptions);
	MPIU_THREADSAFE_INIT_BLOCK_END(initRMAoptions);
    }

79 80 81 82
    comm_size = comm_ptr->local_size;
    rank = comm_ptr->rank;
    
    *win_ptr = (MPID_Win *)MPIU_Handle_obj_alloc( &MPID_Win_mem );
83 84
    MPIU_ERR_CHKANDJUMP1(!(*win_ptr),mpi_errno,MPI_ERR_OTHER,"**nomem",
			 "**nomem %s","MPID_Win_mem");
85

86 87
    MPIU_Object_set_ref(*win_ptr, 1);

88 89 90 91 92 93 94
    (*win_ptr)->fence_cnt = 0;
    (*win_ptr)->base = base;
    (*win_ptr)->size = size;
    (*win_ptr)->disp_unit = disp_unit;
    (*win_ptr)->start_group_ptr = NULL; 
    (*win_ptr)->start_assert = 0; 
    (*win_ptr)->attributes = NULL;
95 96
    (*win_ptr)->rma_ops_list_head = NULL;
    (*win_ptr)->rma_ops_list_tail = NULL;
97 98 99 100 101 102 103
    (*win_ptr)->lock_granted = 0;
    (*win_ptr)->current_lock_type = MPID_LOCK_NONE;
    (*win_ptr)->shared_lock_ref_cnt = 0;
    (*win_ptr)->lock_queue = NULL;
    (*win_ptr)->my_counter = 0;
    (*win_ptr)->my_pt_rma_puts_accs = 0;
    
104 105
    mpi_errno = MPIR_Comm_dup_impl(comm_ptr, &win_comm_ptr);
    if (mpi_errno) MPIU_ERR_POP(mpi_errno);
106 107 108 109
    (*win_ptr)->comm_ptr   = win_comm_ptr;
    (*win_ptr)->myrank = rank;

    MPIU_INSTR_DURATION_START(wincreate_allgather);
110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133
    /* allocate memory for the base addresses, disp_units, and
       completion counters of all processes */ 
    MPIU_CHKPMEM_MALLOC((*win_ptr)->base_addrs, void **,
			comm_size*sizeof(void *), 
			mpi_errno, "(*win_ptr)->base_addrs");

    MPIU_CHKPMEM_MALLOC((*win_ptr)->disp_units, int *, comm_size*sizeof(int), 
			mpi_errno, "(*win_ptr)->disp_units");

    MPIU_CHKPMEM_MALLOC((*win_ptr)->all_win_handles, MPI_Win *, 
			comm_size*sizeof(MPI_Win), 
			mpi_errno, "(*win_ptr)->all_win_handles");
    
    MPIU_CHKPMEM_MALLOC((*win_ptr)->pt_rma_puts_accs, int *, 
			comm_size*sizeof(int), 
			mpi_errno, "(*win_ptr)->pt_rma_puts_accs");
    for (i=0; i<comm_size; i++)	(*win_ptr)->pt_rma_puts_accs[i] = 0;
    
    /* get the addresses of the windows, window objects, and completion
       counters of all processes.  allocate temp. buffer for communication */
    MPIU_CHKLMEM_MALLOC(tmp_buf, MPI_Aint *, 3*comm_size*sizeof(MPI_Aint),
			mpi_errno, "tmp_buf");
    
    /* FIXME: This needs to be fixed for heterogeneous systems */
134 135
    /* FIXME: If we wanted to validate the transfer as within range at the
       origin, we'd also need the window size. */
136
    tmp_buf[3*rank]   = MPIU_PtrToAint(base);
137 138 139
    tmp_buf[3*rank+1] = (MPI_Aint) disp_unit;
    tmp_buf[3*rank+2] = (MPI_Aint) (*win_ptr)->handle;
    
