1
1
openmpi/ompi/mca/btl/openib/btl_openib_mca.c

701 строка
28 KiB
C
Исходник Обычный вид История

Bring over all the work from the /tmp/ib-hw-detect branch. In addition to my design and testing, it was conceptually approved by Gil, Gleb, Pasha, Brad, and Galen. Functionally [probably somewhat lightly] tested by Galen. We may still have to shake out some bugs during the next few months, but it seems to be working for all the cases that I can throw at it. Here's a summary of the changes from that branch: * Move MCA parameter registration to a new file (btl_openib_mca.c): * Properly check the retun status of registering MCA params * Check for valid values of MCA parameters * Make help strings better * Otherwise, the only default value of an MCA param that was changed was max_btls; it went from 4 to -1 (meaning: use all available) * Properly prototyped internal functions in _component.c * Made a bunch of functions static that didn't need to be public * Renamed to remove "mca_" prefix from static functions * Call new MCA param registration function * Call new INI file read/lookup/finalize functions * Updated a bunch of macros to be "BTL_" instead of "ORTE_" * Be a little more consistent with return values * Handle -1 for the max_btls MCA param * Fixed a free() that should have been an OBJ_RELEASE() * Some re-indenting * Added INI-file parsing * New flex file: btl_openib_ini.l * New default HCA params .ini file (probably to be expanded over time by other HCA vendors) * Added more show_help messages for parsing problems * Read in INI files and cache the values for later lookup * When component opens an HCA, lookup to see if any corresponding values were found in the INI files (ID'ed by the HCA vendor_id and vendor_part_id) * Added btl_openib_verbose MCA param that shows what the INI-file stuff does (e.g., shows which MTU your HCA ends up using) * Added btl_openib_hca_param_files as a colon-delimited list of INI files to check for values during startup (in order, left-to-right, just like the MCA base directory param). * MTU is currently the only value supported in this framework. * It is not a fatal error if we don't find params for the HCA in the INI file(s). Instead, just print a warning. New MCA param btl_openib_warn_no_hca_params_found can be used to disable printing the warning. * Add MTU to peer negotiation when making a connection * Exchange maximum MTU; select the lesser of the two This commit was SVN r11182.
2006-08-14 23:30:37 +04:00
/*
* Copyright (c) 2004-2005 The Trustees of Indiana University and Indiana
* University Research and Technology
* Corporation. All rights reserved.
* Copyright (c) 2004-2005 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) 2006-2007 Cisco Systems, Inc. All rights reserved.
* Copyright (c) 2006-2007 Mellanox Technologies. All rights reserved.
* Copyright (c) 2006-2007 Los Alamos National Security, LLC. All rights
* reserved.
* Copyright (c) 2006-2007 Voltaire All rights reserved.
Bring over all the work from the /tmp/ib-hw-detect branch. In addition to my design and testing, it was conceptually approved by Gil, Gleb, Pasha, Brad, and Galen. Functionally [probably somewhat lightly] tested by Galen. We may still have to shake out some bugs during the next few months, but it seems to be working for all the cases that I can throw at it. Here's a summary of the changes from that branch: * Move MCA parameter registration to a new file (btl_openib_mca.c): * Properly check the retun status of registering MCA params * Check for valid values of MCA parameters * Make help strings better * Otherwise, the only default value of an MCA param that was changed was max_btls; it went from 4 to -1 (meaning: use all available) * Properly prototyped internal functions in _component.c * Made a bunch of functions static that didn't need to be public * Renamed to remove "mca_" prefix from static functions * Call new MCA param registration function * Call new INI file read/lookup/finalize functions * Updated a bunch of macros to be "BTL_" instead of "ORTE_" * Be a little more consistent with return values * Handle -1 for the max_btls MCA param * Fixed a free() that should have been an OBJ_RELEASE() * Some re-indenting * Added INI-file parsing * New flex file: btl_openib_ini.l * New default HCA params .ini file (probably to be expanded over time by other HCA vendors) * Added more show_help messages for parsing problems * Read in INI files and cache the values for later lookup * When component opens an HCA, lookup to see if any corresponding values were found in the INI files (ID'ed by the HCA vendor_id and vendor_part_id) * Added btl_openib_verbose MCA param that shows what the INI-file stuff does (e.g., shows which MTU your HCA ends up using) * Added btl_openib_hca_param_files as a colon-delimited list of INI files to check for values during startup (in order, left-to-right, just like the MCA base directory param). * MTU is currently the only value supported in this framework. * It is not a fatal error if we don't find params for the HCA in the INI file(s). Instead, just print a warning. New MCA param btl_openib_warn_no_hca_params_found can be used to disable printing the warning. * Add MTU to peer negotiation when making a connection * Exchange maximum MTU; select the lesser of the two This commit was SVN r11182.
2006-08-14 23:30:37 +04:00
* $COPYRIGHT$
*
* Additional copyrights may follow
*
* $HEADER$
*/
#include "ompi_config.h"
#include <string.h>
#include "opal/mca/installdirs/installdirs.h"
Bring over all the work from the /tmp/ib-hw-detect branch. In addition to my design and testing, it was conceptually approved by Gil, Gleb, Pasha, Brad, and Galen. Functionally [probably somewhat lightly] tested by Galen. We may still have to shake out some bugs during the next few months, but it seems to be working for all the cases that I can throw at it. Here's a summary of the changes from that branch: * Move MCA parameter registration to a new file (btl_openib_mca.c): * Properly check the retun status of registering MCA params * Check for valid values of MCA parameters * Make help strings better * Otherwise, the only default value of an MCA param that was changed was max_btls; it went from 4 to -1 (meaning: use all available) * Properly prototyped internal functions in _component.c * Made a bunch of functions static that didn't need to be public * Renamed to remove "mca_" prefix from static functions * Call new MCA param registration function * Call new INI file read/lookup/finalize functions * Updated a bunch of macros to be "BTL_" instead of "ORTE_" * Be a little more consistent with return values * Handle -1 for the max_btls MCA param * Fixed a free() that should have been an OBJ_RELEASE() * Some re-indenting * Added INI-file parsing * New flex file: btl_openib_ini.l * New default HCA params .ini file (probably to be expanded over time by other HCA vendors) * Added more show_help messages for parsing problems * Read in INI files and cache the values for later lookup * When component opens an HCA, lookup to see if any corresponding values were found in the INI files (ID'ed by the HCA vendor_id and vendor_part_id) * Added btl_openib_verbose MCA param that shows what the INI-file stuff does (e.g., shows which MTU your HCA ends up using) * Added btl_openib_hca_param_files as a colon-delimited list of INI files to check for values during startup (in order, left-to-right, just like the MCA base directory param). * MTU is currently the only value supported in this framework. * It is not a fatal error if we don't find params for the HCA in the INI file(s). Instead, just print a warning. New MCA param btl_openib_warn_no_hca_params_found can be used to disable printing the warning. * Add MTU to peer negotiation when making a connection * Exchange maximum MTU; select the lesser of the two This commit was SVN r11182.
2006-08-14 23:30:37 +04:00
#include "opal/util/output.h"
This commit brings in two major things: 1. Galen's fine-grain control of queue pair resources in the openib BTL. 1. Pasha's new implementation of asychronous HCA event handling. Pasha's new implementation doesn't take much explanation, but the new "multifrag" stuff does. Note that "svn merge" was not used to bring this new code from the /tmp/ib_multifrag branch -- something Bad happened in the periodic trunk pulls on that branch making an actual merge back to the trunk effectively impossible (i.e., lots and lots of arbitrary conflicts and artifical changes). :-( == Fine-grain control of queue pair resources == Galen's fine-grain control of queue pair resources to the OpenIB BTL (thanks to Gleb for fixing broken code and providing additional functionality, Pasha for finding broken code, and Jeff for doing all the svn work and regression testing). Prior to this commit, the OpenIB BTL created two queue pairs: one for eager size fragments and one for max send size fragments. When the use of the shared receive queue (SRQ) was specified (via "-mca btl_openib_use_srq 1"), these QPs would use a shared receive queue for receive buffers instead of the default per-peer (PP) receive queues and buffers. One consequence of this design is that receive buffer utilization (the size of the data received as a percentage of the receive buffer used for the data) was quite poor for a number of applications. The new design allows multiple QPs to be specified at runtime. Each QP can be setup to use PP or SRQ receive buffers as well as giving fine-grained control over receive buffer size, number of receive buffers to post, when to replenish the receive queue (low water mark) and for SRQ QPs, the number of outstanding sends can also be specified. The following is an example of the syntax to describe QPs to the OpenIB BTL using the new MCA parameter btl_openib_receive_queues: {{{ -mca btl_openib_receive_queues \ "P,128,16,4;S,1024,256,128,32;S,4096,256,128,32;S,65536,256,128,32" }}} Each QP description is delimited by ";" (semicolon) with individual fields of the QP description delimited by "," (comma). The above example therefore describes 4 QPs. The first QP is: P,128,16,4 Meaning: per-peer receive buffer QPs are indicated by a starting field of "P"; the first QP (shown above) is therefore a per-peer based QP. The second field indicates the size of the receive buffer in bytes (128 bytes). The third field indicates the number of receive buffers to allocate to the QP (16). The fourth field indicates the low watermark for receive buffers at which time the BTL will repost receive buffers to the QP (4). The second QP is: S,1024,256,128,32 Shared receive queue based QPs are indicated by a starting field of "S"; the second QP (shown above) is therefore a shared receive queue based QP. The second, third and fourth fields are the same as in the per-peer based QP. The fifth field is the number of outstanding sends that are allowed at a given time on the QP (32). This provides a "good enough" mechanism of flow control for some regular communication patterns. QPs MUST be specified in ascending receive buffer size order. This requirement may be removed prior to 1.3 release. This commit was SVN r15474.
2007-07-18 05:15:59 +04:00
#include "opal/util/argv.h"
#include "opal/util/show_help.h"
Bring over all the work from the /tmp/ib-hw-detect branch. In addition to my design and testing, it was conceptually approved by Gil, Gleb, Pasha, Brad, and Galen. Functionally [probably somewhat lightly] tested by Galen. We may still have to shake out some bugs during the next few months, but it seems to be working for all the cases that I can throw at it. Here's a summary of the changes from that branch: * Move MCA parameter registration to a new file (btl_openib_mca.c): * Properly check the retun status of registering MCA params * Check for valid values of MCA parameters * Make help strings better * Otherwise, the only default value of an MCA param that was changed was max_btls; it went from 4 to -1 (meaning: use all available) * Properly prototyped internal functions in _component.c * Made a bunch of functions static that didn't need to be public * Renamed to remove "mca_" prefix from static functions * Call new MCA param registration function * Call new INI file read/lookup/finalize functions * Updated a bunch of macros to be "BTL_" instead of "ORTE_" * Be a little more consistent with return values * Handle -1 for the max_btls MCA param * Fixed a free() that should have been an OBJ_RELEASE() * Some re-indenting * Added INI-file parsing * New flex file: btl_openib_ini.l * New default HCA params .ini file (probably to be expanded over time by other HCA vendors) * Added more show_help messages for parsing problems * Read in INI files and cache the values for later lookup * When component opens an HCA, lookup to see if any corresponding values were found in the INI files (ID'ed by the HCA vendor_id and vendor_part_id) * Added btl_openib_verbose MCA param that shows what the INI-file stuff does (e.g., shows which MTU your HCA ends up using) * Added btl_openib_hca_param_files as a colon-delimited list of INI files to check for values during startup (in order, left-to-right, just like the MCA base directory param). * MTU is currently the only value supported in this framework. * It is not a fatal error if we don't find params for the HCA in the INI file(s). Instead, just print a warning. New MCA param btl_openib_warn_no_hca_params_found can be used to disable printing the warning. * Add MTU to peer negotiation when making a connection * Exchange maximum MTU; select the lesser of the two This commit was SVN r11182.
2006-08-14 23:30:37 +04:00
#include "opal/mca/base/mca_base_param.h"
#include "btl_openib.h"
#include "btl_openib_mca.h"
#include "connect/base.h"
Bring over all the work from the /tmp/ib-hw-detect branch. In addition to my design and testing, it was conceptually approved by Gil, Gleb, Pasha, Brad, and Galen. Functionally [probably somewhat lightly] tested by Galen. We may still have to shake out some bugs during the next few months, but it seems to be working for all the cases that I can throw at it. Here's a summary of the changes from that branch: * Move MCA parameter registration to a new file (btl_openib_mca.c): * Properly check the retun status of registering MCA params * Check for valid values of MCA parameters * Make help strings better * Otherwise, the only default value of an MCA param that was changed was max_btls; it went from 4 to -1 (meaning: use all available) * Properly prototyped internal functions in _component.c * Made a bunch of functions static that didn't need to be public * Renamed to remove "mca_" prefix from static functions * Call new MCA param registration function * Call new INI file read/lookup/finalize functions * Updated a bunch of macros to be "BTL_" instead of "ORTE_" * Be a little more consistent with return values * Handle -1 for the max_btls MCA param * Fixed a free() that should have been an OBJ_RELEASE() * Some re-indenting * Added INI-file parsing * New flex file: btl_openib_ini.l * New default HCA params .ini file (probably to be expanded over time by other HCA vendors) * Added more show_help messages for parsing problems * Read in INI files and cache the values for later lookup * When component opens an HCA, lookup to see if any corresponding values were found in the INI files (ID'ed by the HCA vendor_id and vendor_part_id) * Added btl_openib_verbose MCA param that shows what the INI-file stuff does (e.g., shows which MTU your HCA ends up using) * Added btl_openib_hca_param_files as a colon-delimited list of INI files to check for values during startup (in order, left-to-right, just like the MCA base directory param). * MTU is currently the only value supported in this framework. * It is not a fatal error if we don't find params for the HCA in the INI file(s). Instead, just print a warning. New MCA param btl_openib_warn_no_hca_params_found can be used to disable printing the warning. * Add MTU to peer negotiation when making a connection * Exchange maximum MTU; select the lesser of the two This commit was SVN r11182.
2006-08-14 23:30:37 +04:00
/*
* Local flags
*/
enum {
REGINT_NEG_ONE_OK = 0x01,
REGINT_GE_ZERO = 0x02,
REGINT_GE_ONE = 0x04,
REGINT_NONZERO = 0x08,
REGINT_MAX = 0x88
};
enum {
REGSTR_EMPTY_OK = 0x01,
REGSTR_MAX = 0x88
};
This commit brings in two major things: 1. Galen's fine-grain control of queue pair resources in the openib BTL. 1. Pasha's new implementation of asychronous HCA event handling. Pasha's new implementation doesn't take much explanation, but the new "multifrag" stuff does. Note that "svn merge" was not used to bring this new code from the /tmp/ib_multifrag branch -- something Bad happened in the periodic trunk pulls on that branch making an actual merge back to the trunk effectively impossible (i.e., lots and lots of arbitrary conflicts and artifical changes). :-( == Fine-grain control of queue pair resources == Galen's fine-grain control of queue pair resources to the OpenIB BTL (thanks to Gleb for fixing broken code and providing additional functionality, Pasha for finding broken code, and Jeff for doing all the svn work and regression testing). Prior to this commit, the OpenIB BTL created two queue pairs: one for eager size fragments and one for max send size fragments. When the use of the shared receive queue (SRQ) was specified (via "-mca btl_openib_use_srq 1"), these QPs would use a shared receive queue for receive buffers instead of the default per-peer (PP) receive queues and buffers. One consequence of this design is that receive buffer utilization (the size of the data received as a percentage of the receive buffer used for the data) was quite poor for a number of applications. The new design allows multiple QPs to be specified at runtime. Each QP can be setup to use PP or SRQ receive buffers as well as giving fine-grained control over receive buffer size, number of receive buffers to post, when to replenish the receive queue (low water mark) and for SRQ QPs, the number of outstanding sends can also be specified. The following is an example of the syntax to describe QPs to the OpenIB BTL using the new MCA parameter btl_openib_receive_queues: {{{ -mca btl_openib_receive_queues \ "P,128,16,4;S,1024,256,128,32;S,4096,256,128,32;S,65536,256,128,32" }}} Each QP description is delimited by ";" (semicolon) with individual fields of the QP description delimited by "," (comma). The above example therefore describes 4 QPs. The first QP is: P,128,16,4 Meaning: per-peer receive buffer QPs are indicated by a starting field of "P"; the first QP (shown above) is therefore a per-peer based QP. The second field indicates the size of the receive buffer in bytes (128 bytes). The third field indicates the number of receive buffers to allocate to the QP (16). The fourth field indicates the low watermark for receive buffers at which time the BTL will repost receive buffers to the QP (4). The second QP is: S,1024,256,128,32 Shared receive queue based QPs are indicated by a starting field of "S"; the second QP (shown above) is therefore a shared receive queue based QP. The second, third and fourth fields are the same as in the per-peer based QP. The fifth field is the number of outstanding sends that are allowed at a given time on the QP (32). This provides a "good enough" mechanism of flow control for some regular communication patterns. QPs MUST be specified in ascending receive buffer size order. This requirement may be removed prior to 1.3 release. This commit was SVN r15474.
2007-07-18 05:15:59 +04:00
static int mca_btl_openib_mca_setup_qps(void);
Bring over all the work from the /tmp/ib-hw-detect branch. In addition to my design and testing, it was conceptually approved by Gil, Gleb, Pasha, Brad, and Galen. Functionally [probably somewhat lightly] tested by Galen. We may still have to shake out some bugs during the next few months, but it seems to be working for all the cases that I can throw at it. Here's a summary of the changes from that branch: * Move MCA parameter registration to a new file (btl_openib_mca.c): * Properly check the retun status of registering MCA params * Check for valid values of MCA parameters * Make help strings better * Otherwise, the only default value of an MCA param that was changed was max_btls; it went from 4 to -1 (meaning: use all available) * Properly prototyped internal functions in _component.c * Made a bunch of functions static that didn't need to be public * Renamed to remove "mca_" prefix from static functions * Call new MCA param registration function * Call new INI file read/lookup/finalize functions * Updated a bunch of macros to be "BTL_" instead of "ORTE_" * Be a little more consistent with return values * Handle -1 for the max_btls MCA param * Fixed a free() that should have been an OBJ_RELEASE() * Some re-indenting * Added INI-file parsing * New flex file: btl_openib_ini.l * New default HCA params .ini file (probably to be expanded over time by other HCA vendors) * Added more show_help messages for parsing problems * Read in INI files and cache the values for later lookup * When component opens an HCA, lookup to see if any corresponding values were found in the INI files (ID'ed by the HCA vendor_id and vendor_part_id) * Added btl_openib_verbose MCA param that shows what the INI-file stuff does (e.g., shows which MTU your HCA ends up using) * Added btl_openib_hca_param_files as a colon-delimited list of INI files to check for values during startup (in order, left-to-right, just like the MCA base directory param). * MTU is currently the only value supported in this framework. * It is not a fatal error if we don't find params for the HCA in the INI file(s). Instead, just print a warning. New MCA param btl_openib_warn_no_hca_params_found can be used to disable printing the warning. * Add MTU to peer negotiation when making a connection * Exchange maximum MTU; select the lesser of the two This commit was SVN r11182.
2006-08-14 23:30:37 +04:00
/*
* utility routine for string parameter registration
*/
static inline int reg_string(const char* param_name, const char* param_desc,
const char* default_value, char **out_value,
int flags)
{
char *value;
mca_base_param_reg_string(&mca_btl_openib_component.super.btl_version,
param_name, param_desc, false, false,
default_value, &value);
if (0 != (flags & REGSTR_EMPTY_OK) && 0 == strlen(value)) {
return OMPI_ERR_BAD_PARAM;
}
*out_value = value;
return OMPI_SUCCESS;
}
/*
* utility routine for integer parameter registration
*/
static inline int reg_int(const char* param_name, const char* param_desc,
int default_value, int *out_value, int flags)
{
int value;
mca_base_param_reg_int(&mca_btl_openib_component.super.btl_version,
param_name, param_desc, false, false,
default_value, &value);
if (0 != (flags & REGINT_NEG_ONE_OK) && -1 == value) {
*out_value = value;
return OMPI_SUCCESS;
}
if ((0 != (flags & REGINT_GE_ZERO) && value < 0) ||
(0 != (flags & REGINT_GE_ONE) && value < 1) ||
(0 != (flags & REGINT_NONZERO) && 0 == value)) {
return OMPI_ERR_BAD_PARAM;
}
*out_value = value;
return OMPI_SUCCESS;
}
/*
* Register and check all MCA parameters
*/
int btl_openib_register_mca_params(void)
{
char *msg, *str;
int ival, ival2, ret, tmp;
Bring over all the work from the /tmp/ib-hw-detect branch. In addition to my design and testing, it was conceptually approved by Gil, Gleb, Pasha, Brad, and Galen. Functionally [probably somewhat lightly] tested by Galen. We may still have to shake out some bugs during the next few months, but it seems to be working for all the cases that I can throw at it. Here's a summary of the changes from that branch: * Move MCA parameter registration to a new file (btl_openib_mca.c): * Properly check the retun status of registering MCA params * Check for valid values of MCA parameters * Make help strings better * Otherwise, the only default value of an MCA param that was changed was max_btls; it went from 4 to -1 (meaning: use all available) * Properly prototyped internal functions in _component.c * Made a bunch of functions static that didn't need to be public * Renamed to remove "mca_" prefix from static functions * Call new MCA param registration function * Call new INI file read/lookup/finalize functions * Updated a bunch of macros to be "BTL_" instead of "ORTE_" * Be a little more consistent with return values * Handle -1 for the max_btls MCA param * Fixed a free() that should have been an OBJ_RELEASE() * Some re-indenting * Added INI-file parsing * New flex file: btl_openib_ini.l * New default HCA params .ini file (probably to be expanded over time by other HCA vendors) * Added more show_help messages for parsing problems * Read in INI files and cache the values for later lookup * When component opens an HCA, lookup to see if any corresponding values were found in the INI files (ID'ed by the HCA vendor_id and vendor_part_id) * Added btl_openib_verbose MCA param that shows what the INI-file stuff does (e.g., shows which MTU your HCA ends up using) * Added btl_openib_hca_param_files as a colon-delimited list of INI files to check for values during startup (in order, left-to-right, just like the MCA base directory param). * MTU is currently the only value supported in this framework. * It is not a fatal error if we don't find params for the HCA in the INI file(s). Instead, just print a warning. New MCA param btl_openib_warn_no_hca_params_found can be used to disable printing the warning. * Add MTU to peer negotiation when making a connection * Exchange maximum MTU; select the lesser of the two This commit was SVN r11182.
