1 2 3 4 5 6 7 8 9 10 11 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 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396
|
/* -*- Mode: C; c-basic-offset:4 ; indent-tabs-mode:nil -*- */
/*
* Copyright (c) 2004-2005 The Trustees of Indiana University and Indiana
* University Research and Technology
* Corporation. All rights reserved.
* Copyright (c) 2004-2007 The University of Tennessee and The University
* of Tennessee Research Foundation. All rights
* reserved.
* Copyright (c) 2004-2005 High Performance Computing Center Stuttgart,
* University of Stuttgart. All rights reserved.
* Copyright (c) 2004-2005 The Regents of the University of California.
* All rights reserved.
* Copyright (c) 2007 Sun Microsystems, Inc. All rights reserved.
* Copyright (c) 2015-2017 Research Organization for Information Science
* and Technology (RIST). All rights reserved.
* Copyright (c) 2015 Los Alamos National Security, LLC. All rights
* reserved.
* Copyright (c) 2017 IBM Corporation. All rights reserved.
* Copyright (c) 2018 Triad National Security, LLC. All rights
* reserved.
* $COPYRIGHT$
*
* Additional copyrights may follow
*
* $HEADER$
*/
#include "ompi_config.h"
#include "opal/mca/threads/mutex.h"
#include "opal/mca/threads/condition.h"
#include "ompi/datatype/ompi_datatype.h"
#include "opal/mca/allocator/base/base.h"
#include "opal/mca/allocator/allocator.h"
#include "ompi/mca/pml/pml.h"
#include "ompi/mca/pml/base/pml_base_request.h"
#include "ompi/mca/pml/base/pml_base_sendreq.h"
#include "ompi/mca/pml/base/pml_base_bsend.h"
#include "opal/mca/mpool/mpool.h"
#include "ompi/runtime/mpiruntime.h"
#ifdef HAVE_UNISTD_H
#include <unistd.h>
#endif /* HAVE_UNISTD_H */
static opal_mutex_t mca_pml_bsend_mutex; /* lock for thread safety */
static opal_condition_t mca_pml_bsend_condition; /* condition variable to block on detach */
static mca_allocator_base_component_t* mca_pml_bsend_allocator_component;
static mca_allocator_base_module_t* mca_pml_bsend_allocator; /* sub-allocator to manage users buffer */
static size_t mca_pml_bsend_usersize; /* user provided buffer size */
unsigned char *mca_pml_bsend_userbase=NULL;/* user provided buffer base */
unsigned char *mca_pml_bsend_base = NULL;/* adjusted base of user buffer */
unsigned char *mca_pml_bsend_addr = NULL;/* current offset into user buffer */
static size_t mca_pml_bsend_size; /* adjusted size of user buffer */
static size_t mca_pml_bsend_count; /* number of outstanding requests */
static size_t mca_pml_bsend_pagesz; /* mmap page size */
static int mca_pml_bsend_pagebits; /* number of bits in pagesz */
static opal_atomic_int32_t mca_pml_bsend_init = 0;
/* defined in pml_base_open.c */
extern char *ompi_pml_base_bsend_allocator_name;
static int mca_pml_base_bsend_fini (void);
/*
* Routine to return pages to sub-allocator as needed
*/
static void* mca_pml_bsend_alloc_segment(void *ctx, size_t *size_inout)
{
void *addr;
size_t size = *size_inout;
if(mca_pml_bsend_addr + size > mca_pml_bsend_base + mca_pml_bsend_size) {
return NULL;
}
/* allocate all that is left */
size = mca_pml_bsend_size - (mca_pml_bsend_addr - mca_pml_bsend_base);
addr = mca_pml_bsend_addr;
mca_pml_bsend_addr += size;
*size_inout = size;
return addr;
}
/*
* One time initialization at startup
*/
int mca_pml_base_bsend_init (void)
{
size_t tmp;
if(OPAL_THREAD_ADD_FETCH32(&mca_pml_bsend_init, 1) > 1)
return OMPI_SUCCESS;
/* initialize static objects */
OBJ_CONSTRUCT(&mca_pml_bsend_mutex, opal_mutex_t);
OBJ_CONSTRUCT(&mca_pml_bsend_condition, opal_condition_t);
/* lookup name of the allocator to use for buffered sends */
if(NULL == (mca_pml_bsend_allocator_component = mca_allocator_component_lookup(ompi_pml_base_bsend_allocator_name))) {
return OMPI_ERR_BUFFER;
}
/* determine page size */
tmp = mca_pml_bsend_pagesz = sysconf(_SC_PAGESIZE);
mca_pml_bsend_pagebits = 0;
while( tmp != 0 ) {
tmp >>= 1;
mca_pml_bsend_pagebits++;
}
ompi_mpi_instance_append_finalize (mca_pml_base_bsend_fini);
return OMPI_SUCCESS;
}
/*
* One-time cleanup at shutdown - release any resources.