140 141
    mpi_errno = MPIR_Allgather_impl(MPI_IN_PLACE, 0, MPI_DATATYPE_NULL,
                                    tmp_buf, 3 * sizeof(MPI_Aint), MPI_BYTE,
142
                                    comm_ptr, &errflag);
143
    MPIU_INSTR_DURATION_END(wincreate_allgather);
144
    if (mpi_errno) { MPIU_ERR_POP(mpi_errno); }
145 146
    MPIU_ERR_CHKANDJUMP(errflag, mpi_errno, MPI_ERR_OTHER, "**coll_fail");

147 148

    k = 0;
149 150
    for (i=0; i<comm_size; i++)
    {
151 152 153
	(*win_ptr)->base_addrs[i] = MPIU_AintToPtr(tmp_buf[k++]);
	(*win_ptr)->disp_units[i] = (int) tmp_buf[k++];
	(*win_ptr)->all_win_handles[i] = (MPI_Win) tmp_buf[k++];
154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175
    }
        
 fn_exit:
    MPIU_CHKLMEM_FREEALL();
    MPIDI_RMA_FUNC_EXIT(MPID_STATE_MPIDI_WIN_CREATE);
    return mpi_errno;
    /* --BEGIN ERROR HANDLING-- */
 fn_fail:
    MPIU_CHKPMEM_REAP();
    goto fn_exit;
    /* --END ERROR HANDLING-- */
}




#undef FUNCNAME
#define FUNCNAME MPIDI_Win_free
#undef FCNAME
#define FCNAME MPIDI_QUOTE(FUNCNAME)
int MPIDI_Win_free(MPID_Win **win_ptr)
{
176
    int mpi_errno=MPI_SUCCESS, total_pt_rma_puts_accs;
177
    int in_use;
178
    MPID_Comm *comm_ptr;
179
    int errflag = FALSE;
180 181 182 183
    MPIDI_STATE_DECL(MPID_STATE_MPIDI_WIN_FREE);
        
    MPIDI_RMA_FUNC_ENTER(MPID_STATE_MPIDI_WIN_FREE);
        
184 185
    comm_ptr = (*win_ptr)->comm_ptr;
    MPIU_INSTR_DURATION_START(winfree_rs);
186 187
    mpi_errno = MPIR_Reduce_scatter_block_impl((*win_ptr)->pt_rma_puts_accs, 
                                               &total_pt_rma_puts_accs, 1, 
188
                                               MPI_INT, MPI_SUM, comm_ptr, &errflag);
189
    if (mpi_errno) { MPIU_ERR_POP(mpi_errno); }
190
    MPIU_ERR_CHKANDJUMP(errflag, mpi_errno, MPI_ERR_OTHER, "**coll_fail");
191
    MPIU_INSTR_DURATION_END(winfree_rs);
192 193 194 195 196 197

    if (total_pt_rma_puts_accs != (*win_ptr)->my_pt_rma_puts_accs)
    {
	MPID_Progress_state progress_state;
            
	/* poke the progress engine until the two are equal */
198
	MPIU_INSTR_DURATION_START(winfree_complete);
199 200 201 202 203 204 205 206
	MPID_Progress_start(&progress_state);
	while (total_pt_rma_puts_accs != (*win_ptr)->my_pt_rma_puts_accs)
	{
	    mpi_errno = MPID_Progress_wait(&progress_state);
	    /* --BEGIN ERROR HANDLING-- */
	    if (mpi_errno != MPI_SUCCESS)
	    {
		MPID_Progress_end(&progress_state);
207
		MPIU_ERR_SETANDJUMP(mpi_errno,MPI_ERR_OTHER,"**winnoprogress");
208 209 210 211
	    }
	    /* --END ERROR HANDLING-- */
	}
	MPID_Progress_end(&progress_state);
212
	MPIU_INSTR_DURATION_END(winfree_complete);
213 214
    }

215 216 217
    
    mpi_errno = MPIR_Comm_free_impl(comm_ptr);
    if (mpi_errno) MPIU_ERR_POP(mpi_errno);
218