2006-08-14 23:30:37 +04:00
ret = OMPI_SUCCESS;
#define CHECK(expr) \
tmp = (expr); \
if (OMPI_SUCCESS != tmp) ret = tmp;
/* register IB component parameters */
CHECK(reg_int("verbose",
"Output some verbose OpenIB BTL information "
"(0 = no output, nonzero = output)", 0, &ival, 0));
mca_btl_openib_component.verbose = (0 != ival);
CHECK(reg_int("warn_no_hca_params_found",
"Warn when no HCA-specific parameters are found in the INI file specified by the btl_openib_hca_param_files MCA parameter (0 = do not warn; any other value = warn)",
1, &ival, 0));
mca_btl_openib_component.warn_no_hca_params_found = (0 != ival);
CHECK(reg_int("warn_default_gid_prefix",
"Warn when there is more than one active ports and at least one of them connected to the network with only default GID prefix configured (0 = do not warn; any other value = warn)",
1, &ival, 0));
mca_btl_openib_component.warn_default_gid_prefix = (0 != ival);
CHECK(reg_int("warn_nonexistent_if",
"Warn if non-existent HCAs and/or ports are specified in the btl_openib_if_[in|ex]clude MCA parameters (0 = do not warn; any other value = warn)",
1, &ival, 0));
mca_btl_openib_component.warn_nonexistent_if = (0 != ival);
#ifdef HAVE_IBV_FORK_INIT
ival2 = -1;
#else
ival2 = 0;
#endif
CHECK(reg_int("want_fork_support",
"Whether fork support is desired or not "
"(negative = try to enable fork support, but continue even if it is not available, 0 = do not enable fork support, positive = try to enable fork support and fail if it is not available)",
ival2, &ival, 0));
#ifdef HAVE_IBV_FORK_INIT
mca_btl_openib_component.want_fork_support = ival;
#else
if (0 != ival) {
opal_show_help("help-mpi-btl-openib.txt",
"ibv_fork requested but not supported", true,
orte_system_info.nodename);
return OMPI_ERROR;
}
#endif
asprintf(&str, "%s/mca-btl-openib-hca-params.ini",
opal_install_dirs.pkgdatadir);
Bring over all the work from the /tmp/ib-hw-detect branch. In addition to my design and testing, it was conceptually approved by Gil, Gleb, Pasha, Brad, and Galen. Functionally [probably somewhat lightly] tested by Galen. We may still have to shake out some bugs during the next few months, but it seems to be working for all the cases that I can throw at it. Here's a summary of the changes from that branch: * Move MCA parameter registration to a new file (btl_openib_mca.c): * Properly check the retun status of registering MCA params * Check for valid values of MCA parameters * Make help strings better * Otherwise, the only default value of an MCA param that was changed was max_btls; it went from 4 to -1 (meaning: use all available) * Properly prototyped internal functions in _component.c * Made a bunch of functions static that didn't need to be public * Renamed to remove "mca_" prefix from static functions * Call new MCA param registration function * Call new INI file read/lookup/finalize functions * Updated a bunch of macros to be "BTL_" instead of "ORTE_" * Be a little more consistent with return values * Handle -1 for the max_btls MCA param * Fixed a free() that should have been an OBJ_RELEASE() * Some re-indenting * Added INI-file parsing * New flex file: btl_openib_ini.l * New default HCA params .ini file (probably to be expanded over time by other HCA vendors) * Added more show_help messages for parsing problems * Read in INI files and cache the values for later lookup * When component opens an HCA, lookup to see if any corresponding values were found in the INI files (ID'ed by the HCA vendor_id and vendor_part_id) * Added btl_openib_verbose MCA param that shows what the INI-file stuff does (e.g., shows which MTU your HCA ends up using) * Added btl_openib_hca_param_files as a colon-delimited list of INI files to check for values during startup (in order, left-to-right, just like the MCA base directory param). * MTU is currently the only value supported in this framework. * It is not a fatal error if we don't find params for the HCA in the INI file(s). Instead, just print a warning. New MCA param btl_openib_warn_no_hca_params_found can be used to disable printing the warning. * Add MTU to peer negotiation when making a connection * Exchange maximum MTU; select the lesser of the two This commit was SVN r11182.
2006-08-14 23:30:37 +04:00
if (NULL == str) {
return OMPI_ERR_OUT_OF_RESOURCE;
}
#ifdef HAVE_IBV_FORK_INIT
ival2 = -1;
#else
ival2 = 0;
#endif
CHECK(reg_int("want_fork_support",
"Whether fork support is desired or not "
"(negative = try to enable fork support, but continue even if it is not available, 0 = do not enable fork support, positive = try to enable fork support and fail if it is not available)",
ival2, &ival, 0));
#ifdef HAVE_IBV_FORK_INIT
mca_btl_openib_component.want_fork_support = ival;
#else
if (0 != ival) {
opal_show_help("help-mpi-btl-openib.txt",
"ibv_fork requested but not supported", true,
orte_system_info.nodename);
return OMPI_ERROR;
}
#endif
Bring over all the work from the /tmp/ib-hw-detect branch. In addition to my design and testing, it was conceptually approved by Gil, Gleb, Pasha, Brad, and Galen. Functionally [probably somewhat lightly] tested by Galen. We may still have to shake out some bugs during the next few months, but it seems to be working for all the cases that I can throw at it. Here's a summary of the changes from that branch: * Move MCA parameter registration to a new file (btl_openib_mca.c): * Properly check the retun status of registering MCA params * Check for valid values of MCA parameters * Make help strings better * Otherwise, the only default value of an MCA param that was changed was max_btls; it went from 4 to -1 (meaning: use all available) * Properly prototyped internal functions in _component.c * Made a bunch of functions static that didn't need to be public * Renamed to remove "mca_" prefix from static functions * Call new MCA param registration function * Call new INI file read/lookup/finalize functions * Updated a bunch of macros to be "BTL_" instead of "ORTE_" * Be a little more consistent with return values * Handle -1 for the max_btls MCA param * Fixed a free() that should have been an OBJ_RELEASE() * Some re-indenting * Added INI-file parsing * New flex file: btl_openib_ini.l * New default HCA params .ini file (probably to be expanded over time by other HCA vendors) * Added more show_help messages for parsing problems * Read in INI files and cache the values for later lookup * When component opens an HCA, lookup to see if any corresponding values were found in the INI files (ID'ed by the HCA vendor_id and vendor_part_id) * Added btl_openib_verbose MCA param that shows what the INI-file stuff does (e.g., shows which MTU your HCA ends up using) * Added btl_openib_hca_param_files as a colon-delimited list of INI files to check for values during startup (in order, left-to-right, just like the MCA base directory param). * MTU is currently the only value supported in this framework. * It is not a fatal error if we don't find params for the HCA in the INI file(s). Instead, just print a warning. New MCA param btl_openib_warn_no_hca_params_found can be used to disable printing the warning. * Add MTU to peer negotiation when making a connection * Exchange maximum MTU; select the lesser of the two This commit was SVN r11182.
2006-08-14 23:30:37 +04:00
CHECK(reg_string("hca_param_files",
"Colon-delimited list of INI-style files that contain HCA vendor/part-specific parameters",
str, &mca_btl_openib_component.hca_params_file_names, 0));
free(str);
CHECK(reg_int("max_btls",
"Maximum number of HCA ports to use "
"(-1 = use all available, otherwise must be >= 1)",
-1, &mca_btl_openib_component.ib_max_btls,
REGINT_NEG_ONE_OK | REGINT_GE_ONE));
CHECK(reg_int("free_list_num",
"Intial size of free lists (must be >= 1)",
8, &mca_btl_openib_component.ib_free_list_num,
REGINT_GE_ONE));
CHECK(reg_int("free_list_max",
"Maximum size of free lists "
"(-1 = infinite, otherwise must be >= 0)",
-1, &mca_btl_openib_component.ib_free_list_max,
REGINT_NEG_ONE_OK | REGINT_GE_ONE));
CHECK(reg_int("free_list_inc",
"Increment size of free lists (must be >= 1)",
32, &mca_btl_openib_component.ib_free_list_inc,
REGINT_GE_ONE));
CHECK(reg_string("mpool",
"Name of the memory pool to be used (it is unlikely that you will ever want to change this",
"rdma", &mca_btl_openib_component.ib_mpool_name,
Bring over all the work from the /tmp/ib-hw-detect branch. In addition to my design and testing, it was conceptually approved by Gil, Gleb, Pasha, Brad, and Galen. Functionally [probably somewhat lightly] tested by Galen. We may still have to shake out some bugs during the next few months, but it seems to be working for all the cases that I can throw at it. Here's a summary of the changes from that branch: * Move MCA parameter registration to a new file (btl_openib_mca.c): * Properly check the retun status of registering MCA params * Check for valid values of MCA parameters * Make help strings better * Otherwise, the only default value of an MCA param that was changed was max_btls; it went from 4 to -1 (meaning: use all available) * Properly prototyped internal functions in _component.c * Made a bunch of functions static that didn't need to be public * Renamed to remove "mca_" prefix from static functions * Call new MCA param registration function * Call new INI file read/lookup/finalize functions * Updated a bunch of macros to be "BTL_" instead of "ORTE_" * Be a little more consistent with return values * Handle -1 for the max_btls MCA param * Fixed a free() that should have been an OBJ_RELEASE() * Some re-indenting * Added INI-file parsing * New flex file: btl_openib_ini.l * New default HCA params .ini file (probably to be expanded over time by other HCA vendors) * Added more show_help messages for parsing problems * Read in INI files and cache the values for later lookup * When component opens an HCA, lookup to see if any corresponding values were found in the INI files (ID'ed by the HCA vendor_id and vendor_part_id) * Added btl_openib_verbose MCA param that shows what the INI-file stuff does (e.g., shows which MTU your HCA ends up using) * Added btl_openib_hca_param_files as a colon-delimited list of INI files to check for values during startup (in order, left-to-right, just like the MCA base directory param). * MTU is currently the only value supported in this framework. * It is not a fatal error if we don't find params for the HCA in the INI file(s). Instead, just print a warning. New MCA param btl_openib_warn_no_hca_params_found can be used to disable printing the warning. * Add MTU to peer negotiation when making a connection * Exchange maximum MTU; select the lesser of the two This commit was SVN r11182.
2006-08-14 23:30:37 +04:00
0));
CHECK(reg_int("reg_mru_len",
"Length of the registration cache most recently used list "
"(must be >= 1)",
16, (int*) &mca_btl_openib_component.reg_mru_len,
REGINT_GE_ONE));
CHECK(reg_int("ib_cq_size", "Size of the IB completion "
"queue (will automatically be set to a minimum of "
"(2 * number_of_peers * btl_openib_rd_num))",
1000, &ival, REGINT_GE_ONE));
mca_btl_openib_component.ib_cq_size[BTL_OPENIB_LP_CQ] =
mca_btl_openib_component.ib_cq_size[BTL_OPENIB_HP_CQ] = (uint32_t) ival;
Bring over all the work from the /tmp/ib-hw-detect branch. In addition to my design and testing, it was conceptually approved by Gil, Gleb, Pasha, Brad, and Galen. Functionally [probably somewhat lightly] tested by Galen. We may still have to shake out some bugs during the next few months, but it seems to be working for all the cases that I can throw at it. Here's a summary of the changes from that branch: * Move MCA parameter registration to a new file (btl_openib_mca.c): * Properly check the retun status of registering MCA params * Check for valid values of MCA parameters * Make help strings better * Otherwise, the only default value of an MCA param that was changed was max_btls; it went from 4 to -1 (meaning: use all available) * Properly prototyped internal functions in _component.c * Made a bunch of functions static that didn't need to be public * Renamed to remove "mca_" prefix from static functions * Call new MCA param registration function * Call new INI file read/lookup/finalize functions * Updated a bunch of macros to be "BTL_" instead of "ORTE_" * Be a little more consistent with return values * Handle -1 for the max_btls MCA param * Fixed a free() that should have been an OBJ_RELEASE() * Some re-indenting * Added INI-file parsing * New flex file: btl_openib_ini.l * New default HCA params .ini file (probably to be expanded over time by other HCA vendors) * Added more show_help messages for parsing problems * Read in INI files and cache the values for later lookup * When component opens an HCA, lookup to see if any corresponding values were found in the INI files (ID'ed by the HCA vendor_id and vendor_part_id) * Added btl_openib_verbose MCA param that shows what the INI-file stuff does (e.g., shows which MTU your HCA ends up using) * Added btl_openib_hca_param_files as a colon-delimited list of INI files to check for values during startup (in order, left-to-right, just like the MCA base directory param). * MTU is currently the only value supported in this framework. * It is not a fatal error if we don't find params for the HCA in the INI file(s). Instead, just print a warning. New MCA param btl_openib_warn_no_hca_params_found can be used to disable printing the warning. * Add MTU to peer negotiation when making a connection * Exchange maximum MTU; select the lesser of the two This commit was SVN r11182.
2006-08-14 23:30:37 +04:00
CHECK(reg_int("ib_sg_list_size", "Size of IB segment list "
"(must be >= 1)",
4, &ival, REGINT_GE_ONE));
mca_btl_openib_component.ib_sg_list_size = (uint32_t) ival;
CHECK(reg_int("ib_pkey_ix", "InfiniBand pkey index "
"(must be >= 0)",
0, &ival, REGINT_GE_ZERO));
mca_btl_openib_component.ib_pkey_ix = (uint32_t) ival;
CHECK(reg_int("ib_pkey_val", "InfiniBand pkey value"
"(must be > 0 and < 0xffff)",
0, &ival, REGINT_GE_ZERO));
if (ival > 0xffff) {
This commit brings in two major things: 1. Galen's fine-grain control of queue pair resources in the openib BTL. 1. Pasha's new implementation of asychronous HCA event handling. Pasha's new implementation doesn't take much explanation, but the new "multifrag" stuff does. Note that "svn merge" was not used to bring this new code from the /tmp/ib_multifrag branch -- something Bad happened in the periodic trunk pulls on that branch making an actual merge back to the trunk effectively impossible (i.e., lots and lots of arbitrary conflicts and artifical changes). :-( == Fine-grain control of queue pair resources == Galen's fine-grain control of queue pair resources to the OpenIB BTL (thanks to Gleb for fixing broken code and providing additional functionality, Pasha for finding broken code, and Jeff for doing all the svn work and regression testing). Prior to this commit, the OpenIB BTL created two queue pairs: one for eager size fragments and one for max send size fragments. When the use of the shared receive queue (SRQ) was specified (via "-mca btl_openib_use_srq 1"), these QPs would use a shared receive queue for receive buffers instead of the default per-peer (PP) receive queues and buffers. One consequence of this design is that receive buffer utilization (the size of the data received as a percentage of the receive buffer used for the data) was quite poor for a number of applications. The new design allows multiple QPs to be specified at runtime. Each QP can be setup to use PP or SRQ receive buffers as well as giving fine-grained control over receive buffer size, number of receive buffers to post, when to replenish the receive queue (low water mark) and for SRQ QPs, the number of outstanding sends can also be specified. The following is an example of the syntax to describe QPs to the OpenIB BTL using the new MCA parameter btl_openib_receive_queues: {{{ -mca btl_openib_receive_queues \ "P,128,16,4;S,1024,256,128,32;S,4096,256,128,32;S,65536,256,128,32" }}} Each QP description is delimited by ";" (semicolon) with individual fields of the QP description delimited by "," (comma). The above example therefore describes 4 QPs. The first QP is: P,128,16,4 Meaning: per-peer receive buffer QPs are indicated by a starting field of "P"; the first QP (shown above) is therefore a per-peer based QP. The second field indicates the size of the receive buffer in bytes (128 bytes). The third field indicates the number of receive buffers to allocate to the QP (16). The fourth field indicates the low watermark for receive buffers at which time the BTL will repost receive buffers to the QP (4). The second QP is: S,1024,256,128,32 Shared receive queue based QPs are indicated by a starting field of "S"; the second QP (shown above) is therefore a shared receive queue based QP. The second, third and fourth fields are the same as in the per-peer based QP. The fifth field is the number of outstanding sends that are allowed at a given time on the QP (32). This provides a "good enough" mechanism of flow control for some regular communication patterns. QPs MUST be specified in ascending receive buffer size order. This requirement may be removed prior to 1.3 release. This commit was SVN r15474.
2007-07-18 05:15:59 +04:00
opal_show_help("help-mpi-btl-openib.txt", "invalid mca param value",
true, "invalid value for btl_openib_ib_pkey_val",
"btl_openib_ib_pkey_val ignored");
} else {
mca_btl_openib_component.ib_pkey_val = (uint32_t) ival;
}
Bring over all the work from the /tmp/ib-hw-detect branch. In addition to my design and testing, it was conceptually approved by Gil, Gleb, Pasha, Brad, and Galen. Functionally [probably somewhat lightly] tested by Galen. We may still have to shake out some bugs during the next few months, but it seems to be working for all the cases that I can throw at it. Here's a summary of the changes from that branch: * Move MCA parameter registration to a new file (btl_openib_mca.c): * Properly check the retun status of registering MCA params * Check for valid values of MCA parameters * Make help strings better * Otherwise, the only default value of an MCA param that was changed was max_btls; it went from 4 to -1 (meaning: use all available) * Properly prototyped internal functions in _component.c * Made a bunch of functions static that didn't need to be public * Renamed to remove "mca_" prefix from static functions * Call new MCA param registration function * Call new INI file read/lookup/finalize functions * Updated a bunch of macros to be "BTL_" instead of "ORTE_" * Be a little more consistent with return values * Handle -1 for the max_btls MCA param * Fixed a free() that should have been an OBJ_RELEASE() * Some re-indenting * Added INI-file parsing * New flex file: btl_openib_ini.l * New default HCA params .ini file (probably to be expanded over time by other HCA vendors) * Added more show_help messages for parsing problems * Read in INI files and cache the values for later lookup * When component opens an HCA, lookup to see if any corresponding values were found in the INI files (ID'ed by the HCA vendor_id and vendor_part_id) * Added btl_openib_verbose MCA param that shows what the INI-file stuff does (e.g., shows which MTU your HCA ends up using) * Added btl_openib_hca_param_files as a colon-delimited list of INI files to check for values during startup (in order, left-to-right, just like the MCA base directory param). * MTU is currently the only value supported in this framework. * It is not a fatal error if we don't find params for the HCA in the INI file(s). Instead, just print a warning. New MCA param btl_openib_warn_no_hca_params_found can be used to disable printing the warning. * Add MTU to peer negotiation when making a connection * Exchange maximum MTU; select the lesser of the two This commit was SVN r11182.
2006-08-14 23:30:37 +04:00
CHECK(reg_int("ib_psn", "InfiniBand packet sequence starting number "
"(must be >= 0)",
0, &ival, REGINT_GE_ZERO));
mca_btl_openib_component.ib_psn = (uint32_t) ival;
CHECK(reg_int("ib_qp_ous_rd_atom", "InfiniBand outstanding atomic reads "
"(must be >= 0)",
4, &ival, REGINT_GE_ZERO));
mca_btl_openib_component.ib_qp_ous_rd_atom = (uint32_t) ival;
asprintf(&msg, "IB MTU, in bytes (if not specified in INI files). Valid values are: %d=256 bytes, %d=512 bytes, %d=1024 bytes, %d=2048 bytes, %d=4096 bytes",
IBV_MTU_256,
IBV_MTU_512,
IBV_MTU_1024,
IBV_MTU_2048,
IBV_MTU_4096);
if (NULL == msg) {
/* Don't try to recover from this */
return OMPI_ERR_OUT_OF_RESOURCE;
}
CHECK(reg_int("ib_mtu", msg, IBV_MTU_1024, &ival, 0));
free(msg);
if (ival < IBV_MTU_1024 || ival > IBV_MTU_4096) {
This commit brings in two major things: 1. Galen's fine-grain control of queue pair resources in the openib BTL. 1. Pasha's new implementation of asychronous HCA event handling. Pasha's new implementation doesn't take much explanation, but the new "multifrag" stuff does. Note that "svn merge" was not used to bring this new code from the /tmp/ib_multifrag branch -- something Bad happened in the periodic trunk pulls on that branch making an actual merge back to the trunk effectively impossible (i.e., lots and lots of arbitrary conflicts and artifical changes). :-( == Fine-grain control of queue pair resources == Galen's fine-grain control of queue pair resources to the OpenIB BTL (thanks to Gleb for fixing broken code and providing additional functionality, Pasha for finding broken code, and Jeff for doing all the svn work and regression testing). Prior to this commit, the OpenIB BTL created two queue pairs: one for eager size fragments and one for max send size fragments. When the use of the shared receive queue (SRQ) was specified (via "-mca btl_openib_use_srq 1"), these QPs would use a shared receive queue for receive buffers instead of the default per-peer (PP) receive queues and buffers. One consequence of this design is that receive buffer utilization (the size of the data received as a percentage of the receive buffer used for the data) was quite poor for a number of applications. The new design allows multiple QPs to be specified at runtime. Each QP can be setup to use PP or SRQ receive buffers as well as giving fine-grained control over receive buffer size, number of receive buffers to post, when to replenish the receive queue (low water mark) and for SRQ QPs, the number of outstanding sends can also be specified. The following is an example of the syntax to describe QPs to the OpenIB BTL using the new MCA parameter btl_openib_receive_queues: {{{ -mca btl_openib_receive_queues \ "P,128,16,4;S,1024,256,128,32;S,4096,256,128,32;S,65536,256,128,32" }}} Each QP description is delimited by ";" (semicolon) with individual fields of the QP description delimited by "," (comma). The above example therefore describes 4 QPs. The first QP is: P,128,16,4 Meaning: per-peer receive buffer QPs are indicated by a starting field of "P"; the first QP (shown above) is therefore a per-peer based QP. The second field indicates the size of the receive buffer in bytes (128 bytes). The third field indicates the number of receive buffers to allocate to the QP (16). The fourth field indicates the low watermark for receive buffers at which time the BTL will repost receive buffers to the QP (4). The second QP is: S,1024,256,128,32 Shared receive queue based QPs are indicated by a starting field of "S"; the second QP (shown above) is therefore a shared receive queue based QP. The second, third and fourth fields are the same as in the per-peer based QP. The fifth field is the number of outstanding sends that are allowed at a given time on the QP (32). This provides a "good enough" mechanism of flow control for some regular communication patterns. QPs MUST be specified in ascending receive buffer size order. This requirement may be removed prior to 1.3 release. This commit was SVN r15474.
