1
1
openmpi/ompi/mca/osc/rdma/osc_rdma_comm.c

296 строки
11 KiB
C
Исходник Обычный вид История

/*
* Copyright (c) 2004-2005 The Trustees of Indiana University.
* All rights reserved.
* Copyright (c) 2004-2005 The Trustees of the University of Tennessee.
* 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 Los Alamos National Security, LLC. All rights
* reserved.
* Copyright (c) 2010 Cisco Systems, Inc. All rights reserved.
* $COPYRIGHT$
*
* Additional copyrights may follow
*
* $HEADER$
*/
#include "ompi_config.h"
#include "mpi.h"
#include <stdio.h>
#include "osc_rdma.h"
#include "osc_rdma_sendreq.h"
#include "osc_rdma_header.h"
#include "osc_rdma_data_move.h"
#include "ompi/memchecker.h"
#include "ompi/mca/osc/base/osc_base_obj_convert.h"
#include "opal_stdint.h"
static int
enqueue_sendreq(ompi_osc_rdma_module_t *module,
ompi_osc_rdma_sendreq_t *sendreq)
{
OPAL_THREAD_LOCK(&(module->m_lock));
opal_list_append(&(module->m_pending_sendreqs),
(opal_list_item_t*) sendreq);
module->m_num_pending_sendreqs[sendreq->req_target_rank]++;
OPAL_THREAD_UNLOCK(&(module->m_lock));
return OMPI_SUCCESS;
}
int
ompi_osc_rdma_module_accumulate(void *origin_addr, int origin_count,
struct ompi_datatype_t *origin_dt,
int target, OPAL_PTRDIFF_TYPE target_disp,
int target_count,
struct ompi_datatype_t *target_dt,
struct ompi_op_t *op, ompi_win_t *win)
{
int ret;
ompi_osc_rdma_sendreq_t *sendreq;
ompi_osc_rdma_module_t *module = GET_MODULE(win);
if ((OMPI_WIN_STARTED & ompi_win_get_mode(win)) &&
(!module->m_sc_remote_active_ranks[target])) {
return MPI_ERR_RMA_SYNC;
}
if (OMPI_WIN_FENCE & ompi_win_get_mode(win)) {
/* well, we're definitely in an access epoch now */
ompi_win_set_mode(win, OMPI_WIN_FENCE | OMPI_WIN_ACCESS_EPOCH |
OMPI_WIN_EXPOSE_EPOCH);
}
/* shortcut 0 count case */
if (0 == origin_count || 0 == target_count) {
return OMPI_SUCCESS;
}
/* create sendreq */
ret = ompi_osc_rdma_sendreq_alloc_init(OMPI_OSC_RDMA_ACC,
origin_addr,
origin_count,
origin_dt,
target,
target_disp,
target_count,
target_dt,
module,
&sendreq);
MEMCHECKER(
memchecker_convertor_call(&opal_memchecker_base_mem_noaccess,
&sendreq->req_origin_convertor);
);
if (OMPI_SUCCESS != ret) return ret;
sendreq->req_op_id = op->o_f_to_c_index;
if (module->m_eager_send_active) {
/* accumulate semantics require send to self, which is bloody
expensive with the extra copies. Put a shortcut in for the
common case. */
if (target == ompi_comm_rank(sendreq->req_module->m_comm) &&
- Split the datatype engine into two parts: an MPI specific part in OMPI and a language agnostic part in OPAL. The convertor is completely moved into OPAL. This offers several benefits as described in RFC http://www.open-mpi.org/community/lists/devel/2009/07/6387.php namely: - Fewer basic types (int* and float* types, boolean and wchar - Fixing naming scheme to ompi-nomenclature. - Usability outside of the ompi-layer. - Due to the fixed nature of simple opal types, their information is completely known at compile time and therefore constified - With fewer datatypes (22), the actual sizes of bit-field types may be reduced from 64 to 32 bits, allowing reorganizing the opal_datatype structure, eliminating holes and keeping data required in convertor (upon send/recv) in one cacheline... This has implications to the convertor-datastructure and other parts of the code. - Several performance tests have been run, the netpipe latency does not change with this patch on Linux/x86-64 on the smoky cluster. - Extensive tests have been done to verify correctness (no new regressions) using: 1. mpi_test_suite on linux/x86-64 using clean ompi-trunk and ompi-ddt: a. running both trunk and ompi-ddt resulted in no differences (except for MPI_SHORT_INT and MPI_TYPE_MIX_LB_UB do now run correctly). b. with --enable-memchecker and running under valgrind (one buglet when run with static found in test-suite, commited) 2. ibm testsuite on linux/x86-64 using clean ompi-trunk and ompi-ddt: all passed (except for the dynamic/ tests failed!! as trunk/MTT) 3. compilation and usage of HDF5 tests on Jaguar using PGI and PathScale compilers. 4. compilation and usage on Scicortex. - Please note, that for the heterogeneous case, (-m32 compiled binaries/ompi), neither ompi-trunk, nor ompi-ddt branch would successfully launch. This commit was SVN r21641.