219 220 221 222
    MPIU_Free((*win_ptr)->base_addrs);
    MPIU_Free((*win_ptr)->disp_units);
    MPIU_Free((*win_ptr)->all_win_handles);
    MPIU_Free((*win_ptr)->pt_rma_puts_accs);
223 224 225 226

    MPIU_Object_release_ref(*win_ptr, &in_use);
    /* MPI windows don't have reference count semantics, so this should always be true */
    MPIU_Assert(!in_use);
227
    MPIU_Handle_obj_free( &MPID_Win_mem, *win_ptr );
228

229 230 231
 fn_exit:
    MPIDI_RMA_FUNC_EXIT(MPID_STATE_MPIDI_WIN_FREE);
    return mpi_errno;
232

233 234 235 236 237 238 239 240 241 242 243 244 245 246 247
 fn_fail:
    goto fn_exit;
}



#undef FUNCNAME
#define FUNCNAME MPIDI_Put
#undef FCNAME
#define FCNAME MPIDI_QUOTE(FUNCNAME)
int MPIDI_Put(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;
Pavan Balaji's avatar
Pavan Balaji committed
248
    int dt_contig ATTRIBUTE((unused)), rank, predefined;
249
    MPIDI_RMA_ops *new_ptr;
250
    MPID_Datatype *dtp;
Pavan Balaji's avatar
Pavan Balaji committed
251
    MPI_Aint dt_true_lb ATTRIBUTE((unused));
252 253 254 255 256 257 258 259 260 261 262 263 264 265
    MPIDI_msg_sz_t data_sz;
    MPIU_CHKPMEM_DECL(1);
    MPIDI_STATE_DECL(MPID_STATE_MPIDI_PUT);
        
    MPIDI_RMA_FUNC_ENTER(MPID_STATE_MPIDI_PUT);

    MPIDI_Datatype_get_info(origin_count, origin_datatype,
			    dt_contig, data_sz, dtp,dt_true_lb); 
    
    if ((data_sz == 0) || (target_rank == MPI_PROC_NULL))
    {
	goto fn_exit;
    }

266
    rank = win_ptr->myrank;
267 268 269 270 271 272 273 274 275 276 277
    
    /* If the put is a local operation, do it here */
    if (target_rank == rank)
    {
	mpi_errno = MPIR_Localcopy(origin_addr, origin_count, origin_datatype,
				   (char *) win_ptr->base + win_ptr->disp_unit *
				   target_disp, target_count, target_datatype); 
    }
    else
    {
	/* queue it up */
278 279
	/* FIXME: For short operations, should we use a (per-thread) pool? */
	MPIU_INSTR_DURATION_START(rmaqueue_alloc);
280 281
	MPIU_CHKPMEM_MALLOC(new_ptr, MPIDI_RMA_ops *, sizeof(MPIDI_RMA_ops), 
			    mpi_errno, "RMA operation entry");
282 283 284 285 286 287 288
	MPIU_INSTR_DURATION_END(rmaqueue_alloc);
	if (win_ptr->rma_ops_list_tail) 
	    win_ptr->rma_ops_list_tail->next = new_ptr;
	else
	    win_ptr->rma_ops_list_head = new_ptr;
	win_ptr->rma_ops_list_tail = new_ptr;

289
	MPIU_INSTR_DURATION_START(rmaqueue_set);
290
	/* FIXME: For contig and very short operations, use a streamlined op */
291 292 293 294 295 296 297 298 299
	new_ptr->next = NULL;  
	new_ptr->type = MPIDI_RMA_PUT;
	new_ptr->origin_addr = origin_addr;
	new_ptr->origin_count = origin_count;
	new_ptr->origin_datatype = origin_datatype;
	new_ptr->target_rank = target_rank;
	new_ptr->target_disp = target_disp;
	new_ptr->target_count = target_count;
	new_ptr->target_datatype = target_datatype;
300 301
	MPIU_INSTR_DURATION_END(rmaqueue_set);