*/
static int mca_pml_base_bsend_fini (void)
{
if(OPAL_THREAD_ADD_FETCH32(&mca_pml_bsend_init,-1) > 0)
return OMPI_SUCCESS;
if(NULL != mca_pml_bsend_allocator)
mca_pml_bsend_allocator->alc_finalize(mca_pml_bsend_allocator);
mca_pml_bsend_allocator = NULL;
OBJ_DESTRUCT(&mca_pml_bsend_condition);
OBJ_DESTRUCT(&mca_pml_bsend_mutex);
return OMPI_SUCCESS;
}
/*
* User-level call to attach buffer.
*/
int mca_pml_base_bsend_attach(void* addr, int size)
{
int align;
if(NULL == addr || size <= 0) {
return OMPI_ERR_BUFFER;
}
/* check for buffer already attached */
OPAL_THREAD_LOCK(&mca_pml_bsend_mutex);
if(NULL != mca_pml_bsend_allocator) {
OPAL_THREAD_UNLOCK(&mca_pml_bsend_mutex);
return OMPI_ERR_BUFFER;
}
/* try to create an instance of the allocator - to determine thread safety level */
mca_pml_bsend_allocator = mca_pml_bsend_allocator_component->allocator_init(ompi_mpi_thread_multiple, mca_pml_bsend_alloc_segment, NULL, NULL);
if(NULL == mca_pml_bsend_allocator) {
OPAL_THREAD_UNLOCK(&mca_pml_bsend_mutex);
return OMPI_ERR_BUFFER;
}
/*
* Save away what the user handed in. This is done in case the
* base and size are modified for alignment issues.
*/
mca_pml_bsend_userbase = (unsigned char*)addr;
mca_pml_bsend_usersize = size;
/*
* Align to pointer boundaries. The bsend overhead is large enough
* to account for this. Compute any alignment that needs to be done.
*/
align = sizeof(void *) - ((size_t)addr & (sizeof(void *) - 1));
/* setup local variables */
mca_pml_bsend_base = (unsigned char *)addr + align;
mca_pml_bsend_addr = (unsigned char *)addr + align;
mca_pml_bsend_size = size - align;
mca_pml_bsend_count = 0;
OPAL_THREAD_UNLOCK(&mca_pml_bsend_mutex);
return OMPI_SUCCESS;
}
/*
* User-level call to detach buffer
*/
int mca_pml_base_bsend_detach(void* addr, int* size)
{
OPAL_THREAD_LOCK(&mca_pml_bsend_mutex);
/* is buffer attached */
if(NULL == mca_pml_bsend_allocator) {
OPAL_THREAD_UNLOCK(&mca_pml_bsend_mutex);
return OMPI_ERR_BUFFER;
}
/* wait on any pending requests */
while(mca_pml_bsend_count != 0)
opal_condition_wait(&mca_pml_bsend_condition, &mca_pml_bsend_mutex);
/* free resources associated with the allocator */
mca_pml_bsend_allocator->alc_finalize(mca_pml_bsend_allocator);
mca_pml_bsend_allocator = NULL;
/* return current settings */
if(NULL != addr)
*((void**)addr) = mca_pml_bsend_userbase;
if(NULL != size)
*size = (int)mca_pml_bsend_usersize;
/* reset local variables */
mca_pml_bsend_userbase = NULL;
mca_pml_bsend_usersize = 0;
mca_pml_bsend_base = NULL;
mca_pml_bsend_addr = NULL;
mca_pml_bsend_size = 0;
mca_pml_bsend_count = 0;
OPAL_THREAD_UNLOCK(&mca_pml_bsend_mutex);
return OMPI_SUCCESS;
}
/*
* pack send buffer into buffer
*/
int mca_pml_base_bsend_request_start(ompi_request_t* request)
{
mca_pml_base_send_request_t* sendreq = (mca_pml_base_send_request_t*)request;
struct iovec iov;
unsigned int iov_count;
size_t max_data;
int rc;
if(sendreq->req_bytes_packed > 0) {
/* has a buffer been provided */
OPAL_THREAD_LOCK(&mca_pml_bsend_mutex);
if(NULL == mca_pml_bsend_addr) {
sendreq->req_addr = NULL;
OPAL_THREAD_UNLOCK(&mca_pml_bsend_mutex);
return OMPI_ERR_BUFFER;
}
/* allocate a buffer to hold packed message */
sendreq->req_addr = mca_pml_bsend_allocator->alc_alloc(
mca_pml_bsend_allocator, sendreq->req_bytes_packed, 0);
if(NULL == sendreq->req_addr) {
/* release resources when request is freed */
sendreq->req_base.req_pml_complete = true;
OPAL_THREAD_UNLOCK(&mca_pml_bsend_mutex);
return OMPI_ERR_BUFFER;
}
OPAL_THREAD_UNLOCK(&mca_pml_bsend_mutex);
/* The convertor is already initialized in the beginning so we just have to
* pack the data in the newly allocated buffer.