2009-07-13 08:56:31 +04:00
ompi_datatype_is_contiguous_memory_layout(sendreq->req_target_datatype,
sendreq->req_target_count) &&
- Split the datatype engine into two parts: an MPI specific part in OMPI and a language agnostic part in OPAL. The convertor is completely moved into OPAL. This offers several benefits as described in RFC http://www.open-mpi.org/community/lists/devel/2009/07/6387.php namely: - Fewer basic types (int* and float* types, boolean and wchar - Fixing naming scheme to ompi-nomenclature. - Usability outside of the ompi-layer. - Due to the fixed nature of simple opal types, their information is completely known at compile time and therefore constified - With fewer datatypes (22), the actual sizes of bit-field types may be reduced from 64 to 32 bits, allowing reorganizing the opal_datatype structure, eliminating holes and keeping data required in convertor (upon send/recv) in one cacheline... This has implications to the convertor-datastructure and other parts of the code. - Several performance tests have been run, the netpipe latency does not change with this patch on Linux/x86-64 on the smoky cluster. - Extensive tests have been done to verify correctness (no new regressions) using: 1. mpi_test_suite on linux/x86-64 using clean ompi-trunk and ompi-ddt: a. running both trunk and ompi-ddt resulted in no differences (except for MPI_SHORT_INT and MPI_TYPE_MIX_LB_UB do now run correctly). b. with --enable-memchecker and running under valgrind (one buglet when run with static found in test-suite, commited) 2. ibm testsuite on linux/x86-64 using clean ompi-trunk and ompi-ddt: all passed (except for the dynamic/ tests failed!! as trunk/MTT) 3. compilation and usage of HDF5 tests on Jaguar using PGI and PathScale compilers. 4. compilation and usage on Scicortex. - Please note, that for the heterogeneous case, (-m32 compiled binaries/ompi), neither ompi-trunk, nor ompi-ddt branch would successfully launch. This commit was SVN r21641.