302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340
	/* if source or target datatypes are derived, increment their
	   reference counts */ 
	MPIDI_CH3I_DATATYPE_IS_PREDEFINED(origin_datatype, predefined);
	if (!predefined)
	{
	    MPID_Datatype_get_ptr(origin_datatype, dtp);
	    MPID_Datatype_add_ref(dtp);
	}
	MPIDI_CH3I_DATATYPE_IS_PREDEFINED(target_datatype, predefined);
	if (!predefined)
	{
	    MPID_Datatype_get_ptr(target_datatype, dtp);
	    MPID_Datatype_add_ref(dtp);
	}
    }

  fn_exit:
    MPIDI_RMA_FUNC_EXIT(MPID_STATE_MPIDI_PUT);    
    return mpi_errno;

    /* --BEGIN ERROR HANDLING-- */
  fn_fail:
    MPIU_CHKPMEM_REAP();
    goto fn_exit;
    /* --END ERROR HANDLING-- */
}



#undef FUNCNAME
#define FUNCNAME MPIDI_Get
#undef FCNAME
#define FCNAME MPIDI_QUOTE(FUNCNAME)
int MPIDI_Get(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;
    MPIDI_msg_sz_t data_sz;
Pavan Balaji's avatar
Pavan Balaji committed
341 342
    int dt_contig ATTRIBUTE((unused)), rank, predefined;
    MPI_Aint dt_true_lb ATTRIBUTE((unused));
343
    MPIDI_RMA_ops *new_ptr;
344 345 346 347 348 349 350 351 352 353 354 355 356 357
    MPID_Datatype *dtp;
    MPIU_CHKPMEM_DECL(1);
    MPIDI_STATE_DECL(MPID_STATE_MPIDI_GET);
        
    MPIDI_RMA_FUNC_ENTER(MPID_STATE_MPIDI_GET);

    MPIDI_Datatype_get_info(origin_count, origin_datatype,
			    dt_contig, data_sz, dtp, dt_true_lb); 

    if ((data_sz == 0) || (target_rank == MPI_PROC_NULL))
    {
	goto fn_exit;
    }

358
    rank = win_ptr->myrank;
359 360 361 362 363 364 365 366 367 368 369 370 371
    
    /* If the get is a local operation, do it here */
    if (target_rank == rank)
    {
	mpi_errno = MPIR_Localcopy((char *) win_ptr->base +
				   win_ptr->disp_unit * target_disp,
				   target_count, target_datatype,
				   origin_addr, origin_count,
				   origin_datatype);  
    }
    else
    {
	/* queue it up */
372
	MPIU_INSTR_DURATION_START(rmaqueue_alloc);
373 374
	MPIU_CHKPMEM_MALLOC(new_ptr, MPIDI_RMA_ops *, sizeof(MPIDI_RMA_ops), 
			    mpi_errno, "RMA operation entry");
375 376 377
	MPIU_INSTR_DURATION_END(rmaqueue_alloc);
	if (win_ptr->rma_ops_list_tail) 
	    win_ptr->rma_ops_list_tail->next = new_ptr;
378
	else
379 380
	    win_ptr->rma_ops_list_head = new_ptr;
	win_ptr->rma_ops_list_tail = new_ptr;
381
            
382
	MPIU_INSTR_DURATION_START(rmaqueue_set);
383
	/* FIXME: For contig and very short operations, use a streamlined op */
384 385 386 387 388 389 390 391 392
	new_ptr->next = NULL;  
	new_ptr->type = MPIDI_RMA_GET;
	new_ptr->origin_addr = origin_addr;
	new_ptr->origin_count = origin_count;
	new_ptr->origin_datatype = origin_datatype;
	new_ptr->target_rank = target_rank;
	new_ptr->target_disp = target_disp;
	new_ptr->target_count = target_count;
	new_ptr->target_datatype = target_datatype;
393
	MPIU_INSTR_DURATION_END(rmaqueue_set);
394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434
	
