ports/openmpi
ports/openmpi
1
1
Форкнуть 0
Код Релизы Активность

Merge pull request #4323 from aravindksg/fix_help_text

Просмотр исходного кода 
Move help text output regarding PSM2_CUDA environment variable
Этот коммит содержится в:
Ralph Castain 2017-10-27 10:10:01 -05:00 коммит произвёл GitHub
родитель 5d208a177d bea4503f95
Коммит df48ddd2a1
2 изменённых файлов: 55 добавлений и 35 удалений
Показать статистику Скачать Patch файл Скачать Diff файл Показать все файлы Свернуть все файлы

18
ompi/mca/mtl/psm2/help-mtl-psm2.txt
Просмотреть файл

@ -1,7 +1,7 @@
# -*- text -*-
#
# Copyright (C) 2009. QLogic Corporation. All rights reserved.
# Copyright (c) 2013-2015 Intel, Inc. All rights reserved.
# Copyright (c) 2013-2017 Intel, Inc. All rights reserved.
# $COPYRIGHT$
#
# Additional copyrights may follow
@ -47,5 +47,17 @@ Unknown path record query mechanism %s. Supported mechanisms are %s.
Message size %llu bigger than supported by PSM2 API. Max = %llu
#
[no psm2 cuda env]
Using CUDA enabled OpenMPI but PSM2_CUDA environment variable is %s.
This is not a recommended combination. If the application uses %s.
Warning: Open MPI has detected that you are running in an environment with CUDA
devices present and that you are using Intel(r) Ompi-Path networking. However,
the environment variable PSM2_CUDA was not set, meaning that the PSM2 Omni-Path
networking library was not told how to handle CUDA support.
If your application uses CUDA buffers, you should set the environment variable
PSM2_CUDA to 1; otherwise, set it to 0. Setting the variable to the wrong value
can have performance implications on your application, or even cause it to
crash.
Since it was not set, Open MPI has defaulted to setting the PSM2_CUDA
environment variable to 1.
Local hostname: %s