2007-07-18 05:15:59 +04:00
opal_show_help("help-mpi-btl-openib.txt", "invalid mca param value",
true, "invalid value for btl_openib_ib_mtu",
"btl_openib_ib_mtu reset to 1024");
Bring over all the work from the /tmp/ib-hw-detect branch. In addition to my design and testing, it was conceptually approved by Gil, Gleb, Pasha, Brad, and Galen. Functionally [probably somewhat lightly] tested by Galen. We may still have to shake out some bugs during the next few months, but it seems to be working for all the cases that I can throw at it. Here's a summary of the changes from that branch: * Move MCA parameter registration to a new file (btl_openib_mca.c): * Properly check the retun status of registering MCA params * Check for valid values of MCA parameters * Make help strings better * Otherwise, the only default value of an MCA param that was changed was max_btls; it went from 4 to -1 (meaning: use all available) * Properly prototyped internal functions in _component.c * Made a bunch of functions static that didn't need to be public * Renamed to remove "mca_" prefix from static functions * Call new MCA param registration function * Call new INI file read/lookup/finalize functions * Updated a bunch of macros to be "BTL_" instead of "ORTE_" * Be a little more consistent with return values * Handle -1 for the max_btls MCA param * Fixed a free() that should have been an OBJ_RELEASE() * Some re-indenting * Added INI-file parsing * New flex file: btl_openib_ini.l * New default HCA params .ini file (probably to be expanded over time by other HCA vendors) * Added more show_help messages for parsing problems * Read in INI files and cache the values for later lookup * When component opens an HCA, lookup to see if any corresponding values were found in the INI files (ID'ed by the HCA vendor_id and vendor_part_id) * Added btl_openib_verbose MCA param that shows what the INI-file stuff does (e.g., shows which MTU your HCA ends up using) * Added btl_openib_hca_param_files as a colon-delimited list of INI files to check for values during startup (in order, left-to-right, just like the MCA base directory param). * MTU is currently the only value supported in this framework. * It is not a fatal error if we don't find params for the HCA in the INI file(s). Instead, just print a warning. New MCA param btl_openib_warn_no_hca_params_found can be used to disable printing the warning. * Add MTU to peer negotiation when making a connection * Exchange maximum MTU; select the lesser of the two This commit was SVN r11182.
2006-08-14 23:30:37 +04:00
mca_btl_openib_component.ib_mtu = IBV_MTU_1024;
} else {
mca_btl_openib_component.ib_mtu = (uint32_t) ival;
}
CHECK(reg_int("ib_min_rnr_timer", "InfiniBand minimum "
"\"receiver not ready\" timer, in seconds "
This commit brings in two major things: 1. Galen's fine-grain control of queue pair resources in the openib BTL. 1. Pasha's new implementation of asychronous HCA event handling. Pasha's new implementation doesn't take much explanation, but the new "multifrag" stuff does. Note that "svn merge" was not used to bring this new code from the /tmp/ib_multifrag branch -- something Bad happened in the periodic trunk pulls on that branch making an actual merge back to the trunk effectively impossible (i.e., lots and lots of arbitrary conflicts and artifical changes). :-( == Fine-grain control of queue pair resources == Galen's fine-grain control of queue pair resources to the OpenIB BTL (thanks to Gleb for fixing broken code and providing additional functionality, Pasha for finding broken code, and Jeff for doing all the svn work and regression testing). Prior to this commit, the OpenIB BTL created two queue pairs: one for eager size fragments and one for max send size fragments. When the use of the shared receive queue (SRQ) was specified (via "-mca btl_openib_use_srq 1"), these QPs would use a shared receive queue for receive buffers instead of the default per-peer (PP) receive queues and buffers. One consequence of this design is that receive buffer utilization (the size of the data received as a percentage of the receive buffer used for the data) was quite poor for a number of applications. The new design allows multiple QPs to be specified at runtime. Each QP can be setup to use PP or SRQ receive buffers as well as giving fine-grained control over receive buffer size, number of receive buffers to post, when to replenish the receive queue (low water mark) and for SRQ QPs, the number of outstanding sends can also be specified. The following is an example of the syntax to describe QPs to the OpenIB BTL using the new MCA parameter btl_openib_receive_queues: {{{ -mca btl_openib_receive_queues \ "P,128,16,4;S,1024,256,128,32;S,4096,256,128,32;S,65536,256,128,32" }}} Each QP description is delimited by ";" (semicolon) with individual fields of the QP description delimited by "," (comma). The above example therefore describes 4 QPs. The first QP is: P,128,16,4 Meaning: per-peer receive buffer QPs are indicated by a starting field of "P"; the first QP (shown above) is therefore a per-peer based QP. The second field indicates the size of the receive buffer in bytes (128 bytes). The third field indicates the number of receive buffers to allocate to the QP (16). The fourth field indicates the low watermark for receive buffers at which time the BTL will repost receive buffers to the QP (4). The second QP is: S,1024,256,128,32 Shared receive queue based QPs are indicated by a starting field of "S"; the second QP (shown above) is therefore a shared receive queue based QP. The second, third and fourth fields are the same as in the per-peer based QP. The fifth field is the number of outstanding sends that are allowed at a given time on the QP (32). This provides a "good enough" mechanism of flow control for some regular communication patterns. QPs MUST be specified in ascending receive buffer size order. This requirement may be removed prior to 1.3 release. This commit was SVN r15474.
2007-07-18 05:15:59 +04:00
"(must be >= 0 and <= 31)",
5, &ival, 0));
if (ival > 31) {
opal_show_help("help-mpi-btl-openib.txt", "invalid mca param value",
true, "btl_openib_ib_min_rnr_timer > 31",
"btl_openib_ib_min_rnr_timer reset to 31");
ival = 31;
} else if (ival < 0){
opal_show_help("help-mpi-btl-openib.txt", "invalid mca param value",
true, "btl_openib_ib_min_rnr_timer < 0",
"btl_openib_ib_min_rnr_timer reset to 0");
ival = 0;
}
Bring over all the work from the /tmp/ib-hw-detect branch. In addition to my design and testing, it was conceptually approved by Gil, Gleb, Pasha, Brad, and Galen. Functionally [probably somewhat lightly] tested by Galen. We may still have to shake out some bugs during the next few months, but it seems to be working for all the cases that I can throw at it. Here's a summary of the changes from that branch: * Move MCA parameter registration to a new file (btl_openib_mca.c): * Properly check the retun status of registering MCA params * Check for valid values of MCA parameters * Make help strings better * Otherwise, the only default value of an MCA param that was changed was max_btls; it went from 4 to -1 (meaning: use all available) * Properly prototyped internal functions in _component.c * Made a bunch of functions static that didn't need to be public * Renamed to remove "mca_" prefix from static functions * Call new MCA param registration function * Call new INI file read/lookup/finalize functions * Updated a bunch of macros to be "BTL_" instead of "ORTE_" * Be a little more consistent with return values * Handle -1 for the max_btls MCA param * Fixed a free() that should have been an OBJ_RELEASE() * Some re-indenting * Added INI-file parsing * New flex file: btl_openib_ini.l * New default HCA params .ini file (probably to be expanded over time by other HCA vendors) * Added more show_help messages for parsing problems * Read in INI files and cache the values for later lookup * When component opens an HCA, lookup to see if any corresponding values were found in the INI files (ID'ed by the HCA vendor_id and vendor_part_id) * Added btl_openib_verbose MCA param that shows what the INI-file stuff does (e.g., shows which MTU your HCA ends up using) * Added btl_openib_hca_param_files as a colon-delimited list of INI files to check for values during startup (in order, left-to-right, just like the MCA base directory param). * MTU is currently the only value supported in this framework. * It is not a fatal error if we don't find params for the HCA in the INI file(s). Instead, just print a warning. New MCA param btl_openib_warn_no_hca_params_found can be used to disable printing the warning. * Add MTU to peer negotiation when making a connection * Exchange maximum MTU; select the lesser of the two This commit was SVN r11182.
2006-08-14 23:30:37 +04:00
mca_btl_openib_component.ib_min_rnr_timer = (uint32_t) ival;
CHECK(reg_int("ib_timeout", "InfiniBand transmit timeout, plugged into formula: 4.096 microseconds * (2^btl_openib_ib_timeout)"
This commit brings in two major things: 1. Galen's fine-grain control of queue pair resources in the openib BTL. 1. Pasha's new implementation of asychronous HCA event handling. Pasha's new implementation doesn't take much explanation, but the new "multifrag" stuff does. Note that "svn merge" was not used to bring this new code from the /tmp/ib_multifrag branch -- something Bad happened in the periodic trunk pulls on that branch making an actual merge back to the trunk effectively impossible (i.e., lots and lots of arbitrary conflicts and artifical changes). :-( == Fine-grain control of queue pair resources == Galen's fine-grain control of queue pair resources to the OpenIB BTL (thanks to Gleb for fixing broken code and providing additional functionality, Pasha for finding broken code, and Jeff for doing all the svn work and regression testing). Prior to this commit, the OpenIB BTL created two queue pairs: one for eager size fragments and one for max send size fragments. When the use of the shared receive queue (SRQ) was specified (via "-mca btl_openib_use_srq 1"), these QPs would use a shared receive queue for receive buffers instead of the default per-peer (PP) receive queues and buffers. One consequence of this design is that receive buffer utilization (the size of the data received as a percentage of the receive buffer used for the data) was quite poor for a number of applications. The new design allows multiple QPs to be specified at runtime. Each QP can be setup to use PP or SRQ receive buffers as well as giving fine-grained control over receive buffer size, number of receive buffers to post, when to replenish the receive queue (low water mark) and for SRQ QPs, the number of outstanding sends can also be specified. The following is an example of the syntax to describe QPs to the OpenIB BTL using the new MCA parameter btl_openib_receive_queues: {{{ -mca btl_openib_receive_queues \ "P,128,16,4;S,1024,256,128,32;S,4096,256,128,32;S,65536,256,128,32" }}} Each QP description is delimited by ";" (semicolon) with individual fields of the QP description delimited by "," (comma). The above example therefore describes 4 QPs. The first QP is: P,128,16,4 Meaning: per-peer receive buffer QPs are indicated by a starting field of "P"; the first QP (shown above) is therefore a per-peer based QP. The second field indicates the size of the receive buffer in bytes (128 bytes). The third field indicates the number of receive buffers to allocate to the QP (16). The fourth field indicates the low watermark for receive buffers at which time the BTL will repost receive buffers to the QP (4). The second QP is: S,1024,256,128,32 Shared receive queue based QPs are indicated by a starting field of "S"; the second QP (shown above) is therefore a shared receive queue based QP. The second, third and fourth fields are the same as in the per-peer based QP. The fifth field is the number of outstanding sends that are allowed at a given time on the QP (32). This provides a "good enough" mechanism of flow control for some regular communication patterns. QPs MUST be specified in ascending receive buffer size order. This requirement may be removed prior to 1.3 release. This commit was SVN r15474.
2007-07-18 05:15:59 +04:00
"(must be >= 0 and <= 31)",
10, &ival, 0));
if (ival > 31) {
opal_show_help("help-mpi-btl-openib.txt", "invalid mca param value",
true, "btl_openib_ib_timeout > 31",
"btl_openib_ib_timeout reset to 31");
ival = 31;
} else if (ival < 0) {
opal_show_help("help-mpi-btl-openib.txt", "invalid mca param value",
true, "btl_openib_ib_timeout < 0",
"btl_openib_ib_timeout reset to 0");
ival = 0;
}
Bring over all the work from the /tmp/ib-hw-detect branch. In addition to my design and testing, it was conceptually approved by Gil, Gleb, Pasha, Brad, and Galen. Functionally [probably somewhat lightly] tested by Galen. We may still have to shake out some bugs during the next few months, but it seems to be working for all the cases that I can throw at it. Here's a summary of the changes from that branch: * Move MCA parameter registration to a new file (btl_openib_mca.c): * Properly check the retun status of registering MCA params * Check for valid values of MCA parameters * Make help strings better * Otherwise, the only default value of an MCA param that was changed was max_btls; it went from 4 to -1 (meaning: use all available) * Properly prototyped internal functions in _component.c * Made a bunch of functions static that didn't need to be public * Renamed to remove "mca_" prefix from static functions * Call new MCA param registration function * Call new INI file read/lookup/finalize functions * Updated a bunch of macros to be "BTL_" instead of "ORTE_" * Be a little more consistent with return values * Handle -1 for the max_btls MCA param * Fixed a free() that should have been an OBJ_RELEASE() * Some re-indenting * Added INI-file parsing * New flex file: btl_openib_ini.l * New default HCA params .ini file (probably to be expanded over time by other HCA vendors) * Added more show_help messages for parsing problems * Read in INI files and cache the values for later lookup * When component opens an HCA, lookup to see if any corresponding values were found in the INI files (ID'ed by the HCA vendor_id and vendor_part_id) * Added btl_openib_verbose MCA param that shows what the INI-file stuff does (e.g., shows which MTU your HCA ends up using) * Added btl_openib_hca_param_files as a colon-delimited list of INI files to check for values during startup (in order, left-to-right, just like the MCA base directory param). * MTU is currently the only value supported in this framework. * It is not a fatal error if we don't find params for the HCA in the INI file(s). Instead, just print a warning. New MCA param btl_openib_warn_no_hca_params_found can be used to disable printing the warning. * Add MTU to peer negotiation when making a connection * Exchange maximum MTU; select the lesser of the two This commit was SVN r11182.
2006-08-14 23:30:37 +04:00
mca_btl_openib_component.ib_timeout = (uint32_t) ival;
CHECK(reg_int("ib_retry_count", "InfiniBand transmit retry count "
"(must be >= 0 and <= 7)",
This commit brings in two major things: 1. Galen's fine-grain control of queue pair resources in the openib BTL. 1. Pasha's new implementation of asychronous HCA event handling. Pasha's new implementation doesn't take much explanation, but the new "multifrag" stuff does. Note that "svn merge" was not used to bring this new code from the /tmp/ib_multifrag branch -- something Bad happened in the periodic trunk pulls on that branch making an actual merge back to the trunk effectively impossible (i.e., lots and lots of arbitrary conflicts and artifical changes). :-( == Fine-grain control of queue pair resources == Galen's fine-grain control of queue pair resources to the OpenIB BTL (thanks to Gleb for fixing broken code and providing additional functionality, Pasha for finding broken code, and Jeff for doing all the svn work and regression testing). Prior to this commit, the OpenIB BTL created two queue pairs: one for eager size fragments and one for max send size fragments. When the use of the shared receive queue (SRQ) was specified (via "-mca btl_openib_use_srq 1"), these QPs would use a shared receive queue for receive buffers instead of the default per-peer (PP) receive queues and buffers. One consequence of this design is that receive buffer utilization (the size of the data received as a percentage of the receive buffer used for the data) was quite poor for a number of applications. The new design allows multiple QPs to be specified at runtime. Each QP can be setup to use PP or SRQ receive buffers as well as giving fine-grained control over receive buffer size, number of receive buffers to post, when to replenish the receive queue (low water mark) and for SRQ QPs, the number of outstanding sends can also be specified. The following is an example of the syntax to describe QPs to the OpenIB BTL using the new MCA parameter btl_openib_receive_queues: {{{ -mca btl_openib_receive_queues \ "P,128,16,4;S,1024,256,128,32;S,4096,256,128,32;S,65536,256,128,32" }}} Each QP description is delimited by ";" (semicolon) with individual fields of the QP description delimited by "," (comma). The above example therefore describes 4 QPs. The first QP is: P,128,16,4 Meaning: per-peer receive buffer QPs are indicated by a starting field of "P"; the first QP (shown above) is therefore a per-peer based QP. The second field indicates the size of the receive buffer in bytes (128 bytes). The third field indicates the number of receive buffers to allocate to the QP (16). The fourth field indicates the low watermark for receive buffers at which time the BTL will repost receive buffers to the QP (4). The second QP is: S,1024,256,128,32 Shared receive queue based QPs are indicated by a starting field of "S"; the second QP (shown above) is therefore a shared receive queue based QP. The second, third and fourth fields are the same as in the per-peer based QP. The fifth field is the number of outstanding sends that are allowed at a given time on the QP (32). This provides a "good enough" mechanism of flow control for some regular communication patterns. QPs MUST be specified in ascending receive buffer size order. This requirement may be removed prior to 1.3 release. This commit was SVN r15474.
2007-07-18 05:15:59 +04:00
7, &ival, 0));
if (ival > 7) {
opal_show_help("help-mpi-btl-openib.txt", "invalid mca param value",
true, "btl_openib_ib_retry_count > 7",
"btl_openib_ib_retry_count reset to 7");
ival = 7;
} else if (ival < 0) {
opal_show_help("help-mpi-btl-openib.txt", "invalid mca param value",
true, "btl_openib_ib_retry_count < 0",
"btl_openib_ib_retry_count reset to 0");
ival = 0;
}
Bring over all the work from the /tmp/ib-hw-detect branch. In addition to my design and testing, it was conceptually approved by Gil, Gleb, Pasha, Brad, and Galen. Functionally [probably somewhat lightly] tested by Galen. We may still have to shake out some bugs during the next few months, but it seems to be working for all the cases that I can throw at it. Here's a summary of the changes from that branch: * Move MCA parameter registration to a new file (btl_openib_mca.c): * Properly check the retun status of registering MCA params * Check for valid values of MCA parameters * Make help strings better * Otherwise, the only default value of an MCA param that was changed was max_btls; it went from 4 to -1 (meaning: use all available) * Properly prototyped internal functions in _component.c * Made a bunch of functions static that didn't need to be public * Renamed to remove "mca_" prefix from static functions * Call new MCA param registration function * Call new INI file read/lookup/finalize functions * Updated a bunch of macros to be "BTL_" instead of "ORTE_" * Be a little more consistent with return values * Handle -1 for the max_btls MCA param * Fixed a free() that should have been an OBJ_RELEASE() * Some re-indenting * Added INI-file parsing * New flex file: btl_openib_ini.l * New default HCA params .ini file (probably to be expanded over time by other HCA vendors) * Added more show_help messages for parsing problems * Read in INI files and cache the values for later lookup * When component opens an HCA, lookup to see if any corresponding values were found in the INI files (ID'ed by the HCA vendor_id and vendor_part_id) * Added btl_openib_verbose MCA param that shows what the INI-file stuff does (e.g., shows which MTU your HCA ends up using) * Added btl_openib_hca_param_files as a colon-delimited list of INI files to check for values during startup (in order, left-to-right, just like the MCA base directory param). * MTU is currently the only value supported in this framework. * It is not a fatal error if we don't find params for the HCA in the INI file(s). Instead, just print a warning. New MCA param btl_openib_warn_no_hca_params_found can be used to disable printing the warning. * Add MTU to peer negotiation when making a connection * Exchange maximum MTU; select the lesser of the two This commit was SVN r11182.
2006-08-14 23:30:37 +04:00
mca_btl_openib_component.ib_retry_count = (uint32_t) ival;
CHECK(reg_int("ib_rnr_retry", "InfiniBand \"receiver not ready\" "
"retry count "
"(must be >= 0 and <= 7)",
This commit brings in two major things: 1. Galen's fine-grain control of queue pair resources in the openib BTL. 1. Pasha's new implementation of asychronous HCA event handling. Pasha's new implementation doesn't take much explanation, but the new "multifrag" stuff does. Note that "svn merge" was not used to bring this new code from the /tmp/ib_multifrag branch -- something Bad happened in the periodic trunk pulls on that branch making an actual merge back to the trunk effectively impossible (i.e., lots and lots of arbitrary conflicts and artifical changes). :-( == Fine-grain control of queue pair resources == Galen's fine-grain control of queue pair resources to the OpenIB BTL (thanks to Gleb for fixing broken code and providing additional functionality, Pasha for finding broken code, and Jeff for doing all the svn work and regression testing). Prior to this commit, the OpenIB BTL created two queue pairs: one for eager size fragments and one for max send size fragments. When the use of the shared receive queue (SRQ) was specified (via "-mca btl_openib_use_srq 1"), these QPs would use a shared receive queue for receive buffers instead of the default per-peer (PP) receive queues and buffers. One consequence of this design is that receive buffer utilization (the size of the data received as a percentage of the receive buffer used for the data) was quite poor for a number of applications. The new design allows multiple QPs to be specified at runtime. Each QP can be setup to use PP or SRQ receive buffers as well as giving fine-grained control over receive buffer size, number of receive buffers to post, when to replenish the receive queue (low water mark) and for SRQ QPs, the number of outstanding sends can also be specified. The following is an example of the syntax to describe QPs to the OpenIB BTL using the new MCA parameter btl_openib_receive_queues: {{{ -mca btl_openib_receive_queues \ "P,128,16,4;S,1024,256,128,32;S,4096,256,128,32;S,65536,256,128,32" }}} Each QP description is delimited by ";" (semicolon) with individual fields of the QP description delimited by "," (comma). The above example therefore describes 4 QPs. The first QP is: P,128,16,4 Meaning: per-peer receive buffer QPs are indicated by a starting field of "P"; the first QP (shown above) is therefore a per-peer based QP. The second field indicates the size of the receive buffer in bytes (128 bytes). The third field indicates the number of receive buffers to allocate to the QP (16). The fourth field indicates the low watermark for receive buffers at which time the BTL will repost receive buffers to the QP (4). The second QP is: S,1024,256,128,32 Shared receive queue based QPs are indicated by a starting field of "S"; the second QP (shown above) is therefore a shared receive queue based QP. The second, third and fourth fields are the same as in the per-peer based QP. The fifth field is the number of outstanding sends that are allowed at a given time on the QP (32). This provides a "good enough" mechanism of flow control for some regular communication patterns. QPs MUST be specified in ascending receive buffer size order. This requirement may be removed prior to 1.3 release. This commit was SVN r15474.