	/* if source or target datatypes are derived, increment their
	   reference counts */ 
	MPIDI_CH3I_DATATYPE_IS_PREDEFINED(origin_datatype, predefined);
	if (!predefined)
	{
	    MPID_Datatype_get_ptr(origin_datatype, dtp);
	    MPID_Datatype_add_ref(dtp);
	}
	MPIDI_CH3I_DATATYPE_IS_PREDEFINED(target_datatype, predefined);
	if (!predefined)
	{
	    MPID_Datatype_get_ptr(target_datatype, dtp);
	    MPID_Datatype_add_ref(dtp);
	}
    }

  fn_exit:
    MPIDI_RMA_FUNC_EXIT(MPID_STATE_MPIDI_GET);
    return mpi_errno;

    /* --BEGIN ERROR HANDLING-- */
  fn_fail:
    MPIU_CHKPMEM_REAP();
    goto fn_exit;
    /* --END ERROR HANDLING-- */
}



#undef FUNCNAME
#define FUNCNAME MPIDI_Accumulate
#undef FCNAME
#define FCNAME MPIDI_QUOTE(FUNCNAME)
int MPIDI_Accumulate(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)
{
    int mpi_errno=MPI_SUCCESS;
    MPIDI_msg_sz_t data_sz;
Pavan Balaji's avatar
Pavan Balaji committed
435 436
    int dt_contig ATTRIBUTE((unused)), rank, origin_predefined, target_predefined;
    MPI_Aint dt_true_lb ATTRIBUTE((unused));
437
    MPIDI_RMA_ops *new_ptr;
438 439 440 441 442 443 444 445 446 447 448 449 450 451
    MPID_Datatype *dtp;
    MPIU_CHKLMEM_DECL(2);
    MPIU_CHKPMEM_DECL(1);
    MPIDI_STATE_DECL(MPID_STATE_MPIDI_ACCUMULATE);
    
    MPIDI_RMA_FUNC_ENTER(MPID_STATE_MPIDI_ACCUMULATE);

    MPIDI_Datatype_get_info(origin_count, origin_datatype,
			    dt_contig, data_sz, dtp, dt_true_lb);  
    
    if ((data_sz == 0) || (target_rank == MPI_PROC_NULL))
    {
	goto fn_exit;
    }
452 453

    rank = win_ptr->myrank;
454 455 456 457
    
    MPIDI_CH3I_DATATYPE_IS_PREDEFINED(origin_datatype, origin_predefined);
    MPIDI_CH3I_DATATYPE_IS_PREDEFINED(target_datatype, target_predefined);

458
    /* Do =! rank first (most likely branch?) */
459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476
    if (target_rank == rank)
    {
	MPI_User_function *uop;
	
	if (op == MPI_REPLACE)
	{
	    mpi_errno = MPIR_Localcopy(origin_addr, origin_count, 
				origin_datatype,
				(char *) win_ptr->base + win_ptr->disp_unit *
				target_disp, target_count, target_datatype); 
	    goto fn_exit;
	}
	
	MPIU_ERR_CHKANDJUMP1((HANDLE_GET_KIND(op) != HANDLE_KIND_BUILTIN), 
			     mpi_errno, MPI_ERR_OP, "**opnotpredefined",
			     "**opnotpredefined %d", op );
	
	/* get the function by indexing into the op table */
477
	uop = MPIR_Op_table[((op)&0xf) - 1];
478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501
	
	if (origin_predefined && target_predefined)
	{    
	    (*uop)(origin_addr, (char *) win_ptr->base + win_ptr->disp_unit *
		   target_disp, &target_count, &target_datatype);
	}
	else
	{
	    /* derived datatype */
	    