72
ompi/mca/mtl/psm2/mtl_psm2_component.c
Просмотреть файл

@ -13,7 +13,7 @@
* Copyright (c) 2006-2010 QLogic Corporation. All rights reserved.
* Copyright (c) 2012-2017 Los Alamos National Security, LLC. All rights
* reserved.
* Copyright (c) 2013-2015 Intel, Inc. All rights reserved
* Copyright (c) 2013-2017 Intel, Inc. All rights reserved
* Copyright (c) 2017 Research Organization for Information Science
* and Technology (RIST). All rights reserved.
* $COPYRIGHT$
@ -28,6 +28,7 @@
#include "opal/mca/event/event.h"
#include "opal/util/output.h"
#include "opal/util/show_help.h"
#include "opal/util/opal_environ.h"
#include "ompi/proc/proc.h"
#include "mtl_psm2.h"
@ -45,6 +46,10 @@ static int param_priority;
/* MPI_THREAD_MULTIPLE_SUPPORT */
opal_mutex_t mtl_psm2_mq_mutex = OPAL_MUTEX_STATIC_INIT;
#if OPAL_CUDA_SUPPORT
static bool cuda_envvar_set = false;
#endif
static int ompi_mtl_psm2_component_open(void);
static int ompi_mtl_psm2_component_close(void);
static int ompi_mtl_psm2_component_query(mca_base_module_t **module, int *priority);
@ -201,9 +206,6 @@ static int
ompi_mtl_psm2_component_register(void)
{
int num_local_procs, num_total_procs;
#if OPAL_CUDA_SUPPORT
char *cuda_env;
#endif
ompi_mtl_psm2.connect_timeout = 180;
(void) mca_base_component_var_register(&mca_mtl_psm2_component.super.mtl_version,
@ -228,30 +230,6 @@ ompi_mtl_psm2_component_register(void)
param_priority = 40;
}
#if OPAL_CUDA_SUPPORT
/*
* If using CUDA enabled OpenMPI, the user likely intends to
* run with CUDA buffers. So, force-set the envvar here if user failed
* to set it.
*/
cuda_env = getenv("PSM2_CUDA");
if (!cuda_env) {
opal_show_help("help-mtl-psm2.txt",
"no psm2 cuda env", true,
"not set",
"Host buffers,\nthere will be a performance penalty"
" due to OMPI force setting this variable now.\n"
"Set environment variable to 0 if using Host buffers" );
setenv("PSM2_CUDA", "1", 0);
} else if (strcmp(cuda_env, "0") == 0) {
opal_show_help("help-mtl-psm2.txt",
"no psm2 cuda env", true,
"set to 0",
"CUDA buffers,\nthe execution will SEGFAULT."
" Set environment variable to 1 if using CUDA buffers");
}
#endif
(void) mca_base_component_var_register (&mca_mtl_psm2_component.super.mtl_version,
"priority", "Priority of the PSM2 MTL component",
MCA_BASE_VAR_TYPE_INT, NULL, 0, 0,
@ -272,17 +250,16 @@ static int
ompi_mtl_psm2_component_open(void)
{
int res;
glob_t globbuf;
globbuf.gl_offs = 0;
glob_t globbuf = {0};
/* Component available only if Omni-Path hardware is present */
res = glob("/dev/hfi1_[0-9]", GLOB_DOOFFS, NULL, &globbuf);
if (0 == res || GLOB_NOMATCH == res) {
if (globbuf.gl_pathc > 0) {
globfree(&globbuf);
}
if (0 != res) {
res = glob("/dev/hfi1_[0-9][0-9]", GLOB_APPEND, NULL, &globbuf);
if (0 == res || GLOB_NOMATCH == res) {
if (globbuf.gl_pathc > 0) {
globfree(&globbuf);
}
if (0 != res) {
@ -336,6 +313,11 @@ ompi_mtl_psm2_component_query(mca_base_module_t **module, int *priority)
static int
ompi_mtl_psm2_component_close(void)
{
#if OPAL_CUDA_SUPPORT
if (cuda_envvar_set) {
opal_unsetenv("PSM2_CUDA", &environ);
}
#endif
return OMPI_SUCCESS;
}
@ -362,6 +344,11 @@ ompi_mtl_psm2_component_init(bool enable_progress_threads,
int verno_major = PSM2_VERNO_MAJOR;
int verno_minor = PSM2_VERNO_MINOR;
int local_rank = -1, num_local_procs = 0;
#if OPAL_CUDA_SUPPORT
int ret;
char *cuda_env;
glob_t globbuf = {0};
#endif
/* Compute the total number of processes on this host and our local rank
* on that node. We need to provide PSM2 with these values so it can
@ -389,6 +376,27 @@ ompi_mtl_psm2_component_init(bool enable_progress_threads,
ompi_mtl_psm2_set_shadow_env (ompi_mtl_psm2_shadow_variables + i);
}
#if OPAL_CUDA_SUPPORT
/*
* If using CUDA enabled Open MPI, the user likely intends to
* run with CUDA buffers. So, force-set the envvar here if user failed
* to set it.
*/
ret = glob("/sys/module/nvidia", GLOB_DOOFFS, NULL, &globbuf);
if (globbuf.gl_pathc > 0) {
globfree(&globbuf);
}
cuda_env = getenv("PSM2_CUDA");
if (!cuda_env && (0 == ret)) {
opal_show_help("help-mtl-psm2.txt",
"no psm2 cuda env", true,
ompi_process_info.nodename);
opal_setenv("PSM2_CUDA", "1", false, &environ);
cuda_envvar_set = true;
}
#endif
err = psm2_init(&verno_major, &verno_minor);
if (err) {
opal_show_help("help-mtl-psm2.txt",