2007-07-18 05:15:59 +04:00
7, &ival, 0));
if (ival > 7) {
opal_show_help("help-mpi-btl-openib.txt", "invalid mca param value",
true, "btl_openib_ib_rnr_retry > 7",
"btl_openib_ib_rnr_retry reset to 7");
ival = 7;
} else if (ival < 0) {
opal_show_help("help-mpi-btl-openib.txt", "invalid mca param value",
true, "btl_openib_ib_rnr_retry < 0",
"btl_openib_ib_rnr_retry reset to 0");
ival = 0;
}
Bring over all the work from the /tmp/ib-hw-detect branch. In addition to my design and testing, it was conceptually approved by Gil, Gleb, Pasha, Brad, and Galen. Functionally [probably somewhat lightly] tested by Galen. We may still have to shake out some bugs during the next few months, but it seems to be working for all the cases that I can throw at it. Here's a summary of the changes from that branch: * Move MCA parameter registration to a new file (btl_openib_mca.c): * Properly check the retun status of registering MCA params * Check for valid values of MCA parameters * Make help strings better * Otherwise, the only default value of an MCA param that was changed was max_btls; it went from 4 to -1 (meaning: use all available) * Properly prototyped internal functions in _component.c * Made a bunch of functions static that didn't need to be public * Renamed to remove "mca_" prefix from static functions * Call new MCA param registration function * Call new INI file read/lookup/finalize functions * Updated a bunch of macros to be "BTL_" instead of "ORTE_" * Be a little more consistent with return values * Handle -1 for the max_btls MCA param * Fixed a free() that should have been an OBJ_RELEASE() * Some re-indenting * Added INI-file parsing * New flex file: btl_openib_ini.l * New default HCA params .ini file (probably to be expanded over time by other HCA vendors) * Added more show_help messages for parsing problems * Read in INI files and cache the values for later lookup * When component opens an HCA, lookup to see if any corresponding values were found in the INI files (ID'ed by the HCA vendor_id and vendor_part_id) * Added btl_openib_verbose MCA param that shows what the INI-file stuff does (e.g., shows which MTU your HCA ends up using) * Added btl_openib_hca_param_files as a colon-delimited list of INI files to check for values during startup (in order, left-to-right, just like the MCA base directory param). * MTU is currently the only value supported in this framework. * It is not a fatal error if we don't find params for the HCA in the INI file(s). Instead, just print a warning. New MCA param btl_openib_warn_no_hca_params_found can be used to disable printing the warning. * Add MTU to peer negotiation when making a connection * Exchange maximum MTU; select the lesser of the two This commit was SVN r11182.
2006-08-14 23:30:37 +04:00
mca_btl_openib_component.ib_rnr_retry = (uint32_t) ival;
CHECK(reg_int("ib_max_rdma_dst_ops", "InfiniBand maximum pending RDMA "
"destination operations "
"(must be >= 0)",
4, &ival, REGINT_GE_ZERO));
Bring over all the work from the /tmp/ib-hw-detect branch. In addition to my design and testing, it was conceptually approved by Gil, Gleb, Pasha, Brad, and Galen. Functionally [probably somewhat lightly] tested by Galen. We may still have to shake out some bugs during the next few months, but it seems to be working for all the cases that I can throw at it. Here's a summary of the changes from that branch: * Move MCA parameter registration to a new file (btl_openib_mca.c): * Properly check the retun status of registering MCA params * Check for valid values of MCA parameters * Make help strings better * Otherwise, the only default value of an MCA param that was changed was max_btls; it went from 4 to -1 (meaning: use all available) * Properly prototyped internal functions in _component.c * Made a bunch of functions static that didn't need to be public * Renamed to remove "mca_" prefix from static functions * Call new MCA param registration function * Call new INI file read/lookup/finalize functions * Updated a bunch of macros to be "BTL_" instead of "ORTE_" * Be a little more consistent with return values * Handle -1 for the max_btls MCA param * Fixed a free() that should have been an OBJ_RELEASE() * Some re-indenting * Added INI-file parsing * New flex file: btl_openib_ini.l * New default HCA params .ini file (probably to be expanded over time by other HCA vendors) * Added more show_help messages for parsing problems * Read in INI files and cache the values for later lookup * When component opens an HCA, lookup to see if any corresponding values were found in the INI files (ID'ed by the HCA vendor_id and vendor_part_id) * Added btl_openib_verbose MCA param that shows what the INI-file stuff does (e.g., shows which MTU your HCA ends up using) * Added btl_openib_hca_param_files as a colon-delimited list of INI files to check for values during startup (in order, left-to-right, just like the MCA base directory param). * MTU is currently the only value supported in this framework. * It is not a fatal error if we don't find params for the HCA in the INI file(s). Instead, just print a warning. New MCA param btl_openib_warn_no_hca_params_found can be used to disable printing the warning. * Add MTU to peer negotiation when making a connection * Exchange maximum MTU; select the lesser of the two This commit was SVN r11182.
2006-08-14 23:30:37 +04:00
mca_btl_openib_component.ib_max_rdma_dst_ops = (uint32_t) ival;
CHECK(reg_int("ib_service_level", "InfiniBand service level "
"(must be >= 0 and <= 15)",
0, &ival, 0));
if (ival > 15) {
opal_show_help("help-mpi-btl-openib.txt", "invalid mca param value",
true, "btl_openib_ib_service_level > 15",
"btl_openib_ib_service_level reset to 15");
ival = 15;
} else if (ival < 0) {
opal_show_help("help-mpi-btl-openib.txt", "invalid mca param value",
true, "btl_openib_ib_service_level < 0",
"btl_openib_ib_service_level reset to 0");
ival = 0;
}
Bring over all the work from the /tmp/ib-hw-detect branch. In addition to my design and testing, it was conceptually approved by Gil, Gleb, Pasha, Brad, and Galen. Functionally [probably somewhat lightly] tested by Galen. We may still have to shake out some bugs during the next few months, but it seems to be working for all the cases that I can throw at it. Here's a summary of the changes from that branch: * Move MCA parameter registration to a new file (btl_openib_mca.c): * Properly check the retun status of registering MCA params * Check for valid values of MCA parameters * Make help strings better * Otherwise, the only default value of an MCA param that was changed was max_btls; it went from 4 to -1 (meaning: use all available) * Properly prototyped internal functions in _component.c * Made a bunch of functions static that didn't need to be public * Renamed to remove "mca_" prefix from static functions * Call new MCA param registration function * Call new INI file read/lookup/finalize functions * Updated a bunch of macros to be "BTL_" instead of "ORTE_" * Be a little more consistent with return values * Handle -1 for the max_btls MCA param * Fixed a free() that should have been an OBJ_RELEASE() * Some re-indenting * Added INI-file parsing * New flex file: btl_openib_ini.l * New default HCA params .ini file (probably to be expanded over time by other HCA vendors) * Added more show_help messages for parsing problems * Read in INI files and cache the values for later lookup * When component opens an HCA, lookup to see if any corresponding values were found in the INI files (ID'ed by the HCA vendor_id and vendor_part_id) * Added btl_openib_verbose MCA param that shows what the INI-file stuff does (e.g., shows which MTU your HCA ends up using) * Added btl_openib_hca_param_files as a colon-delimited list of INI files to check for values during startup (in order, left-to-right, just like the MCA base directory param). * MTU is currently the only value supported in this framework. * It is not a fatal error if we don't find params for the HCA in the INI file(s). Instead, just print a warning. New MCA param btl_openib_warn_no_hca_params_found can be used to disable printing the warning. * Add MTU to peer negotiation when making a connection * Exchange maximum MTU; select the lesser of the two This commit was SVN r11182.
2006-08-14 23:30:37 +04:00
mca_btl_openib_component.ib_service_level = (uint32_t) ival;
CHECK(reg_int("use_eager_rdma", "Use RDMA for eager messages",
Bring over all the work from the /tmp/ib-hw-detect branch. In addition to my design and testing, it was conceptually approved by Gil, Gleb, Pasha, Brad, and Galen. Functionally [probably somewhat lightly] tested by Galen. We may still have to shake out some bugs during the next few months, but it seems to be working for all the cases that I can throw at it. Here's a summary of the changes from that branch: * Move MCA parameter registration to a new file (btl_openib_mca.c): * Properly check the retun status of registering MCA params * Check for valid values of MCA parameters * Make help strings better * Otherwise, the only default value of an MCA param that was changed was max_btls; it went from 4 to -1 (meaning: use all available) * Properly prototyped internal functions in _component.c * Made a bunch of functions static that didn't need to be public * Renamed to remove "mca_" prefix from static functions * Call new MCA param registration function * Call new INI file read/lookup/finalize functions * Updated a bunch of macros to be "BTL_" instead of "ORTE_" * Be a little more consistent with return values * Handle -1 for the max_btls MCA param * Fixed a free() that should have been an OBJ_RELEASE() * Some re-indenting * Added INI-file parsing * New flex file: btl_openib_ini.l * New default HCA params .ini file (probably to be expanded over time by other HCA vendors) * Added more show_help messages for parsing problems * Read in INI files and cache the values for later lookup * When component opens an HCA, lookup to see if any corresponding values were found in the INI files (ID'ed by the HCA vendor_id and vendor_part_id) * Added btl_openib_verbose MCA param that shows what the INI-file stuff does (e.g., shows which MTU your HCA ends up using) * Added btl_openib_hca_param_files as a colon-delimited list of INI files to check for values during startup (in order, left-to-right, just like the MCA base directory param). * MTU is currently the only value supported in this framework. * It is not a fatal error if we don't find params for the HCA in the INI file(s). Instead, just print a warning. New MCA param btl_openib_warn_no_hca_params_found can be used to disable printing the warning. * Add MTU to peer negotiation when making a connection * Exchange maximum MTU; select the lesser of the two This commit was SVN r11182.
2006-08-14 23:30:37 +04:00
1, &ival, 0));
mca_btl_openib_component.use_eager_rdma = (uint32_t) (ival != 0);
#if OMPI_ENABLE_PROGRESS_THREADS == 1
/* Fast rdma path isn't supported by PROGRESS_THREAD */
mca_btl_openib_component.use_eager_rdma = 0;
#endif
Bring over all the work from the /tmp/ib-hw-detect branch. In addition to my design and testing, it was conceptually approved by Gil, Gleb, Pasha, Brad, and Galen. Functionally [probably somewhat lightly] tested by Galen. We may still have to shake out some bugs during the next few months, but it seems to be working for all the cases that I can throw at it. Here's a summary of the changes from that branch: * Move MCA parameter registration to a new file (btl_openib_mca.c): * Properly check the retun status of registering MCA params * Check for valid values of MCA parameters * Make help strings better * Otherwise, the only default value of an MCA param that was changed was max_btls; it went from 4 to -1 (meaning: use all available) * Properly prototyped internal functions in _component.c * Made a bunch of functions static that didn't need to be public * Renamed to remove "mca_" prefix from static functions * Call new MCA param registration function * Call new INI file read/lookup/finalize functions * Updated a bunch of macros to be "BTL_" instead of "ORTE_" * Be a little more consistent with return values * Handle -1 for the max_btls MCA param * Fixed a free() that should have been an OBJ_RELEASE() * Some re-indenting * Added INI-file parsing * New flex file: btl_openib_ini.l * New default HCA params .ini file (probably to be expanded over time by other HCA vendors) * Added more show_help messages for parsing problems * Read in INI files and cache the values for later lookup * When component opens an HCA, lookup to see if any corresponding values were found in the INI files (ID'ed by the HCA vendor_id and vendor_part_id) * Added btl_openib_verbose MCA param that shows what the INI-file stuff does (e.g., shows which MTU your HCA ends up using) * Added btl_openib_hca_param_files as a colon-delimited list of INI files to check for values during startup (in order, left-to-right, just like the MCA base directory param). * MTU is currently the only value supported in this framework. * It is not a fatal error if we don't find params for the HCA in the INI file(s). Instead, just print a warning. New MCA param btl_openib_warn_no_hca_params_found can be used to disable printing the warning. * Add MTU to peer negotiation when making a connection * Exchange maximum MTU; select the lesser of the two This commit was SVN r11182.
2006-08-14 23:30:37 +04:00
CHECK(reg_int("eager_rdma_threshold",
"Use RDMA for short messages after this number of "
"messages are received from a given peer "
"(must be >= 1)",
16, &ival, REGINT_GE_ONE));
mca_btl_openib_component.eager_rdma_threshold = (int32_t) ival;
Bring over all the work from the /tmp/ib-hw-detect branch. In addition to my design and testing, it was conceptually approved by Gil, Gleb, Pasha, Brad, and Galen. Functionally [probably somewhat lightly] tested by Galen. We may still have to shake out some bugs during the next few months, but it seems to be working for all the cases that I can throw at it. Here's a summary of the changes from that branch: * Move MCA parameter registration to a new file (btl_openib_mca.c): * Properly check the retun status of registering MCA params * Check for valid values of MCA parameters * Make help strings better * Otherwise, the only default value of an MCA param that was changed was max_btls; it went from 4 to -1 (meaning: use all available) * Properly prototyped internal functions in _component.c * Made a bunch of functions static that didn't need to be public * Renamed to remove "mca_" prefix from static functions * Call new MCA param registration function * Call new INI file read/lookup/finalize functions * Updated a bunch of macros to be "BTL_" instead of "ORTE_" * Be a little more consistent with return values * Handle -1 for the max_btls MCA param * Fixed a free() that should have been an OBJ_RELEASE() * Some re-indenting * Added INI-file parsing * New flex file: btl_openib_ini.l * New default HCA params .ini file (probably to be expanded over time by other HCA vendors) * Added more show_help messages for parsing problems * Read in INI files and cache the values for later lookup * When component opens an HCA, lookup to see if any corresponding values were found in the INI files (ID'ed by the HCA vendor_id and vendor_part_id) * Added btl_openib_verbose MCA param that shows what the INI-file stuff does (e.g., shows which MTU your HCA ends up using) * Added btl_openib_hca_param_files as a colon-delimited list of INI files to check for values during startup (in order, left-to-right, just like the MCA base directory param). * MTU is currently the only value supported in this framework. * It is not a fatal error if we don't find params for the HCA in the INI file(s). Instead, just print a warning. New MCA param btl_openib_warn_no_hca_params_found can be used to disable printing the warning. * Add MTU to peer negotiation when making a connection * Exchange maximum MTU; select the lesser of the two This commit was SVN r11182.
2006-08-14 23:30:37 +04:00
CHECK(reg_int("max_eager_rdma", "Maximum number of peers allowed to use "
"RDMA for short messages (RDMA is used for all long "
"messages, except if explicitly disabled, such as "
"with the \"dr\" pml) "
"(must be >= 0)",
16, &ival, REGINT_GE_ZERO));
mca_btl_openib_component.max_eager_rdma = (int32_t) ival;
Bring over all the work from the /tmp/ib-hw-detect branch. In addition to my design and testing, it was conceptually approved by Gil, Gleb, Pasha, Brad, and Galen. Functionally [probably somewhat lightly] tested by Galen. We may still have to shake out some bugs during the next few months, but it seems to be working for all the cases that I can throw at it. Here's a summary of the changes from that branch: * Move MCA parameter registration to a new file (btl_openib_mca.c): * Properly check the retun status of registering MCA params * Check for valid values of MCA parameters * Make help strings better * Otherwise, the only default value of an MCA param that was changed was max_btls; it went from 4 to -1 (meaning: use all available) * Properly prototyped internal functions in _component.c * Made a bunch of functions static that didn't need to be public * Renamed to remove "mca_" prefix from static functions * Call new MCA param registration function * Call new INI file read/lookup/finalize functions * Updated a bunch of macros to be "BTL_" instead of "ORTE_" * Be a little more consistent with return values * Handle -1 for the max_btls MCA param * Fixed a free() that should have been an OBJ_RELEASE() * Some re-indenting * Added INI-file parsing * New flex file: btl_openib_ini.l * New default HCA params .ini file (probably to be expanded over time by other HCA vendors) * Added more show_help messages for parsing problems * Read in INI files and cache the values for later lookup * When component opens an HCA, lookup to see if any corresponding values were found in the INI files (ID'ed by the HCA vendor_id and vendor_part_id) * Added btl_openib_verbose MCA param that shows what the INI-file stuff does (e.g., shows which MTU your HCA ends up using) * Added btl_openib_hca_param_files as a colon-delimited list of INI files to check for values during startup (in order, left-to-right, just like the MCA base directory param). * MTU is currently the only value supported in this framework. * It is not a fatal error if we don't find params for the HCA in the INI file(s). Instead, just print a warning. New MCA param btl_openib_warn_no_hca_params_found can be used to disable printing the warning. * Add MTU to peer negotiation when making a connection * Exchange maximum MTU; select the lesser of the two This commit was SVN r11182.
2006-08-14 23:30:37 +04:00
CHECK(reg_int("eager_rdma_num", "Number of RDMA buffers to allocate "
"for small messages"
"(must be >= 1)",
16, &ival, REGINT_GE_ONE));
mca_btl_openib_component.eager_rdma_num = (int32_t) (ival + 1);
Bring over all the work from the /tmp/ib-hw-detect branch. In addition to my design and testing, it was conceptually approved by Gil, Gleb, Pasha, Brad, and Galen. Functionally [probably somewhat lightly] tested by Galen. We may still have to shake out some bugs during the next few months, but it seems to be working for all the cases that I can throw at it. Here's a summary of the changes from that branch: * Move MCA parameter registration to a new file (btl_openib_mca.c): * Properly check the retun status of registering MCA params * Check for valid values of MCA parameters * Make help strings better * Otherwise, the only default value of an MCA param that was changed was max_btls; it went from 4 to -1 (meaning: use all available) * Properly prototyped internal functions in _component.c * Made a bunch of functions static that didn't need to be public * Renamed to remove "mca_" prefix from static functions * Call new MCA param registration function * Call new INI file read/lookup/finalize functions * Updated a bunch of macros to be "BTL_" instead of "ORTE_" * Be a little more consistent with return values * Handle -1 for the max_btls MCA param * Fixed a free() that should have been an OBJ_RELEASE() * Some re-indenting * Added INI-file parsing * New flex file: btl_openib_ini.l * New default HCA params .ini file (probably to be expanded over time by other HCA vendors) * Added more show_help messages for parsing problems * Read in INI files and cache the values for later lookup * When component opens an HCA, lookup to see if any corresponding values were found in the INI files (ID'ed by the HCA vendor_id and vendor_part_id) * Added btl_openib_verbose MCA param that shows what the INI-file stuff does (e.g., shows which MTU your HCA ends up using) * Added btl_openib_hca_param_files as a colon-delimited list of INI files to check for values during startup (in order, left-to-right, just like the MCA base directory param). * MTU is currently the only value supported in this framework. * It is not a fatal error if we don't find params for the HCA in the INI file(s). Instead, just print a warning. New MCA param btl_openib_warn_no_hca_params_found can be used to disable printing the warning. * Add MTU to peer negotiation when making a connection * Exchange maximum MTU; select the lesser of the two This commit was SVN r11182.
2006-08-14 23:30:37 +04:00
CHECK(reg_int("btls_per_lid", "Number of BTLs to create for each "
"InfiniBand LID "
"(must be >= 1)",
1, &ival, REGINT_GE_ONE));
mca_btl_openib_component.btls_per_lid = (uint32_t) ival;
CHECK(reg_int("max_lmc", "Maximum number of LIDs to use for each HCA port "
"(must be >= 0, where 0 = use all available)",
0, &ival, REGINT_GE_ZERO));
mca_btl_openib_component.max_lmc = (uint32_t) ival;
This commit brings in two major things: 1. Galen's fine-grain control of queue pair resources in the openib BTL. 1. Pasha's new implementation of asychronous HCA event handling. Pasha's new implementation doesn't take much explanation, but the new "multifrag" stuff does. Note that "svn merge" was not used to bring this new code from the /tmp/ib_multifrag branch -- something Bad happened in the periodic trunk pulls on that branch making an actual merge back to the trunk effectively impossible (i.e., lots and lots of arbitrary conflicts and artifical changes). :-( == Fine-grain control of queue pair resources == Galen's fine-grain control of queue pair resources to the OpenIB BTL (thanks to Gleb for fixing broken code and providing additional functionality, Pasha for finding broken code, and Jeff for doing all the svn work and regression testing). Prior to this commit, the OpenIB BTL created two queue pairs: one for eager size fragments and one for max send size fragments. When the use of the shared receive queue (SRQ) was specified (via "-mca btl_openib_use_srq 1"), these QPs would use a shared receive queue for receive buffers instead of the default per-peer (PP) receive queues and buffers. One consequence of this design is that receive buffer utilization (the size of the data received as a percentage of the receive buffer used for the data) was quite poor for a number of applications. The new design allows multiple QPs to be specified at runtime. Each QP can be setup to use PP or SRQ receive buffers as well as giving fine-grained control over receive buffer size, number of receive buffers to post, when to replenish the receive queue (low water mark) and for SRQ QPs, the number of outstanding sends can also be specified. The following is an example of the syntax to describe QPs to the OpenIB BTL using the new MCA parameter btl_openib_receive_queues: {{{ -mca btl_openib_receive_queues \ "P,128,16,4;S,1024,256,128,32;S,4096,256,128,32;S,65536,256,128,32" }}} Each QP description is delimited by ";" (semicolon) with individual fields of the QP description delimited by "," (comma). The above example therefore describes 4 QPs. The first QP is: P,128,16,4 Meaning: per-peer receive buffer QPs are indicated by a starting field of "P"; the first QP (shown above) is therefore a per-peer based QP. The second field indicates the size of the receive buffer in bytes (128 bytes). The third field indicates the number of receive buffers to allocate to the QP (16). The fourth field indicates the low watermark for receive buffers at which time the BTL will repost receive buffers to the QP (4). The second QP is: S,1024,256,128,32 Shared receive queue based QPs are indicated by a starting field of "S"; the second QP (shown above) is therefore a shared receive queue based QP. The second, third and fourth fields are the same as in the per-peer based QP. The fifth field is the number of outstanding sends that are allowed at a given time on the QP (32). This provides a "good enough" mechanism of flow control for some regular communication patterns. QPs MUST be specified in ascending receive buffer size order. This requirement may be removed prior to 1.3 release. This commit was SVN r15474.