	    MPID_Segment *segp;
	    DLOOP_VECTOR *dloop_vec;
	    MPI_Aint first, last;
	    int vec_len, i, type_size, count;
	    MPI_Datatype type;
	    MPI_Aint true_lb, true_extent, extent;
	    void *tmp_buf=NULL, *source_buf, *target_buf;
	    
	    if (origin_datatype != target_datatype)
	    {
		/* first copy the data into a temporary buffer with
		   the same datatype as the target. Then do the
		   accumulate operation. */
		
502
		MPIR_Type_get_true_extent_impl(target_datatype, &true_lb, &true_extent);
503 504 505 506 507 508 509 510 511 512 513 514 515
		MPID_Datatype_get_extent_macro(target_datatype, extent); 
		
		MPIU_CHKLMEM_MALLOC(tmp_buf, void *, 
			target_count * (MPIR_MAX(extent,true_extent)), 
			mpi_errno, "temporary buffer");
		/* adjust for potential negative lower bound in datatype */
		tmp_buf = (void *)((char*)tmp_buf - true_lb);
		
		mpi_errno = MPIR_Localcopy(origin_addr, origin_count,
					   origin_datatype, tmp_buf,
					   target_count, target_datatype);  
		if (mpi_errno) { MPIU_ERR_POP(mpi_errno); }
	    }
516 517 518 519 520 521

	    if (target_predefined) { 
		/* target predefined type, origin derived datatype */

		(*uop)(tmp_buf, (char *) win_ptr->base + win_ptr->disp_unit *
		   target_disp, &target_count, &target_datatype);
522
	    }
523
	    else {
524
	    
525
		segp = MPID_Segment_alloc();
526 527
		MPIU_ERR_CHKANDJUMP1((!segp), mpi_errno, MPI_ERR_OTHER, 
				    "**nomem","**nomem %s","MPID_Segment_alloc"); 
528 529 530 531 532
		MPID_Segment_init(NULL, target_count, target_datatype, segp, 0);
		first = 0;
		last  = SEGMENT_IGNORE_LAST;
		
		MPID_Datatype_get_ptr(target_datatype, dtp);
533
		vec_len = dtp->max_contig_blocks * target_count + 1; 
534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555
		/* +1 needed because Rob says so */
		MPIU_CHKLMEM_MALLOC(dloop_vec, DLOOP_VECTOR *, 
				    vec_len * sizeof(DLOOP_VECTOR), 
				    mpi_errno, "dloop vector");
		
		MPID_Segment_pack_vector(segp, first, &last, dloop_vec, &vec_len);
		
		source_buf = (tmp_buf != NULL) ? tmp_buf : origin_addr;
		target_buf = (char *) win_ptr->base + 
		    win_ptr->disp_unit * target_disp;
		type = dtp->eltype;
		type_size = MPID_Datatype_get_basic_size(type);
		for (i=0; i<vec_len; i++)
		{
		    count = (dloop_vec[i].DLOOP_VECTOR_LEN)/type_size;
		    (*uop)((char *)source_buf + MPIU_PtrToAint(dloop_vec[i].DLOOP_VECTOR_BUF),
			   (char *)target_buf + MPIU_PtrToAint(dloop_vec[i].DLOOP_VECTOR_BUF),
			   &count, &type);
		}
		
		MPID_Segment_free(segp);
	    }
556 557 558 559 560
	}
    }
    else
    {
	/* queue it up */
561
	MPIU_INSTR_DURATION_START(rmaqueue_alloc);
562 563
	MPIU_CHKPMEM_MALLOC(new_ptr, MPIDI_RMA_ops *, sizeof(MPIDI_RMA_ops), 
			    mpi_errno, "RMA operation entry");
564 565 566
	MPIU_INSTR_DURATION_END(rmaqueue_alloc);
	if (win_ptr->rma_ops_list_tail) 
	    win_ptr->rma_ops_list_tail->next = new_ptr;
567
	else
568 569 570 571 572
	    win_ptr->rma_ops_list_head = new_ptr;
	win_ptr->rma_ops_list_tail = new_ptr;