2009-07-13 08:56:31 +04:00
!opal_convertor_need_buffers(&sendreq->req_origin_convertor) &&
0 == OPAL_THREAD_TRYLOCK(&module->m_acc_lock)) {
void *target_buffer = (unsigned char*) module->m_win->w_baseptr +
((unsigned long) target_disp *
module->m_win->w_disp_unit);
struct iovec iov;
uint32_t iov_count = 1;
size_t max_data = sendreq->req_origin_bytes_packed;
iov.iov_len = max_data;
iov.iov_base = NULL;
- Split the datatype engine into two parts: an MPI specific part in OMPI and a language agnostic part in OPAL. The convertor is completely moved into OPAL. This offers several benefits as described in RFC http://www.open-mpi.org/community/lists/devel/2009/07/6387.php namely: - Fewer basic types (int* and float* types, boolean and wchar - Fixing naming scheme to ompi-nomenclature. - Usability outside of the ompi-layer. - Due to the fixed nature of simple opal types, their information is completely known at compile time and therefore constified - With fewer datatypes (22), the actual sizes of bit-field types may be reduced from 64 to 32 bits, allowing reorganizing the opal_datatype structure, eliminating holes and keeping data required in convertor (upon send/recv) in one cacheline... This has implications to the convertor-datastructure and other parts of the code. - Several performance tests have been run, the netpipe latency does not change with this patch on Linux/x86-64 on the smoky cluster. - Extensive tests have been done to verify correctness (no new regressions) using: 1. mpi_test_suite on linux/x86-64 using clean ompi-trunk and ompi-ddt: a. running both trunk and ompi-ddt resulted in no differences (except for MPI_SHORT_INT and MPI_TYPE_MIX_LB_UB do now run correctly). b. with --enable-memchecker and running under valgrind (one buglet when run with static found in test-suite, commited) 2. ibm testsuite on linux/x86-64 using clean ompi-trunk and ompi-ddt: all passed (except for the dynamic/ tests failed!! as trunk/MTT) 3. compilation and usage of HDF5 tests on Jaguar using PGI and PathScale compilers. 4. compilation and usage on Scicortex. - Please note, that for the heterogeneous case, (-m32 compiled binaries/ompi), neither ompi-trunk, nor ompi-ddt branch would successfully launch. This commit was SVN r21641.
2009-07-13 08:56:31 +04:00
ret = opal_convertor_pack(&sendreq->req_origin_convertor,
&iov, &iov_count,
&max_data);
if (ret < 0) {
OPAL_THREAD_UNLOCK(&module->m_acc_lock);
return OMPI_ERR_FATAL;
}
ret = ompi_osc_base_process_op(target_buffer,
iov.iov_base,
max_data,
target_dt,
target_count,
op);
/* unlock the window for accumulates */
OPAL_THREAD_UNLOCK(&module->m_acc_lock);
ompi_osc_rdma_sendreq_free(sendreq);
return ret;
}
OPAL_THREAD_LOCK(&module->m_lock);
sendreq->req_module->m_num_pending_out += 1;
module->m_num_pending_sendreqs[sendreq->req_target_rank] += 1;
OPAL_THREAD_UNLOCK(&(module->m_lock));
ret = ompi_osc_rdma_sendreq_send(module, sendreq);
if (OMPI_ERR_TEMP_OUT_OF_RESOURCE == OPAL_SOS_GET_ERROR_CODE(ret)) {
OPAL_THREAD_LOCK(&module->m_lock);
sendreq->req_module->m_num_pending_out -= 1;
opal_list_append(&(module->m_pending_sendreqs),
(opal_list_item_t*) sendreq);
OPAL_THREAD_UNLOCK(&module->m_lock);
ret = OMPI_SUCCESS;
}
} else {
/* enqueue sendreq */
ret = enqueue_sendreq(module, sendreq);
}
return ret;
}
int
ompi_osc_rdma_module_get(void *origin_addr,