2007-07-18 05:15:59 +04:00
#if OMPI_HAVE_THREADS
CHECK(reg_int("use_async_event_thread",
"If nonzero, use the thread that will handle InfiniBand asyncihronous events ",
1, &ival, 0));
mca_btl_openib_component.use_async_event_thread = (0 != ival);
#endif
Bring over all the work from the /tmp/ib-hw-detect branch. In addition to my design and testing, it was conceptually approved by Gil, Gleb, Pasha, Brad, and Galen. Functionally [probably somewhat lightly] tested by Galen. We may still have to shake out some bugs during the next few months, but it seems to be working for all the cases that I can throw at it. Here's a summary of the changes from that branch: * Move MCA parameter registration to a new file (btl_openib_mca.c): * Properly check the retun status of registering MCA params * Check for valid values of MCA parameters * Make help strings better * Otherwise, the only default value of an MCA param that was changed was max_btls; it went from 4 to -1 (meaning: use all available) * Properly prototyped internal functions in _component.c * Made a bunch of functions static that didn't need to be public * Renamed to remove "mca_" prefix from static functions * Call new MCA param registration function * Call new INI file read/lookup/finalize functions * Updated a bunch of macros to be "BTL_" instead of "ORTE_" * Be a little more consistent with return values * Handle -1 for the max_btls MCA param * Fixed a free() that should have been an OBJ_RELEASE() * Some re-indenting * Added INI-file parsing * New flex file: btl_openib_ini.l * New default HCA params .ini file (probably to be expanded over time by other HCA vendors) * Added more show_help messages for parsing problems * Read in INI files and cache the values for later lookup * When component opens an HCA, lookup to see if any corresponding values were found in the INI files (ID'ed by the HCA vendor_id and vendor_part_id) * Added btl_openib_verbose MCA param that shows what the INI-file stuff does (e.g., shows which MTU your HCA ends up using) * Added btl_openib_hca_param_files as a colon-delimited list of INI files to check for values during startup (in order, left-to-right, just like the MCA base directory param). * MTU is currently the only value supported in this framework. * It is not a fatal error if we don't find params for the HCA in the INI file(s). Instead, just print a warning. New MCA param btl_openib_warn_no_hca_params_found can be used to disable printing the warning. * Add MTU to peer negotiation when making a connection * Exchange maximum MTU; select the lesser of the two This commit was SVN r11182.
2006-08-14 23:30:37 +04:00
CHECK(reg_int("buffer_alignment",
"Prefered communication buffer alignment, in bytes "
"(must be > 0 and power of two)",
Bring over all the work from the /tmp/ib-hw-detect branch. In addition to my design and testing, it was conceptually approved by Gil, Gleb, Pasha, Brad, and Galen. Functionally [probably somewhat lightly] tested by Galen. We may still have to shake out some bugs during the next few months, but it seems to be working for all the cases that I can throw at it. Here's a summary of the changes from that branch: * Move MCA parameter registration to a new file (btl_openib_mca.c): * Properly check the retun status of registering MCA params * Check for valid values of MCA parameters * Make help strings better * Otherwise, the only default value of an MCA param that was changed was max_btls; it went from 4 to -1 (meaning: use all available) * Properly prototyped internal functions in _component.c * Made a bunch of functions static that didn't need to be public * Renamed to remove "mca_" prefix from static functions * Call new MCA param registration function * Call new INI file read/lookup/finalize functions * Updated a bunch of macros to be "BTL_" instead of "ORTE_" * Be a little more consistent with return values * Handle -1 for the max_btls MCA param * Fixed a free() that should have been an OBJ_RELEASE() * Some re-indenting * Added INI-file parsing * New flex file: btl_openib_ini.l * New default HCA params .ini file (probably to be expanded over time by other HCA vendors) * Added more show_help messages for parsing problems * Read in INI files and cache the values for later lookup * When component opens an HCA, lookup to see if any corresponding values were found in the INI files (ID'ed by the HCA vendor_id and vendor_part_id) * Added btl_openib_verbose MCA param that shows what the INI-file stuff does (e.g., shows which MTU your HCA ends up using) * Added btl_openib_hca_param_files as a colon-delimited list of INI files to check for values during startup (in order, left-to-right, just like the MCA base directory param). * MTU is currently the only value supported in this framework. * It is not a fatal error if we don't find params for the HCA in the INI file(s). Instead, just print a warning. New MCA param btl_openib_warn_no_hca_params_found can be used to disable printing the warning. * Add MTU to peer negotiation when making a connection * Exchange maximum MTU; select the lesser of the two This commit was SVN r11182.
2006-08-14 23:30:37 +04:00
64, &ival, REGINT_GE_ZERO));
if(ival <= 1 || (ival & (ival - 1))) {
opal_show_help("help-mpi-btl-openib.txt", "wrong buffer alignment",
true, ival, orte_system_info.nodename, 64);
mca_btl_openib_component.buffer_alignment = 64;
} else {
mca_btl_openib_component.buffer_alignment = (uint32_t) ival;
}
This commit brings in two major things: 1. Galen's fine-grain control of queue pair resources in the openib BTL. 1. Pasha's new implementation of asychronous HCA event handling. Pasha's new implementation doesn't take much explanation, but the new "multifrag" stuff does. Note that "svn merge" was not used to bring this new code from the /tmp/ib_multifrag branch -- something Bad happened in the periodic trunk pulls on that branch making an actual merge back to the trunk effectively impossible (i.e., lots and lots of arbitrary conflicts and artifical changes). :-( == Fine-grain control of queue pair resources == Galen's fine-grain control of queue pair resources to the OpenIB BTL (thanks to Gleb for fixing broken code and providing additional functionality, Pasha for finding broken code, and Jeff for doing all the svn work and regression testing). Prior to this commit, the OpenIB BTL created two queue pairs: one for eager size fragments and one for max send size fragments. When the use of the shared receive queue (SRQ) was specified (via "-mca btl_openib_use_srq 1"), these QPs would use a shared receive queue for receive buffers instead of the default per-peer (PP) receive queues and buffers. One consequence of this design is that receive buffer utilization (the size of the data received as a percentage of the receive buffer used for the data) was quite poor for a number of applications. The new design allows multiple QPs to be specified at runtime. Each QP can be setup to use PP or SRQ receive buffers as well as giving fine-grained control over receive buffer size, number of receive buffers to post, when to replenish the receive queue (low water mark) and for SRQ QPs, the number of outstanding sends can also be specified. The following is an example of the syntax to describe QPs to the OpenIB BTL using the new MCA parameter btl_openib_receive_queues: {{{ -mca btl_openib_receive_queues \ "P,128,16,4;S,1024,256,128,32;S,4096,256,128,32;S,65536,256,128,32" }}} Each QP description is delimited by ";" (semicolon) with individual fields of the QP description delimited by "," (comma). The above example therefore describes 4 QPs. The first QP is: P,128,16,4 Meaning: per-peer receive buffer QPs are indicated by a starting field of "P"; the first QP (shown above) is therefore a per-peer based QP. The second field indicates the size of the receive buffer in bytes (128 bytes). The third field indicates the number of receive buffers to allocate to the QP (16). The fourth field indicates the low watermark for receive buffers at which time the BTL will repost receive buffers to the QP (4). The second QP is: S,1024,256,128,32 Shared receive queue based QPs are indicated by a starting field of "S"; the second QP (shown above) is therefore a shared receive queue based QP. The second, third and fourth fields are the same as in the per-peer based QP. The fifth field is the number of outstanding sends that are allowed at a given time on the QP (32). This provides a "good enough" mechanism of flow control for some regular communication patterns. QPs MUST be specified in ascending receive buffer size order. This requirement may be removed prior to 1.3 release. This commit was SVN r15474.
2007-07-18 05:15:59 +04:00
CHECK(reg_int("use_message_coalescing",
"Use message coalescing", 1, &ival, 0));
mca_btl_openib_component.use_message_coalescing = (0 != ival);
/* Info only */
mca_base_param_reg_int(&mca_btl_openib_component.super.btl_version,
"have_fork_support",
"Whether the OpenFabrics stack supports applications that invoke the \"fork()\" system call or not (0 = no, 1 = yes). Note that this value does NOT indicate whether the system being run on supports \"fork()\" with OpenFabrics applications or not.",
false, true,
#ifdef HAVE_IBV_FORK_INIT
1,
#else
0,
#endif
NULL);
mca_btl_openib_module.super.btl_exclusivity = MCA_BTL_EXCLUSIVITY_DEFAULT;
This commit brings in two major things: 1. Galen's fine-grain control of queue pair resources in the openib BTL. 1. Pasha's new implementation of asychronous HCA event handling. Pasha's new implementation doesn't take much explanation, but the new "multifrag" stuff does. Note that "svn merge" was not used to bring this new code from the /tmp/ib_multifrag branch -- something Bad happened in the periodic trunk pulls on that branch making an actual merge back to the trunk effectively impossible (i.e., lots and lots of arbitrary conflicts and artifical changes). :-( == Fine-grain control of queue pair resources == Galen's fine-grain control of queue pair resources to the OpenIB BTL (thanks to Gleb for fixing broken code and providing additional functionality, Pasha for finding broken code, and Jeff for doing all the svn work and regression testing). Prior to this commit, the OpenIB BTL created two queue pairs: one for eager size fragments and one for max send size fragments. When the use of the shared receive queue (SRQ) was specified (via "-mca btl_openib_use_srq 1"), these QPs would use a shared receive queue for receive buffers instead of the default per-peer (PP) receive queues and buffers. One consequence of this design is that receive buffer utilization (the size of the data received as a percentage of the receive buffer used for the data) was quite poor for a number of applications. The new design allows multiple QPs to be specified at runtime. Each QP can be setup to use PP or SRQ receive buffers as well as giving fine-grained control over receive buffer size, number of receive buffers to post, when to replenish the receive queue (low water mark) and for SRQ QPs, the number of outstanding sends can also be specified. The following is an example of the syntax to describe QPs to the OpenIB BTL using the new MCA parameter btl_openib_receive_queues: {{{ -mca btl_openib_receive_queues \ "P,128,16,4;S,1024,256,128,32;S,4096,256,128,32;S,65536,256,128,32" }}} Each QP description is delimited by ";" (semicolon) with individual fields of the QP description delimited by "," (comma). The above example therefore describes 4 QPs. The first QP is: P,128,16,4 Meaning: per-peer receive buffer QPs are indicated by a starting field of "P"; the first QP (shown above) is therefore a per-peer based QP. The second field indicates the size of the receive buffer in bytes (128 bytes). The third field indicates the number of receive buffers to allocate to the QP (16). The fourth field indicates the low watermark for receive buffers at which time the BTL will repost receive buffers to the QP (4). The second QP is: S,1024,256,128,32 Shared receive queue based QPs are indicated by a starting field of "S"; the second QP (shown above) is therefore a shared receive queue based QP. The second, third and fourth fields are the same as in the per-peer based QP. The fifth field is the number of outstanding sends that are allowed at a given time on the QP (32). This provides a "good enough" mechanism of flow control for some regular communication patterns. QPs MUST be specified in ascending receive buffer size order. This requirement may be removed prior to 1.3 release. This commit was SVN r15474.
2007-07-18 05:15:59 +04:00
mca_btl_openib_module.super.btl_eager_limit = 12 * 1024;
mca_btl_openib_module.super.btl_rndv_eager_limit = 12 * 1024;
mca_btl_openib_module.super.btl_max_send_size = 64 * 1024;
mca_btl_openib_module.super.btl_rdma_pipeline_send_length = 1024 * 1024;
mca_btl_openib_module.super.btl_rdma_pipeline_frag_size = 1024 * 1024;
mca_btl_openib_module.super.btl_min_rdma_pipeline_size = 256 * 1024;
mca_btl_openib_module.super.btl_flags = MCA_BTL_FLAGS_RDMA |
MCA_BTL_FLAGS_NEED_ACK | MCA_BTL_FLAGS_NEED_CSUM | MCA_BTL_FLAGS_HETEROGENEOUS_RDMA;
mca_btl_openib_module.super.btl_bandwidth = 800;
mca_btl_openib_module.super.btl_latency = 10;
This commit brings in two major things: 1. Galen's fine-grain control of queue pair resources in the openib BTL. 1. Pasha's new implementation of asychronous HCA event handling. Pasha's new implementation doesn't take much explanation, but the new "multifrag" stuff does. Note that "svn merge" was not used to bring this new code from the /tmp/ib_multifrag branch -- something Bad happened in the periodic trunk pulls on that branch making an actual merge back to the trunk effectively impossible (i.e., lots and lots of arbitrary conflicts and artifical changes). :-( == Fine-grain control of queue pair resources == Galen's fine-grain control of queue pair resources to the OpenIB BTL (thanks to Gleb for fixing broken code and providing additional functionality, Pasha for finding broken code, and Jeff for doing all the svn work and regression testing). Prior to this commit, the OpenIB BTL created two queue pairs: one for eager size fragments and one for max send size fragments. When the use of the shared receive queue (SRQ) was specified (via "-mca btl_openib_use_srq 1"), these QPs would use a shared receive queue for receive buffers instead of the default per-peer (PP) receive queues and buffers. One consequence of this design is that receive buffer utilization (the size of the data received as a percentage of the receive buffer used for the data) was quite poor for a number of applications. The new design allows multiple QPs to be specified at runtime. Each QP can be setup to use PP or SRQ receive buffers as well as giving fine-grained control over receive buffer size, number of receive buffers to post, when to replenish the receive queue (low water mark) and for SRQ QPs, the number of outstanding sends can also be specified. The following is an example of the syntax to describe QPs to the OpenIB BTL using the new MCA parameter btl_openib_receive_queues: {{{ -mca btl_openib_receive_queues \ "P,128,16,4;S,1024,256,128,32;S,4096,256,128,32;S,65536,256,128,32" }}} Each QP description is delimited by ";" (semicolon) with individual fields of the QP description delimited by "," (comma). The above example therefore describes 4 QPs. The first QP is: P,128,16,4 Meaning: per-peer receive buffer QPs are indicated by a starting field of "P"; the first QP (shown above) is therefore a per-peer based QP. The second field indicates the size of the receive buffer in bytes (128 bytes). The third field indicates the number of receive buffers to allocate to the QP (16). The fourth field indicates the low watermark for receive buffers at which time the BTL will repost receive buffers to the QP (4). The second QP is: S,1024,256,128,32 Shared receive queue based QPs are indicated by a starting field of "S"; the second QP (shown above) is therefore a shared receive queue based QP. The second, third and fourth fields are the same as in the per-peer based QP. The fifth field is the number of outstanding sends that are allowed at a given time on the QP (32). This provides a "good enough" mechanism of flow control for some regular communication patterns. QPs MUST be specified in ascending receive buffer size order. This requirement may be removed prior to 1.3 release. This commit was SVN r15474.
2007-07-18 05:15:59 +04:00
ret = mca_btl_base_param_register(
&mca_btl_openib_component.super.btl_version,
&mca_btl_openib_module.super);
if (OMPI_SUCCESS != ret) {
This commit brings in two major things: 1. Galen's fine-grain control of queue pair resources in the openib BTL. 1. Pasha's new implementation of asychronous HCA event handling. Pasha's new implementation doesn't take much explanation, but the new "multifrag" stuff does. Note that "svn merge" was not used to bring this new code from the /tmp/ib_multifrag branch -- something Bad happened in the periodic trunk pulls on that branch making an actual merge back to the trunk effectively impossible (i.e., lots and lots of arbitrary conflicts and artifical changes). :-( == Fine-grain control of queue pair resources == Galen's fine-grain control of queue pair resources to the OpenIB BTL (thanks to Gleb for fixing broken code and providing additional functionality, Pasha for finding broken code, and Jeff for doing all the svn work and regression testing). Prior to this commit, the OpenIB BTL created two queue pairs: one for eager size fragments and one for max send size fragments. When the use of the shared receive queue (SRQ) was specified (via "-mca btl_openib_use_srq 1"), these QPs would use a shared receive queue for receive buffers instead of the default per-peer (PP) receive queues and buffers. One consequence of this design is that receive buffer utilization (the size of the data received as a percentage of the receive buffer used for the data) was quite poor for a number of applications. The new design allows multiple QPs to be specified at runtime. Each QP can be setup to use PP or SRQ receive buffers as well as giving fine-grained control over receive buffer size, number of receive buffers to post, when to replenish the receive queue (low water mark) and for SRQ QPs, the number of outstanding sends can also be specified. The following is an example of the syntax to describe QPs to the OpenIB BTL using the new MCA parameter btl_openib_receive_queues: {{{ -mca btl_openib_receive_queues \ "P,128,16,4;S,1024,256,128,32;S,4096,256,128,32;S,65536,256,128,32" }}} Each QP description is delimited by ";" (semicolon) with individual fields of the QP description delimited by "," (comma). The above example therefore describes 4 QPs. The first QP is: P,128,16,4 Meaning: per-peer receive buffer QPs are indicated by a starting field of "P"; the first QP (shown above) is therefore a per-peer based QP. The second field indicates the size of the receive buffer in bytes (128 bytes). The third field indicates the number of receive buffers to allocate to the QP (16). The fourth field indicates the low watermark for receive buffers at which time the BTL will repost receive buffers to the QP (4). The second QP is: S,1024,256,128,32 Shared receive queue based QPs are indicated by a starting field of "S"; the second QP (shown above) is therefore a shared receive queue based QP. The second, third and fourth fields are the same as in the per-peer based QP. The fifth field is the number of outstanding sends that are allowed at a given time on the QP (32). This provides a "good enough" mechanism of flow control for some regular communication patterns. QPs MUST be specified in ascending receive buffer size order. This requirement may be removed prior to 1.3 release. This commit was SVN r15474.
2007-07-18 05:15:59 +04:00
return ret;
}
This commit brings in two major things: 1. Galen's fine-grain control of queue pair resources in the openib BTL. 1. Pasha's new implementation of asychronous HCA event handling. Pasha's new implementation doesn't take much explanation, but the new "multifrag" stuff does. Note that "svn merge" was not used to bring this new code from the /tmp/ib_multifrag branch -- something Bad happened in the periodic trunk pulls on that branch making an actual merge back to the trunk effectively impossible (i.e., lots and lots of arbitrary conflicts and artifical changes). :-( == Fine-grain control of queue pair resources == Galen's fine-grain control of queue pair resources to the OpenIB BTL (thanks to Gleb for fixing broken code and providing additional functionality, Pasha for finding broken code, and Jeff for doing all the svn work and regression testing). Prior to this commit, the OpenIB BTL created two queue pairs: one for eager size fragments and one for max send size fragments. When the use of the shared receive queue (SRQ) was specified (via "-mca btl_openib_use_srq 1"), these QPs would use a shared receive queue for receive buffers instead of the default per-peer (PP) receive queues and buffers. One consequence of this design is that receive buffer utilization (the size of the data received as a percentage of the receive buffer used for the data) was quite poor for a number of applications. The new design allows multiple QPs to be specified at runtime. Each QP can be setup to use PP or SRQ receive buffers as well as giving fine-grained control over receive buffer size, number of receive buffers to post, when to replenish the receive queue (low water mark) and for SRQ QPs, the number of outstanding sends can also be specified. The following is an example of the syntax to describe QPs to the OpenIB BTL using the new MCA parameter btl_openib_receive_queues: {{{ -mca btl_openib_receive_queues \ "P,128,16,4;S,1024,256,128,32;S,4096,256,128,32;S,65536,256,128,32" }}} Each QP description is delimited by ";" (semicolon) with individual fields of the QP description delimited by "," (comma). The above example therefore describes 4 QPs. The first QP is: P,128,16,4 Meaning: per-peer receive buffer QPs are indicated by a starting field of "P"; the first QP (shown above) is therefore a per-peer based QP. The second field indicates the size of the receive buffer in bytes (128 bytes). The third field indicates the number of receive buffers to allocate to the QP (16). The fourth field indicates the low watermark for receive buffers at which time the BTL will repost receive buffers to the QP (4). The second QP is: S,1024,256,128,32 Shared receive queue based QPs are indicated by a starting field of "S"; the second QP (shown above) is therefore a shared receive queue based QP. The second, third and fourth fields are the same as in the per-peer based QP. The fifth field is the number of outstanding sends that are allowed at a given time on the QP (32). This provides a "good enough" mechanism of flow control for some regular communication patterns. QPs MUST be specified in ascending receive buffer size order. This requirement may be removed prior to 1.3 release. This commit was SVN r15474.