	/* If predefined and contiguous, use a simplified element */
	if (origin_predefined && target_predefined && enableShortACC) {
573
	    MPIU_INSTR_DURATION_START(rmaqueue_set);
574 575 576 577 578 579 580 581 582 583 584
	    new_ptr->next = NULL;
	    new_ptr->type = MPIDI_RMA_ACC_CONTIG;
	    /* Only the information needed for the contig/predefined acc */
	    new_ptr->origin_addr = origin_addr;
	    new_ptr->origin_count = origin_count;
	    new_ptr->origin_datatype = origin_datatype;
	    new_ptr->target_rank = target_rank;
	    new_ptr->target_disp = target_disp;
	    new_ptr->target_count = target_count;
	    new_ptr->target_datatype = target_datatype;
	    new_ptr->op = op;
585
	    MPIU_INSTR_DURATION_END(rmaqueue_set);
586
	    goto fn_exit;
587
	}
588

589
	MPIU_INSTR_DURATION_START(rmaqueue_set);
590 591 592 593 594 595 596 597 598 599
	new_ptr->next = NULL;  
	new_ptr->type = MPIDI_RMA_ACCUMULATE;
	new_ptr->origin_addr = origin_addr;
	new_ptr->origin_count = origin_count;
	new_ptr->origin_datatype = origin_datatype;
	new_ptr->target_rank = target_rank;
	new_ptr->target_disp = target_disp;
	new_ptr->target_count = target_count;
	new_ptr->target_datatype = target_datatype;
	new_ptr->op = op;
600
	MPIU_INSTR_DURATION_END(rmaqueue_set);
601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662
	
	/* if source or target datatypes are derived, increment their
	   reference counts */ 
	if (!origin_predefined)
	{
	    MPID_Datatype_get_ptr(origin_datatype, dtp);
	    MPID_Datatype_add_ref(dtp);
	}
	if (!target_predefined)
	{
	    MPID_Datatype_get_ptr(target_datatype, dtp);
	    MPID_Datatype_add_ref(dtp);
	}
    }

 fn_exit:
    MPIU_CHKLMEM_FREEALL();
    MPIDI_RMA_FUNC_EXIT(MPID_STATE_MPIDI_ACCUMULATE);
    return mpi_errno;

    /* --BEGIN ERROR HANDLING-- */
  fn_fail:
    MPIU_CHKPMEM_REAP();
    goto fn_exit;
    /* --END ERROR HANDLING-- */
}


#undef FUNCNAME
#define FUNCNAME MPIDI_Alloc_mem
#undef FCNAME
#define FCNAME MPIDI_QUOTE(FUNCNAME)
void *MPIDI_Alloc_mem( size_t size, MPID_Info *info_ptr )
{
    void *ap;
    MPIDI_STATE_DECL(MPID_STATE_MPIDI_ALLOC_MEM);

    MPIDI_FUNC_ENTER(MPID_STATE_MPIDI_ALLOC_MEM);

    ap = MPIU_Malloc(size);
    
    MPIDI_FUNC_EXIT(MPID_STATE_MPIDI_ALLOC_MEM);
    return ap;
}


#undef FUNCNAME
#define FUNCNAME MPIDI_Free_mem
#undef FCNAME
#define FCNAME MPIDI_QUOTE(FUNCNAME)
int MPIDI_Free_mem( void *ptr )
{
    int mpi_errno = MPI_SUCCESS;
    MPIDI_STATE_DECL(MPID_STATE_MPIDI_FREE_MEM);

    MPIDI_FUNC_ENTER(MPID_STATE_MPIDI_FREE_MEM);

    MPIU_Free(ptr);
    
    MPIDI_FUNC_EXIT(MPID_STATE_MPIDI_FREE_MEM);
    return mpi_errno;
}