*/
iov.iov_base = (IOVBASE_TYPE*)sendreq->req_addr;
iov.iov_len = sendreq->req_bytes_packed;
iov_count = 1;
max_data = iov.iov_len;
if((rc = opal_convertor_pack( &sendreq->req_base.req_convertor,
&iov,
&iov_count,
&max_data )) < 0) {
return OMPI_ERROR;
}
/* setup convertor to point to packed buffer (at position zero) */
opal_convertor_prepare_for_send( &sendreq->req_base.req_convertor, &(ompi_mpi_packed.dt.super),
max_data, sendreq->req_addr );
/* increment count of pending requests */
mca_pml_bsend_count++;
}
return OMPI_SUCCESS;
}
/*
* allocate buffer
*/
int mca_pml_base_bsend_request_alloc(ompi_request_t* request)
{
mca_pml_base_send_request_t* sendreq = (mca_pml_base_send_request_t*)request;
assert( sendreq->req_bytes_packed > 0 );
/* has a buffer been provided */
OPAL_THREAD_LOCK(&mca_pml_bsend_mutex);
if(NULL == mca_pml_bsend_addr) {
sendreq->req_addr = NULL;
OPAL_THREAD_UNLOCK(&mca_pml_bsend_mutex);
return OMPI_ERR_BUFFER;
}
/* allocate a buffer to hold packed message */
sendreq->req_addr = mca_pml_bsend_allocator->alc_alloc(
mca_pml_bsend_allocator, sendreq->req_bytes_packed, 0);
if(NULL == sendreq->req_addr) {
/* release resources when request is freed */
sendreq->req_base.req_pml_complete = true;
OPAL_THREAD_UNLOCK(&mca_pml_bsend_mutex);
/* progress communications, with the hope that more resources
* will be freed */
opal_progress();
return OMPI_ERR_BUFFER;
}
/* increment count of pending requests */
mca_pml_bsend_count++;
OPAL_THREAD_UNLOCK(&mca_pml_bsend_mutex);
return OMPI_SUCCESS;
}
/*
* allocate buffer
*/
void* mca_pml_base_bsend_request_alloc_buf( size_t length )
{
void* buf = NULL;
/* has a buffer been provided */
OPAL_THREAD_LOCK(&mca_pml_bsend_mutex);
if(NULL == mca_pml_bsend_addr) {
OPAL_THREAD_UNLOCK(&mca_pml_bsend_mutex);
return NULL;
}
/* allocate a buffer to hold packed message */
buf = mca_pml_bsend_allocator->alc_alloc(
mca_pml_bsend_allocator, length, 0);
if(NULL == buf) {
/* release resources when request is freed */
OPAL_THREAD_UNLOCK(&mca_pml_bsend_mutex);
/* progress communications, with the hope that more resources
* will be freed */
opal_progress();
return NULL;
}
/* increment count of pending requests */
mca_pml_bsend_count++;
OPAL_THREAD_UNLOCK(&mca_pml_bsend_mutex);
return buf;
}
/*
* Request completed - free buffer and decrement pending count
*/
int mca_pml_base_bsend_request_free(void* addr)
{
/* remove from list of pending requests */
OPAL_THREAD_LOCK(&mca_pml_bsend_mutex);
/* free buffer */
mca_pml_bsend_allocator->alc_free(mca_pml_bsend_allocator, addr);
/* decrement count of buffered requests */
if(--mca_pml_bsend_count == 0)
opal_condition_signal(&mca_pml_bsend_condition);
OPAL_THREAD_UNLOCK(&mca_pml_bsend_mutex);
return OMPI_SUCCESS;
}
/*
* Request completed - free buffer and decrement pending count
*/
int mca_pml_base_bsend_request_fini(ompi_request_t* request)
{
mca_pml_base_send_request_t* sendreq = (mca_pml_base_send_request_t*)request;
if(sendreq->req_bytes_packed == 0 ||
sendreq->req_addr == NULL ||
sendreq->req_addr == sendreq->req_base.req_addr)
return OMPI_SUCCESS;
/* remove from list of pending requests */
OPAL_THREAD_LOCK(&mca_pml_bsend_mutex);
/* free buffer */
mca_pml_bsend_allocator->alc_free(mca_pml_bsend_allocator, (void *)sendreq->req_addr);
sendreq->req_addr = sendreq->req_base.req_addr;
/* decrement count of buffered requests */
if(--mca_pml_bsend_count == 0)
opal_condition_signal(&mca_pml_bsend_condition);
OPAL_THREAD_UNLOCK(&mca_pml_bsend_mutex);
return OMPI_SUCCESS;
}
|