2007-07-18 05:15:59 +04:00
/* setup all the qp stuff */
if (OMPI_SUCCESS != (ret = mca_btl_openib_mca_setup_qps())) {
This commit brings in two major things: 1. Galen's fine-grain control of queue pair resources in the openib BTL. 1. Pasha's new implementation of asychronous HCA event handling. Pasha's new implementation doesn't take much explanation, but the new "multifrag" stuff does. Note that "svn merge" was not used to bring this new code from the /tmp/ib_multifrag branch -- something Bad happened in the periodic trunk pulls on that branch making an actual merge back to the trunk effectively impossible (i.e., lots and lots of arbitrary conflicts and artifical changes). :-( == Fine-grain control of queue pair resources == Galen's fine-grain control of queue pair resources to the OpenIB BTL (thanks to Gleb for fixing broken code and providing additional functionality, Pasha for finding broken code, and Jeff for doing all the svn work and regression testing). Prior to this commit, the OpenIB BTL created two queue pairs: one for eager size fragments and one for max send size fragments. When the use of the shared receive queue (SRQ) was specified (via "-mca btl_openib_use_srq 1"), these QPs would use a shared receive queue for receive buffers instead of the default per-peer (PP) receive queues and buffers. One consequence of this design is that receive buffer utilization (the size of the data received as a percentage of the receive buffer used for the data) was quite poor for a number of applications. The new design allows multiple QPs to be specified at runtime. Each QP can be setup to use PP or SRQ receive buffers as well as giving fine-grained control over receive buffer size, number of receive buffers to post, when to replenish the receive queue (low water mark) and for SRQ QPs, the number of outstanding sends can also be specified. The following is an example of the syntax to describe QPs to the OpenIB BTL using the new MCA parameter btl_openib_receive_queues: {{{ -mca btl_openib_receive_queues \ "P,128,16,4;S,1024,256,128,32;S,4096,256,128,32;S,65536,256,128,32" }}} Each QP description is delimited by ";" (semicolon) with individual fields of the QP description delimited by "," (comma). The above example therefore describes 4 QPs. The first QP is: P,128,16,4 Meaning: per-peer receive buffer QPs are indicated by a starting field of "P"; the first QP (shown above) is therefore a per-peer based QP. The second field indicates the size of the receive buffer in bytes (128 bytes). The third field indicates the number of receive buffers to allocate to the QP (16). The fourth field indicates the low watermark for receive buffers at which time the BTL will repost receive buffers to the QP (4). The second QP is: S,1024,256,128,32 Shared receive queue based QPs are indicated by a starting field of "S"; the second QP (shown above) is therefore a shared receive queue based QP. The second, third and fourth fields are the same as in the per-peer based QP. The fifth field is the number of outstanding sends that are allowed at a given time on the QP (32). This provides a "good enough" mechanism of flow control for some regular communication patterns. QPs MUST be specified in ascending receive buffer size order. This requirement may be removed prior to 1.3 release. This commit was SVN r15474.
2007-07-18 05:15:59 +04:00
return ret;
}
This commit brings in two major things: 1. Galen's fine-grain control of queue pair resources in the openib BTL. 1. Pasha's new implementation of asychronous HCA event handling. Pasha's new implementation doesn't take much explanation, but the new "multifrag" stuff does. Note that "svn merge" was not used to bring this new code from the /tmp/ib_multifrag branch -- something Bad happened in the periodic trunk pulls on that branch making an actual merge back to the trunk effectively impossible (i.e., lots and lots of arbitrary conflicts and artifical changes). :-( == Fine-grain control of queue pair resources == Galen's fine-grain control of queue pair resources to the OpenIB BTL (thanks to Gleb for fixing broken code and providing additional functionality, Pasha for finding broken code, and Jeff for doing all the svn work and regression testing). Prior to this commit, the OpenIB BTL created two queue pairs: one for eager size fragments and one for max send size fragments. When the use of the shared receive queue (SRQ) was specified (via "-mca btl_openib_use_srq 1"), these QPs would use a shared receive queue for receive buffers instead of the default per-peer (PP) receive queues and buffers. One consequence of this design is that receive buffer utilization (the size of the data received as a percentage of the receive buffer used for the data) was quite poor for a number of applications. The new design allows multiple QPs to be specified at runtime. Each QP can be setup to use PP or SRQ receive buffers as well as giving fine-grained control over receive buffer size, number of receive buffers to post, when to replenish the receive queue (low water mark) and for SRQ QPs, the number of outstanding sends can also be specified. The following is an example of the syntax to describe QPs to the OpenIB BTL using the new MCA parameter btl_openib_receive_queues: {{{ -mca btl_openib_receive_queues \ "P,128,16,4;S,1024,256,128,32;S,4096,256,128,32;S,65536,256,128,32" }}} Each QP description is delimited by ";" (semicolon) with individual fields of the QP description delimited by "," (comma). The above example therefore describes 4 QPs. The first QP is: P,128,16,4 Meaning: per-peer receive buffer QPs are indicated by a starting field of "P"; the first QP (shown above) is therefore a per-peer based QP. The second field indicates the size of the receive buffer in bytes (128 bytes). The third field indicates the number of receive buffers to allocate to the QP (16). The fourth field indicates the low watermark for receive buffers at which time the BTL will repost receive buffers to the QP (4). The second QP is: S,1024,256,128,32 Shared receive queue based QPs are indicated by a starting field of "S"; the second QP (shown above) is therefore a shared receive queue based QP. The second, third and fourth fields are the same as in the per-peer based QP. The fifth field is the number of outstanding sends that are allowed at a given time on the QP (32). This provides a "good enough" mechanism of flow control for some regular communication patterns. QPs MUST be specified in ascending receive buffer size order. This requirement may be removed prior to 1.3 release. This commit was SVN r15474.
2007-07-18 05:15:59 +04:00
CHECK(reg_string("if_include",
"Comma-delimited list of HCAs/ports to be used (e.g. \"mthca0,mthca1:2\"; empty value means to use all ports found). Mutually exclusive with btl_openib_if_exclude.",
NULL, &mca_btl_openib_component.if_include,
0));
CHECK(reg_string("if_exclude",
"Comma-delimited list of HCAs/ports to be excluded (empty value means to not exclude any ports). Mutually exclusive with btl_openib_if_include.",
NULL, &mca_btl_openib_component.if_exclude,
0));
Bring over all the work from the /tmp/ib-hw-detect branch. In addition to my design and testing, it was conceptually approved by Gil, Gleb, Pasha, Brad, and Galen. Functionally [probably somewhat lightly] tested by Galen. We may still have to shake out some bugs during the next few months, but it seems to be working for all the cases that I can throw at it. Here's a summary of the changes from that branch: * Move MCA parameter registration to a new file (btl_openib_mca.c): * Properly check the retun status of registering MCA params * Check for valid values of MCA parameters * Make help strings better * Otherwise, the only default value of an MCA param that was changed was max_btls; it went from 4 to -1 (meaning: use all available) * Properly prototyped internal functions in _component.c * Made a bunch of functions static that didn't need to be public * Renamed to remove "mca_" prefix from static functions * Call new MCA param registration function * Call new INI file read/lookup/finalize functions * Updated a bunch of macros to be "BTL_" instead of "ORTE_" * Be a little more consistent with return values * Handle -1 for the max_btls MCA param * Fixed a free() that should have been an OBJ_RELEASE() * Some re-indenting * Added INI-file parsing * New flex file: btl_openib_ini.l * New default HCA params .ini file (probably to be expanded over time by other HCA vendors) * Added more show_help messages for parsing problems * Read in INI files and cache the values for later lookup * When component opens an HCA, lookup to see if any corresponding values were found in the INI files (ID'ed by the HCA vendor_id and vendor_part_id) * Added btl_openib_verbose MCA param that shows what the INI-file stuff does (e.g., shows which MTU your HCA ends up using) * Added btl_openib_hca_param_files as a colon-delimited list of INI files to check for values during startup (in order, left-to-right, just like the MCA base directory param). * MTU is currently the only value supported in this framework. * It is not a fatal error if we don't find params for the HCA in the INI file(s). Instead, just print a warning. New MCA param btl_openib_warn_no_hca_params_found can be used to disable printing the warning. * Add MTU to peer negotiation when making a connection * Exchange maximum MTU; select the lesser of the two This commit was SVN r11182.
2006-08-14 23:30:37 +04:00
return ret;
}
This commit brings in two major things: 1. Galen's fine-grain control of queue pair resources in the openib BTL. 1. Pasha's new implementation of asychronous HCA event handling. Pasha's new implementation doesn't take much explanation, but the new "multifrag" stuff does. Note that "svn merge" was not used to bring this new code from the /tmp/ib_multifrag branch -- something Bad happened in the periodic trunk pulls on that branch making an actual merge back to the trunk effectively impossible (i.e., lots and lots of arbitrary conflicts and artifical changes). :-( == Fine-grain control of queue pair resources == Galen's fine-grain control of queue pair resources to the OpenIB BTL (thanks to Gleb for fixing broken code and providing additional functionality, Pasha for finding broken code, and Jeff for doing all the svn work and regression testing). Prior to this commit, the OpenIB BTL created two queue pairs: one for eager size fragments and one for max send size fragments. When the use of the shared receive queue (SRQ) was specified (via "-mca btl_openib_use_srq 1"), these QPs would use a shared receive queue for receive buffers instead of the default per-peer (PP) receive queues and buffers. One consequence of this design is that receive buffer utilization (the size of the data received as a percentage of the receive buffer used for the data) was quite poor for a number of applications. The new design allows multiple QPs to be specified at runtime. Each QP can be setup to use PP or SRQ receive buffers as well as giving fine-grained control over receive buffer size, number of receive buffers to post, when to replenish the receive queue (low water mark) and for SRQ QPs, the number of outstanding sends can also be specified. The following is an example of the syntax to describe QPs to the OpenIB BTL using the new MCA parameter btl_openib_receive_queues: {{{ -mca btl_openib_receive_queues \ "P,128,16,4;S,1024,256,128,32;S,4096,256,128,32;S,65536,256,128,32" }}} Each QP description is delimited by ";" (semicolon) with individual fields of the QP description delimited by "," (comma). The above example therefore describes 4 QPs. The first QP is: P,128,16,4 Meaning: per-peer receive buffer QPs are indicated by a starting field of "P"; the first QP (shown above) is therefore a per-peer based QP. The second field indicates the size of the receive buffer in bytes (128 bytes). The third field indicates the number of receive buffers to allocate to the QP (16). The fourth field indicates the low watermark for receive buffers at which time the BTL will repost receive buffers to the QP (4). The second QP is: S,1024,256,128,32 Shared receive queue based QPs are indicated by a starting field of "S"; the second QP (shown above) is therefore a shared receive queue based QP. The second, third and fourth fields are the same as in the per-peer based QP. The fifth field is the number of outstanding sends that are allowed at a given time on the QP (32). This provides a "good enough" mechanism of flow control for some regular communication patterns. QPs MUST be specified in ascending receive buffer size order. This requirement may be removed prior to 1.3 release. This commit was SVN r15474.
2007-07-18 05:15:59 +04:00
static int32_t atoi_param(char *param, int32_t dflt)
{
if(NULL == param || '\0' == param[0])
return dflt ? dflt : 1;
return atoi(param);
}
This commit brings in two major things: 1. Galen's fine-grain control of queue pair resources in the openib BTL. 1. Pasha's new implementation of asychronous HCA event handling. Pasha's new implementation doesn't take much explanation, but the new "multifrag" stuff does. Note that "svn merge" was not used to bring this new code from the /tmp/ib_multifrag branch -- something Bad happened in the periodic trunk pulls on that branch making an actual merge back to the trunk effectively impossible (i.e., lots and lots of arbitrary conflicts and artifical changes). :-( == Fine-grain control of queue pair resources == Galen's fine-grain control of queue pair resources to the OpenIB BTL (thanks to Gleb for fixing broken code and providing additional functionality, Pasha for finding broken code, and Jeff for doing all the svn work and regression testing). Prior to this commit, the OpenIB BTL created two queue pairs: one for eager size fragments and one for max send size fragments. When the use of the shared receive queue (SRQ) was specified (via "-mca btl_openib_use_srq 1"), these QPs would use a shared receive queue for receive buffers instead of the default per-peer (PP) receive queues and buffers. One consequence of this design is that receive buffer utilization (the size of the data received as a percentage of the receive buffer used for the data) was quite poor for a number of applications. The new design allows multiple QPs to be specified at runtime. Each QP can be setup to use PP or SRQ receive buffers as well as giving fine-grained control over receive buffer size, number of receive buffers to post, when to replenish the receive queue (low water mark) and for SRQ QPs, the number of outstanding sends can also be specified. The following is an example of the syntax to describe QPs to the OpenIB BTL using the new MCA parameter btl_openib_receive_queues: {{{ -mca btl_openib_receive_queues \ "P,128,16,4;S,1024,256,128,32;S,4096,256,128,32;S,65536,256,128,32" }}} Each QP description is delimited by ";" (semicolon) with individual fields of the QP description delimited by "," (comma). The above example therefore describes 4 QPs. The first QP is: P,128,16,4 Meaning: per-peer receive buffer QPs are indicated by a starting field of "P"; the first QP (shown above) is therefore a per-peer based QP. The second field indicates the size of the receive buffer in bytes (128 bytes). The third field indicates the number of receive buffers to allocate to the QP (16). The fourth field indicates the low watermark for receive buffers at which time the BTL will repost receive buffers to the QP (4). The second QP is: S,1024,256,128,32 Shared receive queue based QPs are indicated by a starting field of "S"; the second QP (shown above) is therefore a shared receive queue based QP. The second, third and fourth fields are the same as in the per-peer based QP. The fifth field is the number of outstanding sends that are allowed at a given time on the QP (32). This provides a "good enough" mechanism of flow control for some regular communication patterns. QPs MUST be specified in ascending receive buffer size order. This requirement may be removed prior to 1.3 release. This commit was SVN r15474.
2007-07-18 05:15:59 +04:00
static int mca_btl_openib_mca_setup_qps(void)
{
This commit brings in two major things: 1. Galen's fine-grain control of queue pair resources in the openib BTL. 1. Pasha's new implementation of asychronous HCA event handling. Pasha's new implementation doesn't take much explanation, but the new "multifrag" stuff does. Note that "svn merge" was not used to bring this new code from the /tmp/ib_multifrag branch -- something Bad happened in the periodic trunk pulls on that branch making an actual merge back to the trunk effectively impossible (i.e., lots and lots of arbitrary conflicts and artifical changes). :-( == Fine-grain control of queue pair resources == Galen's fine-grain control of queue pair resources to the OpenIB BTL (thanks to Gleb for fixing broken code and providing additional functionality, Pasha for finding broken code, and Jeff for doing all the svn work and regression testing). Prior to this commit, the OpenIB BTL created two queue pairs: one for eager size fragments and one for max send size fragments. When the use of the shared receive queue (SRQ) was specified (via "-mca btl_openib_use_srq 1"), these QPs would use a shared receive queue for receive buffers instead of the default per-peer (PP) receive queues and buffers. One consequence of this design is that receive buffer utilization (the size of the data received as a percentage of the receive buffer used for the data) was quite poor for a number of applications. The new design allows multiple QPs to be specified at runtime. Each QP can be setup to use PP or SRQ receive buffers as well as giving fine-grained control over receive buffer size, number of receive buffers to post, when to replenish the receive queue (low water mark) and for SRQ QPs, the number of outstanding sends can also be specified. The following is an example of the syntax to describe QPs to the OpenIB BTL using the new MCA parameter btl_openib_receive_queues: {{{ -mca btl_openib_receive_queues \ "P,128,16,4;S,1024,256,128,32;S,4096,256,128,32;S,65536,256,128,32" }}} Each QP description is delimited by ";" (semicolon) with individual fields of the QP description delimited by "," (comma). The above example therefore describes 4 QPs. The first QP is: P,128,16,4 Meaning: per-peer receive buffer QPs are indicated by a starting field of "P"; the first QP (shown above) is therefore a per-peer based QP. The second field indicates the size of the receive buffer in bytes (128 bytes). The third field indicates the number of receive buffers to allocate to the QP (16). The fourth field indicates the low watermark for receive buffers at which time the BTL will repost receive buffers to the QP (4). The second QP is: S,1024,256,128,32 Shared receive queue based QPs are indicated by a starting field of "S"; the second QP (shown above) is therefore a shared receive queue based QP. The second, third and fourth fields are the same as in the per-peer based QP. The fifth field is the number of outstanding sends that are allowed at a given time on the QP (32). This provides a "good enough" mechanism of flow control for some regular communication patterns. QPs MUST be specified in ascending receive buffer size order. This requirement may be removed prior to 1.3 release. This commit was SVN r15474.
2007-07-18 05:15:59 +04:00
/* All the multi-qp stuff.. */
char *str;
char **queues, **params = NULL;
int num_xrc_qps = 0, num_pp_qps = 0, num_srq_qps = 0, qp = 0;
char default_qps[100];
uint32_t max_qp_size, max_size_needed;
int32_t min_freelist_size = 0;
int smallest_pp_qp = 0, ret = OMPI_ERROR, i;
uint32_t mid_qp_size;
mid_qp_size = mca_btl_openib_module.super.btl_eager_limit / 4;
/* round mid_qp_size to smallest power of two */
for(i = 31; i > 0; i--) {
if(!(mid_qp_size & (1<<i)))
continue;
mid_qp_size = (1<<i);
break;
}
if(mid_qp_size <= 128)
mid_qp_size = 1024;
snprintf(default_qps, 100,
"P,128,256,192,128:S,%u,256,128,32:S,%u,256,128,32:S,%u,256,128,32",
mid_qp_size,
(uint32_t)mca_btl_openib_module.super.btl_eager_limit,
(uint32_t)mca_btl_openib_module.super.btl_max_send_size);
This commit brings in two major things: 1. Galen's fine-grain control of queue pair resources in the openib BTL. 1. Pasha's new implementation of asychronous HCA event handling. Pasha's new implementation doesn't take much explanation, but the new "multifrag" stuff does. Note that "svn merge" was not used to bring this new code from the /tmp/ib_multifrag branch -- something Bad happened in the periodic trunk pulls on that branch making an actual merge back to the trunk effectively impossible (i.e., lots and lots of arbitrary conflicts and artifical changes). :-( == Fine-grain control of queue pair resources == Galen's fine-grain control of queue pair resources to the OpenIB BTL (thanks to Gleb for fixing broken code and providing additional functionality, Pasha for finding broken code, and Jeff for doing all the svn work and regression testing). Prior to this commit, the OpenIB BTL created two queue pairs: one for eager size fragments and one for max send size fragments. When the use of the shared receive queue (SRQ) was specified (via "-mca btl_openib_use_srq 1"), these QPs would use a shared receive queue for receive buffers instead of the default per-peer (PP) receive queues and buffers. One consequence of this design is that receive buffer utilization (the size of the data received as a percentage of the receive buffer used for the data) was quite poor for a number of applications. The new design allows multiple QPs to be specified at runtime. Each QP can be setup to use PP or SRQ receive buffers as well as giving fine-grained control over receive buffer size, number of receive buffers to post, when to replenish the receive queue (low water mark) and for SRQ QPs, the number of outstanding sends can also be specified. The following is an example of the syntax to describe QPs to the OpenIB BTL using the new MCA parameter btl_openib_receive_queues: {{{ -mca btl_openib_receive_queues \ "P,128,16,4;S,1024,256,128,32;S,4096,256,128,32;S,65536,256,128,32" }}} Each QP description is delimited by ";" (semicolon) with individual fields of the QP description delimited by "," (comma). The above example therefore describes 4 QPs. The first QP is: P,128,16,4 Meaning: per-peer receive buffer QPs are indicated by a starting field of "P"; the first QP (shown above) is therefore a per-peer based QP. The second field indicates the size of the receive buffer in bytes (128 bytes). The third field indicates the number of receive buffers to allocate to the QP (16). The fourth field indicates the low watermark for receive buffers at which time the BTL will repost receive buffers to the QP (4). The second QP is: S,1024,256,128,32 Shared receive queue based QPs are indicated by a starting field of "S"; the second QP (shown above) is therefore a shared receive queue based QP. The second, third and fourth fields are the same as in the per-peer based QP. The fifth field is the number of outstanding sends that are allowed at a given time on the QP (32). This provides a "good enough" mechanism of flow control for some regular communication patterns. QPs MUST be specified in ascending receive buffer size order. This requirement may be removed prior to 1.3 release. This commit was SVN r15474.
2007-07-18 05:15:59 +04:00
reg_string("receive_queues",
"Colon-delimited, coma delimited list of receive queues: P,4096,8,6,4:P,32768,8,6,4",
This commit brings in two major things: 1. Galen's fine-grain control of queue pair resources in the openib BTL. 1. Pasha's new implementation of asychronous HCA event handling. Pasha's new implementation doesn't take much explanation, but the new "multifrag" stuff does. Note that "svn merge" was not used to bring this new code from the /tmp/ib_multifrag branch -- something Bad happened in the periodic trunk pulls on that branch making an actual merge back to the trunk effectively impossible (i.e., lots and lots of arbitrary conflicts and artifical changes). :-( == Fine-grain control of queue pair resources == Galen's fine-grain control of queue pair resources to the OpenIB BTL (thanks to Gleb for fixing broken code and providing additional functionality, Pasha for finding broken code, and Jeff for doing all the svn work and regression testing). Prior to this commit, the OpenIB BTL created two queue pairs: one for eager size fragments and one for max send size fragments. When the use of the shared receive queue (SRQ) was specified (via "-mca btl_openib_use_srq 1"), these QPs would use a shared receive queue for receive buffers instead of the default per-peer (PP) receive queues and buffers. One consequence of this design is that receive buffer utilization (the size of the data received as a percentage of the receive buffer used for the data) was quite poor for a number of applications. The new design allows multiple QPs to be specified at runtime. Each QP can be setup to use PP or SRQ receive buffers as well as giving fine-grained control over receive buffer size, number of receive buffers to post, when to replenish the receive queue (low water mark) and for SRQ QPs, the number of outstanding sends can also be specified. The following is an example of the syntax to describe QPs to the OpenIB BTL using the new MCA parameter btl_openib_receive_queues: {{{ -mca btl_openib_receive_queues \ "P,128,16,4;S,1024,256,128,32;S,4096,256,128,32;S,65536,256,128,32" }}} Each QP description is delimited by ";" (semicolon) with individual fields of the QP description delimited by "," (comma). The above example therefore describes 4 QPs. The first QP is: P,128,16,4 Meaning: per-peer receive buffer QPs are indicated by a starting field of "P"; the first QP (shown above) is therefore a per-peer based QP. The second field indicates the size of the receive buffer in bytes (128 bytes). The third field indicates the number of receive buffers to allocate to the QP (16). The fourth field indicates the low watermark for receive buffers at which time the BTL will repost receive buffers to the QP (4). The second QP is: S,1024,256,128,32 Shared receive queue based QPs are indicated by a starting field of "S"; the second QP (shown above) is therefore a shared receive queue based QP. The second, third and fourth fields are the same as in the per-peer based QP. The fifth field is the number of outstanding sends that are allowed at a given time on the QP (32). This provides a "good enough" mechanism of flow control for some regular communication patterns. QPs MUST be specified in ascending receive buffer size order. This requirement may be removed prior to 1.3 release. This commit was SVN r15474.