int origin_count,
struct ompi_datatype_t *origin_dt,
int target,
OPAL_PTRDIFF_TYPE target_disp,
int target_count,
struct ompi_datatype_t *target_dt,
ompi_win_t *win)
{
int ret;
ompi_osc_rdma_sendreq_t *sendreq;
ompi_osc_rdma_module_t *module = GET_MODULE(win);
if ((OMPI_WIN_STARTED & ompi_win_get_mode(win)) &&
(!module->m_sc_remote_active_ranks[target])) {
return MPI_ERR_RMA_SYNC;
}
if (OMPI_WIN_FENCE & ompi_win_get_mode(win)) {
/* well, we're definitely in an access epoch now */
ompi_win_set_mode(win, OMPI_WIN_FENCE | OMPI_WIN_ACCESS_EPOCH |
OMPI_WIN_EXPOSE_EPOCH);
}
/* shortcut 0 count case */
if (0 == origin_count || 0 == target_count) {
return OMPI_SUCCESS;
}
/* create sendreq */
ret = ompi_osc_rdma_sendreq_alloc_init(OMPI_OSC_RDMA_GET,
origin_addr,
origin_count,
origin_dt,
target,
target_disp,
target_count,
target_dt,
module,
&sendreq);
MEMCHECKER(
memchecker_convertor_call(&opal_memchecker_base_mem_noaccess,
&sendreq->req_origin_convertor);
);
if (OMPI_SUCCESS != ret) return ret;
if (module->m_eager_send_active) {
OPAL_THREAD_LOCK(&module->m_lock);
sendreq->req_module->m_num_pending_out += 1;
module->m_num_pending_sendreqs[sendreq->req_target_rank] += 1;
OPAL_THREAD_UNLOCK(&(module->m_lock));
ret = ompi_osc_rdma_sendreq_send(module, sendreq);
if (OMPI_ERR_TEMP_OUT_OF_RESOURCE == OPAL_SOS_GET_ERROR_CODE(ret)) {
OPAL_THREAD_LOCK(&module->m_lock);
sendreq->req_module->m_num_pending_out -= 1;
opal_list_append(&(module->m_pending_sendreqs),
(opal_list_item_t*) sendreq);
OPAL_THREAD_UNLOCK(&module->m_lock);
ret = OMPI_SUCCESS;
}
} else {
/* enqueue sendreq */
ret = enqueue_sendreq(module, sendreq);
}
return ret;
}
int
ompi_osc_rdma_module_put(void *origin_addr, int origin_count,
struct ompi_datatype_t *origin_dt,
int target, OPAL_PTRDIFF_TYPE target_disp,
int target_count,
struct ompi_datatype_t *target_dt, ompi_win_t *win)
{
int ret;
ompi_osc_rdma_sendreq_t *sendreq;
ompi_osc_rdma_module_t *module = GET_MODULE(win);
if ((OMPI_WIN_STARTED & ompi_win_get_mode(win)) &&
(!module->m_sc_remote_active_ranks[target])) {
return MPI_ERR_RMA_SYNC;
}
if (OMPI_WIN_FENCE & ompi_win_get_mode(win)) {
/* well, we're definitely in an access epoch now */
ompi_win_set_mode(win, OMPI_WIN_FENCE | OMPI_WIN_ACCESS_EPOCH |
OMPI_WIN_EXPOSE_EPOCH);
}
/* shortcut 0 count case */
if (0 == origin_count || 0 == target_count) {
return OMPI_SUCCESS;
}
/* create sendreq */
ret = ompi_osc_rdma_sendreq_alloc_init(OMPI_OSC_RDMA_PUT,
origin_addr,
origin_count,
origin_dt,
target,
target_disp,
target_count,
target_dt,
module,
&sendreq);
MEMCHECKER(
memchecker_convertor_call(&opal_memchecker_base_mem_noaccess,
&sendreq->req_origin_convertor);
);
if (OMPI_SUCCESS != ret) return ret;
if (module->m_eager_send_active) {
OPAL_THREAD_LOCK(&module->m_lock);
sendreq->req_module->m_num_pending_out += 1;
module->m_num_pending_sendreqs[sendreq->req_target_rank] += 1;
OPAL_THREAD_UNLOCK(&(module->m_lock));
ret = ompi_osc_rdma_sendreq_send(module, sendreq);
if (OMPI_ERR_TEMP_OUT_OF_RESOURCE == OPAL_SOS_GET_ERROR_CODE(ret)) {
OPAL_THREAD_LOCK(&module->m_lock);
sendreq->req_module->m_num_pending_out -= 1;
opal_list_append(&(module->m_pending_sendreqs),
(opal_list_item_t*) sendreq);
OPAL_THREAD_UNLOCK(&module->m_lock);
ret = OMPI_SUCCESS;
}
} else {
/* enqueue sendreq */
ret = enqueue_sendreq(module, sendreq);
}
return ret;
}