2007-07-18 05:15:59 +04:00
default_qps, &str, 0);
queues = opal_argv_split(str, ':');
This commit brings in two major things: 1. Galen's fine-grain control of queue pair resources in the openib BTL. 1. Pasha's new implementation of asychronous HCA event handling. Pasha's new implementation doesn't take much explanation, but the new "multifrag" stuff does. Note that "svn merge" was not used to bring this new code from the /tmp/ib_multifrag branch -- something Bad happened in the periodic trunk pulls on that branch making an actual merge back to the trunk effectively impossible (i.e., lots and lots of arbitrary conflicts and artifical changes). :-( == Fine-grain control of queue pair resources == Galen's fine-grain control of queue pair resources to the OpenIB BTL (thanks to Gleb for fixing broken code and providing additional functionality, Pasha for finding broken code, and Jeff for doing all the svn work and regression testing). Prior to this commit, the OpenIB BTL created two queue pairs: one for eager size fragments and one for max send size fragments. When the use of the shared receive queue (SRQ) was specified (via "-mca btl_openib_use_srq 1"), these QPs would use a shared receive queue for receive buffers instead of the default per-peer (PP) receive queues and buffers. One consequence of this design is that receive buffer utilization (the size of the data received as a percentage of the receive buffer used for the data) was quite poor for a number of applications. The new design allows multiple QPs to be specified at runtime. Each QP can be setup to use PP or SRQ receive buffers as well as giving fine-grained control over receive buffer size, number of receive buffers to post, when to replenish the receive queue (low water mark) and for SRQ QPs, the number of outstanding sends can also be specified. The following is an example of the syntax to describe QPs to the OpenIB BTL using the new MCA parameter btl_openib_receive_queues: {{{ -mca btl_openib_receive_queues \ "P,128,16,4;S,1024,256,128,32;S,4096,256,128,32;S,65536,256,128,32" }}} Each QP description is delimited by ";" (semicolon) with individual fields of the QP description delimited by "," (comma). The above example therefore describes 4 QPs. The first QP is: P,128,16,4 Meaning: per-peer receive buffer QPs are indicated by a starting field of "P"; the first QP (shown above) is therefore a per-peer based QP. The second field indicates the size of the receive buffer in bytes (128 bytes). The third field indicates the number of receive buffers to allocate to the QP (16). The fourth field indicates the low watermark for receive buffers at which time the BTL will repost receive buffers to the QP (4). The second QP is: S,1024,256,128,32 Shared receive queue based QPs are indicated by a starting field of "S"; the second QP (shown above) is therefore a shared receive queue based QP. The second, third and fourth fields are the same as in the per-peer based QP. The fifth field is the number of outstanding sends that are allowed at a given time on the QP (32). This provides a "good enough" mechanism of flow control for some regular communication patterns. QPs MUST be specified in ascending receive buffer size order. This requirement may be removed prior to 1.3 release. This commit was SVN r15474.
2007-07-18 05:15:59 +04:00
if (0 == opal_argv_count(queues)) {
opal_show_help("help-mpi-btl-openib.txt",
"no qps in receive_queues", true,
orte_system_info.nodename, str);
This commit brings in two major things: 1. Galen's fine-grain control of queue pair resources in the openib BTL. 1. Pasha's new implementation of asychronous HCA event handling. Pasha's new implementation doesn't take much explanation, but the new "multifrag" stuff does. Note that "svn merge" was not used to bring this new code from the /tmp/ib_multifrag branch -- something Bad happened in the periodic trunk pulls on that branch making an actual merge back to the trunk effectively impossible (i.e., lots and lots of arbitrary conflicts and artifical changes). :-( == Fine-grain control of queue pair resources == Galen's fine-grain control of queue pair resources to the OpenIB BTL (thanks to Gleb for fixing broken code and providing additional functionality, Pasha for finding broken code, and Jeff for doing all the svn work and regression testing). Prior to this commit, the OpenIB BTL created two queue pairs: one for eager size fragments and one for max send size fragments. When the use of the shared receive queue (SRQ) was specified (via "-mca btl_openib_use_srq 1"), these QPs would use a shared receive queue for receive buffers instead of the default per-peer (PP) receive queues and buffers. One consequence of this design is that receive buffer utilization (the size of the data received as a percentage of the receive buffer used for the data) was quite poor for a number of applications. The new design allows multiple QPs to be specified at runtime. Each QP can be setup to use PP or SRQ receive buffers as well as giving fine-grained control over receive buffer size, number of receive buffers to post, when to replenish the receive queue (low water mark) and for SRQ QPs, the number of outstanding sends can also be specified. The following is an example of the syntax to describe QPs to the OpenIB BTL using the new MCA parameter btl_openib_receive_queues: {{{ -mca btl_openib_receive_queues \ "P,128,16,4;S,1024,256,128,32;S,4096,256,128,32;S,65536,256,128,32" }}} Each QP description is delimited by ";" (semicolon) with individual fields of the QP description delimited by "," (comma). The above example therefore describes 4 QPs. The first QP is: P,128,16,4 Meaning: per-peer receive buffer QPs are indicated by a starting field of "P"; the first QP (shown above) is therefore a per-peer based QP. The second field indicates the size of the receive buffer in bytes (128 bytes). The third field indicates the number of receive buffers to allocate to the QP (16). The fourth field indicates the low watermark for receive buffers at which time the BTL will repost receive buffers to the QP (4). The second QP is: S,1024,256,128,32 Shared receive queue based QPs are indicated by a starting field of "S"; the second QP (shown above) is therefore a shared receive queue based QP. The second, third and fourth fields are the same as in the per-peer based QP. The fifth field is the number of outstanding sends that are allowed at a given time on the QP (32). This provides a "good enough" mechanism of flow control for some regular communication patterns. QPs MUST be specified in ascending receive buffer size order. This requirement may be removed prior to 1.3 release. This commit was SVN r15474.
2007-07-18 05:15:59 +04:00
return OMPI_ERROR;
}
while (queues[qp] != NULL) {
if (0 == strncmp("P,", queues[qp], 2)) {
num_pp_qps++;
if(smallest_pp_qp > qp)
smallest_pp_qp = qp;
} else if (0 == strncmp("S,", queues[qp], 2)) {
This commit brings in two major things: 1. Galen's fine-grain control of queue pair resources in the openib BTL. 1. Pasha's new implementation of asychronous HCA event handling. Pasha's new implementation doesn't take much explanation, but the new "multifrag" stuff does. Note that "svn merge" was not used to bring this new code from the /tmp/ib_multifrag branch -- something Bad happened in the periodic trunk pulls on that branch making an actual merge back to the trunk effectively impossible (i.e., lots and lots of arbitrary conflicts and artifical changes). :-( == Fine-grain control of queue pair resources == Galen's fine-grain control of queue pair resources to the OpenIB BTL (thanks to Gleb for fixing broken code and providing additional functionality, Pasha for finding broken code, and Jeff for doing all the svn work and regression testing). Prior to this commit, the OpenIB BTL created two queue pairs: one for eager size fragments and one for max send size fragments. When the use of the shared receive queue (SRQ) was specified (via "-mca btl_openib_use_srq 1"), these QPs would use a shared receive queue for receive buffers instead of the default per-peer (PP) receive queues and buffers. One consequence of this design is that receive buffer utilization (the size of the data received as a percentage of the receive buffer used for the data) was quite poor for a number of applications. The new design allows multiple QPs to be specified at runtime. Each QP can be setup to use PP or SRQ receive buffers as well as giving fine-grained control over receive buffer size, number of receive buffers to post, when to replenish the receive queue (low water mark) and for SRQ QPs, the number of outstanding sends can also be specified. The following is an example of the syntax to describe QPs to the OpenIB BTL using the new MCA parameter btl_openib_receive_queues: {{{ -mca btl_openib_receive_queues \ "P,128,16,4;S,1024,256,128,32;S,4096,256,128,32;S,65536,256,128,32" }}} Each QP description is delimited by ";" (semicolon) with individual fields of the QP description delimited by "," (comma). The above example therefore describes 4 QPs. The first QP is: P,128,16,4 Meaning: per-peer receive buffer QPs are indicated by a starting field of "P"; the first QP (shown above) is therefore a per-peer based QP. The second field indicates the size of the receive buffer in bytes (128 bytes). The third field indicates the number of receive buffers to allocate to the QP (16). The fourth field indicates the low watermark for receive buffers at which time the BTL will repost receive buffers to the QP (4). The second QP is: S,1024,256,128,32 Shared receive queue based QPs are indicated by a starting field of "S"; the second QP (shown above) is therefore a shared receive queue based QP. The second, third and fourth fields are the same as in the per-peer based QP. The fifth field is the number of outstanding sends that are allowed at a given time on the QP (32). This provides a "good enough" mechanism of flow control for some regular communication patterns. QPs MUST be specified in ascending receive buffer size order. This requirement may be removed prior to 1.3 release. This commit was SVN r15474.
2007-07-18 05:15:59 +04:00
num_srq_qps++;
} else if (0 == strncmp("X,", queues[qp], 2)) {
#if HAVE_XRC
num_xrc_qps++;
#else
opal_show_help("help-mpi-btl-openib.txt", "No XRC support", true,
orte_system_info.nodename, str);
goto error;
#endif
This commit brings in two major things: 1. Galen's fine-grain control of queue pair resources in the openib BTL. 1. Pasha's new implementation of asychronous HCA event handling. Pasha's new implementation doesn't take much explanation, but the new "multifrag" stuff does. Note that "svn merge" was not used to bring this new code from the /tmp/ib_multifrag branch -- something Bad happened in the periodic trunk pulls on that branch making an actual merge back to the trunk effectively impossible (i.e., lots and lots of arbitrary conflicts and artifical changes). :-( == Fine-grain control of queue pair resources == Galen's fine-grain control of queue pair resources to the OpenIB BTL (thanks to Gleb for fixing broken code and providing additional functionality, Pasha for finding broken code, and Jeff for doing all the svn work and regression testing). Prior to this commit, the OpenIB BTL created two queue pairs: one for eager size fragments and one for max send size fragments. When the use of the shared receive queue (SRQ) was specified (via "-mca btl_openib_use_srq 1"), these QPs would use a shared receive queue for receive buffers instead of the default per-peer (PP) receive queues and buffers. One consequence of this design is that receive buffer utilization (the size of the data received as a percentage of the receive buffer used for the data) was quite poor for a number of applications. The new design allows multiple QPs to be specified at runtime. Each QP can be setup to use PP or SRQ receive buffers as well as giving fine-grained control over receive buffer size, number of receive buffers to post, when to replenish the receive queue (low water mark) and for SRQ QPs, the number of outstanding sends can also be specified. The following is an example of the syntax to describe QPs to the OpenIB BTL using the new MCA parameter btl_openib_receive_queues: {{{ -mca btl_openib_receive_queues \ "P,128,16,4;S,1024,256,128,32;S,4096,256,128,32;S,65536,256,128,32" }}} Each QP description is delimited by ";" (semicolon) with individual fields of the QP description delimited by "," (comma). The above example therefore describes 4 QPs. The first QP is: P,128,16,4 Meaning: per-peer receive buffer QPs are indicated by a starting field of "P"; the first QP (shown above) is therefore a per-peer based QP. The second field indicates the size of the receive buffer in bytes (128 bytes). The third field indicates the number of receive buffers to allocate to the QP (16). The fourth field indicates the low watermark for receive buffers at which time the BTL will repost receive buffers to the QP (4). The second QP is: S,1024,256,128,32 Shared receive queue based QPs are indicated by a starting field of "S"; the second QP (shown above) is therefore a shared receive queue based QP. The second, third and fourth fields are the same as in the per-peer based QP. The fifth field is the number of outstanding sends that are allowed at a given time on the QP (32). This provides a "good enough" mechanism of flow control for some regular communication patterns. QPs MUST be specified in ascending receive buffer size order. This requirement may be removed prior to 1.3 release. This commit was SVN r15474.
2007-07-18 05:15:59 +04:00
} else {
opal_show_help("help-mpi-btl-openib.txt",
"invalid qp type in receive_queues", true,
orte_system_info.nodename, str, queues[qp]);
This commit brings in two major things: 1. Galen's fine-grain control of queue pair resources in the openib BTL. 1. Pasha's new implementation of asychronous HCA event handling. Pasha's new implementation doesn't take much explanation, but the new "multifrag" stuff does. Note that "svn merge" was not used to bring this new code from the /tmp/ib_multifrag branch -- something Bad happened in the periodic trunk pulls on that branch making an actual merge back to the trunk effectively impossible (i.e., lots and lots of arbitrary conflicts and artifical changes). :-( == Fine-grain control of queue pair resources == Galen's fine-grain control of queue pair resources to the OpenIB BTL (thanks to Gleb for fixing broken code and providing additional functionality, Pasha for finding broken code, and Jeff for doing all the svn work and regression testing). Prior to this commit, the OpenIB BTL created two queue pairs: one for eager size fragments and one for max send size fragments. When the use of the shared receive queue (SRQ) was specified (via "-mca btl_openib_use_srq 1"), these QPs would use a shared receive queue for receive buffers instead of the default per-peer (PP) receive queues and buffers. One consequence of this design is that receive buffer utilization (the size of the data received as a percentage of the receive buffer used for the data) was quite poor for a number of applications. The new design allows multiple QPs to be specified at runtime. Each QP can be setup to use PP or SRQ receive buffers as well as giving fine-grained control over receive buffer size, number of receive buffers to post, when to replenish the receive queue (low water mark) and for SRQ QPs, the number of outstanding sends can also be specified. The following is an example of the syntax to describe QPs to the OpenIB BTL using the new MCA parameter btl_openib_receive_queues: {{{ -mca btl_openib_receive_queues \ "P,128,16,4;S,1024,256,128,32;S,4096,256,128,32;S,65536,256,128,32" }}} Each QP description is delimited by ";" (semicolon) with individual fields of the QP description delimited by "," (comma). The above example therefore describes 4 QPs. The first QP is: P,128,16,4 Meaning: per-peer receive buffer QPs are indicated by a starting field of "P"; the first QP (shown above) is therefore a per-peer based QP. The second field indicates the size of the receive buffer in bytes (128 bytes). The third field indicates the number of receive buffers to allocate to the QP (16). The fourth field indicates the low watermark for receive buffers at which time the BTL will repost receive buffers to the QP (4). The second QP is: S,1024,256,128,32 Shared receive queue based QPs are indicated by a starting field of "S"; the second QP (shown above) is therefore a shared receive queue based QP. The second, third and fourth fields are the same as in the per-peer based QP. The fifth field is the number of outstanding sends that are allowed at a given time on the QP (32). This provides a "good enough" mechanism of flow control for some regular communication patterns. QPs MUST be specified in ascending receive buffer size order. This requirement may be removed prior to 1.3 release. This commit was SVN r15474.
2007-07-18 05:15:59 +04:00
goto error;
}
qp++;
}
/* Current XRC implementation can't used with other QP types - PP and SRQ */
if (num_xrc_qps > 0 && (num_pp_qps > 0 || num_srq_qps > 0)) {
opal_show_help("help-mpi-btl-openib.txt", "XRC with PP or SRQ", true,
orte_system_info.nodename, str);
goto error;
}
/* Current XRC implementation can't used with btls_per_lid > 1 */
if (num_xrc_qps > 0 && mca_btl_openib_component.btls_per_lid > 1) {
opal_show_help("help-mpi-btl-openib.txt", "XRC with BTLs per LID", true,
orte_system_info.nodename, str, num_xrc_qps);
goto error;
}
This commit brings in two major things: 1. Galen's fine-grain control of queue pair resources in the openib BTL. 1. Pasha's new implementation of asychronous HCA event handling. Pasha's new implementation doesn't take much explanation, but the new "multifrag" stuff does. Note that "svn merge" was not used to bring this new code from the /tmp/ib_multifrag branch -- something Bad happened in the periodic trunk pulls on that branch making an actual merge back to the trunk effectively impossible (i.e., lots and lots of arbitrary conflicts and artifical changes). :-( == Fine-grain control of queue pair resources == Galen's fine-grain control of queue pair resources to the OpenIB BTL (thanks to Gleb for fixing broken code and providing additional functionality, Pasha for finding broken code, and Jeff for doing all the svn work and regression testing). Prior to this commit, the OpenIB BTL created two queue pairs: one for eager size fragments and one for max send size fragments. When the use of the shared receive queue (SRQ) was specified (via "-mca btl_openib_use_srq 1"), these QPs would use a shared receive queue for receive buffers instead of the default per-peer (PP) receive queues and buffers. One consequence of this design is that receive buffer utilization (the size of the data received as a percentage of the receive buffer used for the data) was quite poor for a number of applications. The new design allows multiple QPs to be specified at runtime. Each QP can be setup to use PP or SRQ receive buffers as well as giving fine-grained control over receive buffer size, number of receive buffers to post, when to replenish the receive queue (low water mark) and for SRQ QPs, the number of outstanding sends can also be specified. The following is an example of the syntax to describe QPs to the OpenIB BTL using the new MCA parameter btl_openib_receive_queues: {{{ -mca btl_openib_receive_queues \ "P,128,16,4;S,1024,256,128,32;S,4096,256,128,32;S,65536,256,128,32" }}} Each QP description is delimited by ";" (semicolon) with individual fields of the QP description delimited by "," (comma). The above example therefore describes 4 QPs. The first QP is: P,128,16,4 Meaning: per-peer receive buffer QPs are indicated by a starting field of "P"; the first QP (shown above) is therefore a per-peer based QP. The second field indicates the size of the receive buffer in bytes (128 bytes). The third field indicates the number of receive buffers to allocate to the QP (16). The fourth field indicates the low watermark for receive buffers at which time the BTL will repost receive buffers to the QP (4). The second QP is: S,1024,256,128,32 Shared receive queue based QPs are indicated by a starting field of "S"; the second QP (shown above) is therefore a shared receive queue based QP. The second, third and fourth fields are the same as in the per-peer based QP. The fifth field is the number of outstanding sends that are allowed at a given time on the QP (32). This provides a "good enough" mechanism of flow control for some regular communication patterns. QPs MUST be specified in ascending receive buffer size order. This requirement may be removed prior to 1.3 release. This commit was SVN r15474.
2007-07-18 05:15:59 +04:00
mca_btl_openib_component.num_pp_qps = num_pp_qps;
mca_btl_openib_component.num_srq_qps = num_srq_qps;
mca_btl_openib_component.num_xrc_qps = num_xrc_qps;
mca_btl_openib_component.num_qps = num_pp_qps + num_srq_qps + num_xrc_qps;
This commit brings in two major things: 1. Galen's fine-grain control of queue pair resources in the openib BTL. 1. Pasha's new implementation of asychronous HCA event handling. Pasha's new implementation doesn't take much explanation, but the new "multifrag" stuff does. Note that "svn merge" was not used to bring this new code from the /tmp/ib_multifrag branch -- something Bad happened in the periodic trunk pulls on that branch making an actual merge back to the trunk effectively impossible (i.e., lots and lots of arbitrary conflicts and artifical changes). :-( == Fine-grain control of queue pair resources == Galen's fine-grain control of queue pair resources to the OpenIB BTL (thanks to Gleb for fixing broken code and providing additional functionality, Pasha for finding broken code, and Jeff for doing all the svn work and regression testing). Prior to this commit, the OpenIB BTL created two queue pairs: one for eager size fragments and one for max send size fragments. When the use of the shared receive queue (SRQ) was specified (via "-mca btl_openib_use_srq 1"), these QPs would use a shared receive queue for receive buffers instead of the default per-peer (PP) receive queues and buffers. One consequence of this design is that receive buffer utilization (the size of the data received as a percentage of the receive buffer used for the data) was quite poor for a number of applications. The new design allows multiple QPs to be specified at runtime. Each QP can be setup to use PP or SRQ receive buffers as well as giving fine-grained control over receive buffer size, number of receive buffers to post, when to replenish the receive queue (low water mark) and for SRQ QPs, the number of outstanding sends can also be specified. The following is an example of the syntax to describe QPs to the OpenIB BTL using the new MCA parameter btl_openib_receive_queues: {{{ -mca btl_openib_receive_queues \ "P,128,16,4;S,1024,256,128,32;S,4096,256,128,32;S,65536,256,128,32" }}} Each QP description is delimited by ";" (semicolon) with individual fields of the QP description delimited by "," (comma). The above example therefore describes 4 QPs. The first QP is: P,128,16,4 Meaning: per-peer receive buffer QPs are indicated by a starting field of "P"; the first QP (shown above) is therefore a per-peer based QP. The second field indicates the size of the receive buffer in bytes (128 bytes). The third field indicates the number of receive buffers to allocate to the QP (16). The fourth field indicates the low watermark for receive buffers at which time the BTL will repost receive buffers to the QP (4). The second QP is: S,1024,256,128,32 Shared receive queue based QPs are indicated by a starting field of "S"; the second QP (shown above) is therefore a shared receive queue based QP. The second, third and fourth fields are the same as in the per-peer based QP. The fifth field is the number of outstanding sends that are allowed at a given time on the QP (32). This provides a "good enough" mechanism of flow control for some regular communication patterns. QPs MUST be specified in ascending receive buffer size order. This requirement may be removed prior to 1.3 release. This commit was SVN r15474.
2007-07-18 05:15:59 +04:00
mca_btl_openib_component.qp_infos = (mca_btl_openib_qp_info_t*)
malloc(sizeof(mca_btl_openib_qp_info_t) *
mca_btl_openib_component.num_qps);
qp = 0;
#define P(N) (((N) > count)?NULL:params[(N)])
This commit brings in two major things: 1. Galen's fine-grain control of queue pair resources in the openib BTL. 1. Pasha's new implementation of asychronous HCA event handling. Pasha's new implementation doesn't take much explanation, but the new "multifrag" stuff does. Note that "svn merge" was not used to bring this new code from the /tmp/ib_multifrag branch -- something Bad happened in the periodic trunk pulls on that branch making an actual merge back to the trunk effectively impossible (i.e., lots and lots of arbitrary conflicts and artifical changes). :-( == Fine-grain control of queue pair resources == Galen's fine-grain control of queue pair resources to the OpenIB BTL (thanks to Gleb for fixing broken code and providing additional functionality, Pasha for finding broken code, and Jeff for doing all the svn work and regression testing). Prior to this commit, the OpenIB BTL created two queue pairs: one for eager size fragments and one for max send size fragments. When the use of the shared receive queue (SRQ) was specified (via "-mca btl_openib_use_srq 1"), these QPs would use a shared receive queue for receive buffers instead of the default per-peer (PP) receive queues and buffers. One consequence of this design is that receive buffer utilization (the size of the data received as a percentage of the receive buffer used for the data) was quite poor for a number of applications. The new design allows multiple QPs to be specified at runtime. Each QP can be setup to use PP or SRQ receive buffers as well as giving fine-grained control over receive buffer size, number of receive buffers to post, when to replenish the receive queue (low water mark) and for SRQ QPs, the number of outstanding sends can also be specified. The following is an example of the syntax to describe QPs to the OpenIB BTL using the new MCA parameter btl_openib_receive_queues: {{{ -mca btl_openib_receive_queues \ "P,128,16,4;S,1024,256,128,32;S,4096,256,128,32;S,65536,256,128,32" }}} Each QP description is delimited by ";" (semicolon) with individual fields of the QP description delimited by "," (comma). The above example therefore describes 4 QPs. The first QP is: P,128,16,4 Meaning: per-peer receive buffer QPs are indicated by a starting field of "P"; the first QP (shown above) is therefore a per-peer based QP. The second field indicates the size of the receive buffer in bytes (128 bytes). The third field indicates the number of receive buffers to allocate to the QP (16). The fourth field indicates the low watermark for receive buffers at which time the BTL will repost receive buffers to the QP (4). The second QP is: S,1024,256,128,32 Shared receive queue based QPs are indicated by a starting field of "S"; the second QP (shown above) is therefore a shared receive queue based QP. The second, third and fourth fields are the same as in the per-peer based QP. The fifth field is the number of outstanding sends that are allowed at a given time on the QP (32). This provides a "good enough" mechanism of flow control for some regular communication patterns. QPs MUST be specified in ascending receive buffer size order. This requirement may be removed prior to 1.3 release. This commit was SVN r15474.
2007-07-18 05:15:59 +04:00
while(queues[qp] != NULL) {
int i = 0, count;
int32_t rd_low, rd_num;
params = opal_argv_split_with_empty(queues[qp], ',');
count = opal_argv_count(params);
This commit brings in two major things: 1. Galen's fine-grain control of queue pair resources in the openib BTL. 1. Pasha's new implementation of asychronous HCA event handling. Pasha's new implementation doesn't take much explanation, but the new "multifrag" stuff does. Note that "svn merge" was not used to bring this new code from the /tmp/ib_multifrag branch -- something Bad happened in the periodic trunk pulls on that branch making an actual merge back to the trunk effectively impossible (i.e., lots and lots of arbitrary conflicts and artifical changes). :-( == Fine-grain control of queue pair resources == Galen's fine-grain control of queue pair resources to the OpenIB BTL (thanks to Gleb for fixing broken code and providing additional functionality, Pasha for finding broken code, and Jeff for doing all the svn work and regression testing). Prior to this commit, the OpenIB BTL created two queue pairs: one for eager size fragments and one for max send size fragments. When the use of the shared receive queue (SRQ) was specified (via "-mca btl_openib_use_srq 1"), these QPs would use a shared receive queue for receive buffers instead of the default per-peer (PP) receive queues and buffers. One consequence of this design is that receive buffer utilization (the size of the data received as a percentage of the receive buffer used for the data) was quite poor for a number of applications. The new design allows multiple QPs to be specified at runtime. Each QP can be setup to use PP or SRQ receive buffers as well as giving fine-grained control over receive buffer size, number of receive buffers to post, when to replenish the receive queue (low water mark) and for SRQ QPs, the number of outstanding sends can also be specified. The following is an example of the syntax to describe QPs to the OpenIB BTL using the new MCA parameter btl_openib_receive_queues: {{{ -mca btl_openib_receive_queues \ "P,128,16,4;S,1024,256,128,32;S,4096,256,128,32;S,65536,256,128,32" }}} Each QP description is delimited by ";" (semicolon) with individual fields of the QP description delimited by "," (comma). The above example therefore describes 4 QPs. The first QP is: P,128,16,4 Meaning: per-peer receive buffer QPs are indicated by a starting field of "P"; the first QP (shown above) is therefore a per-peer based QP. The second field indicates the size of the receive buffer in bytes (128 bytes). The third field indicates the number of receive buffers to allocate to the QP (16). The fourth field indicates the low watermark for receive buffers at which time the BTL will repost receive buffers to the QP (4). The second QP is: S,1024,256,128,32 Shared receive queue based QPs are indicated by a starting field of "S"; the second QP (shown above) is therefore a shared receive queue based QP. The second, third and fourth fields are the same as in the per-peer based QP. The fifth field is the number of outstanding sends that are allowed at a given time on the QP (32). This provides a "good enough" mechanism of flow control for some regular communication patterns. QPs MUST be specified in ascending receive buffer size order. This requirement may be removed prior to 1.3 release. This commit was SVN r15474.
2007-07-18 05:15:59 +04:00
if ('P' == params[0][0]) {
int32_t rd_win, rd_rsv;
if (count < 3 || count > 6) {
opal_show_help("help-mpi-btl-openib.txt",
"invalid pp qp specification", true,
orte_system_info.nodename, queues[qp]);
This commit brings in two major things: 1. Galen's fine-grain control of queue pair resources in the openib BTL. 1. Pasha's new implementation of asychronous HCA event handling. Pasha's new implementation doesn't take much explanation, but the new "multifrag" stuff does. Note that "svn merge" was not used to bring this new code from the /tmp/ib_multifrag branch -- something Bad happened in the periodic trunk pulls on that branch making an actual merge back to the trunk effectively impossible (i.e., lots and lots of arbitrary conflicts and artifical changes). :-( == Fine-grain control of queue pair resources == Galen's fine-grain control of queue pair resources to the OpenIB BTL (thanks to Gleb for fixing broken code and providing additional functionality, Pasha for finding broken code, and Jeff for doing all the svn work and regression testing). Prior to this commit, the OpenIB BTL created two queue pairs: one for eager size fragments and one for max send size fragments. When the use of the shared receive queue (SRQ) was specified (via "-mca btl_openib_use_srq 1"), these QPs would use a shared receive queue for receive buffers instead of the default per-peer (PP) receive queues and buffers. One consequence of this design is that receive buffer utilization (the size of the data received as a percentage of the receive buffer used for the data) was quite poor for a number of applications. The new design allows multiple QPs to be specified at runtime. Each QP can be setup to use PP or SRQ receive buffers as well as giving fine-grained control over receive buffer size, number of receive buffers to post, when to replenish the receive queue (low water mark) and for SRQ QPs, the number of outstanding sends can also be specified. The following is an example of the syntax to describe QPs to the OpenIB BTL using the new MCA parameter btl_openib_receive_queues: {{{ -mca btl_openib_receive_queues \ "P,128,16,4;S,1024,256,128,32;S,4096,256,128,32;S,65536,256,128,32" }}} Each QP description is delimited by ";" (semicolon) with individual fields of the QP description delimited by "," (comma). The above example therefore describes 4 QPs. The first QP is: P,128,16,4 Meaning: per-peer receive buffer QPs are indicated by a starting field of "P"; the first QP (shown above) is therefore a per-peer based QP. The second field indicates the size of the receive buffer in bytes (128 bytes). The third field indicates the number of receive buffers to allocate to the QP (16). The fourth field indicates the low watermark for receive buffers at which time the BTL will repost receive buffers to the QP (4). The second QP is: S,1024,256,128,32 Shared receive queue based QPs are indicated by a starting field of "S"; the second QP (shown above) is therefore a shared receive queue based QP. The second, third and fourth fields are the same as in the per-peer based QP. The fifth field is the number of outstanding sends that are allowed at a given time on the QP (32). This provides a "good enough" mechanism of flow control for some regular communication patterns. QPs MUST be specified in ascending receive buffer size order. This requirement may be removed prior to 1.3 release. This commit was SVN r15474.
2007-07-18 05:15:59 +04:00
goto error;
}
mca_btl_openib_component.qp_infos[qp].type = MCA_BTL_OPENIB_PP_QP;
mca_btl_openib_component.qp_infos[qp].size = atoi_param(P(1), 0);
rd_num = atoi_param(P(2), 256);
/* by default set rd_low to be 3/4 of rd_num */
rd_low = atoi_param(P(3), rd_num - (rd_num / 4));
rd_win = atoi_param(P(4), (rd_num - rd_low) * 2);
rd_rsv = atoi_param(P(5), (rd_num * 2) / rd_win);
BTL_VERBOSE(("pp: rd_num is %d rd_low is %d rd_win %d rd_rsv %d\n",
rd_num, rd_low, rd_win, rd_rsv));
This commit brings in two major things: 1. Galen's fine-grain control of queue pair resources in the openib BTL. 1. Pasha's new implementation of asychronous HCA event handling. Pasha's new implementation doesn't take much explanation, but the new "multifrag" stuff does. Note that "svn merge" was not used to bring this new code from the /tmp/ib_multifrag branch -- something Bad happened in the periodic trunk pulls on that branch making an actual merge back to the trunk effectively impossible (i.e., lots and lots of arbitrary conflicts and artifical changes). :-( == Fine-grain control of queue pair resources == Galen's fine-grain control of queue pair resources to the OpenIB BTL (thanks to Gleb for fixing broken code and providing additional functionality, Pasha for finding broken code, and Jeff for doing all the svn work and regression testing). Prior to this commit, the OpenIB BTL created two queue pairs: one for eager size fragments and one for max send size fragments. When the use of the shared receive queue (SRQ) was specified (via "-mca btl_openib_use_srq 1"), these QPs would use a shared receive queue for receive buffers instead of the default per-peer (PP) receive queues and buffers. One consequence of this design is that receive buffer utilization (the size of the data received as a percentage of the receive buffer used for the data) was quite poor for a number of applications. The new design allows multiple QPs to be specified at runtime. Each QP can be setup to use PP or SRQ receive buffers as well as giving fine-grained control over receive buffer size, number of receive buffers to post, when to replenish the receive queue (low water mark) and for SRQ QPs, the number of outstanding sends can also be specified. The following is an example of the syntax to describe QPs to the OpenIB BTL using the new MCA parameter btl_openib_receive_queues: {{{ -mca btl_openib_receive_queues \ "P,128,16,4;S,1024,256,128,32;S,4096,256,128,32;S,65536,256,128,32" }}} Each QP description is delimited by ";" (semicolon) with individual fields of the QP description delimited by "," (comma). The above example therefore describes 4 QPs. The first QP is: P,128,16,4 Meaning: per-peer receive buffer QPs are indicated by a starting field of "P"; the first QP (shown above) is therefore a per-peer based QP. The second field indicates the size of the receive buffer in bytes (128 bytes). The third field indicates the number of receive buffers to allocate to the QP (16). The fourth field indicates the low watermark for receive buffers at which time the BTL will repost receive buffers to the QP (4). The second QP is: S,1024,256,128,32 Shared receive queue based QPs are indicated by a starting field of "S"; the second QP (shown above) is therefore a shared receive queue based QP. The second, third and fourth fields are the same as in the per-peer based QP. The fifth field is the number of outstanding sends that are allowed at a given time on the QP (32). This provides a "good enough" mechanism of flow control for some regular communication patterns. QPs MUST be specified in ascending receive buffer size order. This requirement may be removed prior to 1.3 release. This commit was SVN r15474.
2007-07-18 05:15:59 +04:00
/* Calculate the smallest freelist size that can be allowed */
if (rd_num + rd_rsv > min_freelist_size)
min_freelist_size = rd_num + rd_rsv;
mca_btl_openib_component.qp_infos[qp].u.pp_qp.rd_win = rd_win;
mca_btl_openib_component.qp_infos[qp].u.pp_qp.rd_rsv = rd_rsv;
if((rd_num - rd_low) > rd_win)
opal_show_help("help-mpi-btl-openib.txt", "non optimal rd_win",
true, rd_win, rd_num - rd_low);
} else {
int32_t sd_max;
if(count < 3 || count > 5) {
opal_show_help("help-mpi-btl-openib.txt",
"invalid srq specification", true,
orte_system_info.nodename, queues[qp]);
This commit brings in two major things: 1. Galen's fine-grain control of queue pair resources in the openib BTL. 1. Pasha's new implementation of asychronous HCA event handling. Pasha's new implementation doesn't take much explanation, but the new "multifrag" stuff does. Note that "svn merge" was not used to bring this new code from the /tmp/ib_multifrag branch -- something Bad happened in the periodic trunk pulls on that branch making an actual merge back to the trunk effectively impossible (i.e., lots and lots of arbitrary conflicts and artifical changes). :-( == Fine-grain control of queue pair resources == Galen's fine-grain control of queue pair resources to the OpenIB BTL (thanks to Gleb for fixing broken code and providing additional functionality, Pasha for finding broken code, and Jeff for doing all the svn work and regression testing). Prior to this commit, the OpenIB BTL created two queue pairs: one for eager size fragments and one for max send size fragments. When the use of the shared receive queue (SRQ) was specified (via "-mca btl_openib_use_srq 1"), these QPs would use a shared receive queue for receive buffers instead of the default per-peer (PP) receive queues and buffers. One consequence of this design is that receive buffer utilization (the size of the data received as a percentage of the receive buffer used for the data) was quite poor for a number of applications. The new design allows multiple QPs to be specified at runtime. Each QP can be setup to use PP or SRQ receive buffers as well as giving fine-grained control over receive buffer size, number of receive buffers to post, when to replenish the receive queue (low water mark) and for SRQ QPs, the number of outstanding sends can also be specified. The following is an example of the syntax to describe QPs to the OpenIB BTL using the new MCA parameter btl_openib_receive_queues: {{{ -mca btl_openib_receive_queues \ "P,128,16,4;S,1024,256,128,32;S,4096,256,128,32;S,65536,256,128,32" }}} Each QP description is delimited by ";" (semicolon) with individual fields of the QP description delimited by "," (comma). The above example therefore describes 4 QPs. The first QP is: P,128,16,4 Meaning: per-peer receive buffer QPs are indicated by a starting field of "P"; the first QP (shown above) is therefore a per-peer based QP. The second field indicates the size of the receive buffer in bytes (128 bytes). The third field indicates the number of receive buffers to allocate to the QP (16). The fourth field indicates the low watermark for receive buffers at which time the BTL will repost receive buffers to the QP (4). The second QP is: S,1024,256,128,32 Shared receive queue based QPs are indicated by a starting field of "S"; the second QP (shown above) is therefore a shared receive queue based QP. The second, third and fourth fields are the same as in the per-peer based QP. The fifth field is the number of outstanding sends that are allowed at a given time on the QP (32). This provides a "good enough" mechanism of flow control for some regular communication patterns. QPs MUST be specified in ascending receive buffer size order. This requirement may be removed prior to 1.3 release. This commit was SVN r15474.
2007-07-18 05:15:59 +04:00
goto error;
}
mca_btl_openib_component.qp_infos[qp].type = (params[0][0] =='X') ?
MCA_BTL_OPENIB_XRC_QP : MCA_BTL_OPENIB_SRQ_QP;
mca_btl_openib_component.qp_infos[qp].size = atoi_param(P(1), 0);
rd_num = atoi_param(P(2), 256);
/* by default set rd_low to be 3/4 of rd_num */
rd_low = atoi_param(P(3), rd_num - (rd_num / 4));
sd_max = atoi_param(P(4), rd_low / 4);
BTL_VERBOSE(("srq: rd_num is %d rd_low is %d sd_max is %d\n",
rd_num, rd_low, sd_max));
/* Calculate the smallest freelist size that can be allowed */
if (rd_num > min_freelist_size)
min_freelist_size = rd_num;
mca_btl_openib_component.qp_infos[qp].u.srq_qp.sd_max = sd_max;
This commit brings in two major things: 1. Galen's fine-grain control of queue pair resources in the openib BTL. 1. Pasha's new implementation of asychronous HCA event handling. Pasha's new implementation doesn't take much explanation, but the new "multifrag" stuff does. Note that "svn merge" was not used to bring this new code from the /tmp/ib_multifrag branch -- something Bad happened in the periodic trunk pulls on that branch making an actual merge back to the trunk effectively impossible (i.e., lots and lots of arbitrary conflicts and artifical changes). :-( == Fine-grain control of queue pair resources == Galen's fine-grain control of queue pair resources to the OpenIB BTL (thanks to Gleb for fixing broken code and providing additional functionality, Pasha for finding broken code, and Jeff for doing all the svn work and regression testing). Prior to this commit, the OpenIB BTL created two queue pairs: one for eager size fragments and one for max send size fragments. When the use of the shared receive queue (SRQ) was specified (via "-mca btl_openib_use_srq 1"), these QPs would use a shared receive queue for receive buffers instead of the default per-peer (PP) receive queues and buffers. One consequence of this design is that receive buffer utilization (the size of the data received as a percentage of the receive buffer used for the data) was quite poor for a number of applications. The new design allows multiple QPs to be specified at runtime. Each QP can be setup to use PP or SRQ receive buffers as well as giving fine-grained control over receive buffer size, number of receive buffers to post, when to replenish the receive queue (low water mark) and for SRQ QPs, the number of outstanding sends can also be specified. The following is an example of the syntax to describe QPs to the OpenIB BTL using the new MCA parameter btl_openib_receive_queues: {{{ -mca btl_openib_receive_queues \ "P,128,16,4;S,1024,256,128,32;S,4096,256,128,32;S,65536,256,128,32" }}} Each QP description is delimited by ";" (semicolon) with individual fields of the QP description delimited by "," (comma). The above example therefore describes 4 QPs. The first QP is: P,128,16,4 Meaning: per-peer receive buffer QPs are indicated by a starting field of "P"; the first QP (shown above) is therefore a per-peer based QP. The second field indicates the size of the receive buffer in bytes (128 bytes). The third field indicates the number of receive buffers to allocate to the QP (16). The fourth field indicates the low watermark for receive buffers at which time the BTL will repost receive buffers to the QP (4). The second QP is: S,1024,256,128,32 Shared receive queue based QPs are indicated by a starting field of "S"; the second QP (shown above) is therefore a shared receive queue based QP. The second, third and fourth fields are the same as in the per-peer based QP. The fifth field is the number of outstanding sends that are allowed at a given time on the QP (32). This provides a "good enough" mechanism of flow control for some regular communication patterns. QPs MUST be specified in ascending receive buffer size order. This requirement may be removed prior to 1.3 release. This commit was SVN r15474.
2007-07-18 05:15:59 +04:00
}
if (rd_num <= rd_low) {
opal_show_help("help-mpi-btl-openib.txt", "rd_num must be > rd_low",
true, orte_system_info.nodename, queues[qp]);
goto error;
}
mca_btl_openib_component.qp_infos[qp].rd_num = rd_num;
mca_btl_openib_component.qp_infos[qp].rd_low = rd_low;
while (NULL != params[i]) {
This commit brings in two major things: 1. Galen's fine-grain control of queue pair resources in the openib BTL. 1. Pasha's new implementation of asychronous HCA event handling. Pasha's new implementation doesn't take much explanation, but the new "multifrag" stuff does. Note that "svn merge" was not used to bring this new code from the /tmp/ib_multifrag branch -- something Bad happened in the periodic trunk pulls on that branch making an actual merge back to the trunk effectively impossible (i.e., lots and lots of arbitrary conflicts and artifical changes). :-( == Fine-grain control of queue pair resources == Galen's fine-grain control of queue pair resources to the OpenIB BTL (thanks to Gleb for fixing broken code and providing additional functionality, Pasha for finding broken code, and Jeff for doing all the svn work and regression testing). Prior to this commit, the OpenIB BTL created two queue pairs: one for eager size fragments and one for max send size fragments. When the use of the shared receive queue (SRQ) was specified (via "-mca btl_openib_use_srq 1"), these QPs would use a shared receive queue for receive buffers instead of the default per-peer (PP) receive queues and buffers. One consequence of this design is that receive buffer utilization (the size of the data received as a percentage of the receive buffer used for the data) was quite poor for a number of applications. The new design allows multiple QPs to be specified at runtime. Each QP can be setup to use PP or SRQ receive buffers as well as giving fine-grained control over receive buffer size, number of receive buffers to post, when to replenish the receive queue (low water mark) and for SRQ QPs, the number of outstanding sends can also be specified. The following is an example of the syntax to describe QPs to the OpenIB BTL using the new MCA parameter btl_openib_receive_queues: {{{ -mca btl_openib_receive_queues \ "P,128,16,4;S,1024,256,128,32;S,4096,256,128,32;S,65536,256,128,32" }}} Each QP description is delimited by ";" (semicolon) with individual fields of the QP description delimited by "," (comma). The above example therefore describes 4 QPs. The first QP is: P,128,16,4 Meaning: per-peer receive buffer QPs are indicated by a starting field of "P"; the first QP (shown above) is therefore a per-peer based QP. The second field indicates the size of the receive buffer in bytes (128 bytes). The third field indicates the number of receive buffers to allocate to the QP (16). The fourth field indicates the low watermark for receive buffers at which time the BTL will repost receive buffers to the QP (4). The second QP is: S,1024,256,128,32 Shared receive queue based QPs are indicated by a starting field of "S"; the second QP (shown above) is therefore a shared receive queue based QP. The second, third and fourth fields are the same as in the per-peer based QP. The fifth field is the number of outstanding sends that are allowed at a given time on the QP (32). This provides a "good enough" mechanism of flow control for some regular communication patterns. QPs MUST be specified in ascending receive buffer size order. This requirement may be removed prior to 1.3 release. This commit was SVN r15474.
2007-07-18 05:15:59 +04:00
free(params[i++]);
}
This commit brings in two major things: 1. Galen's fine-grain control of queue pair resources in the openib BTL. 1. Pasha's new implementation of asychronous HCA event handling. Pasha's new implementation doesn't take much explanation, but the new "multifrag" stuff does. Note that "svn merge" was not used to bring this new code from the /tmp/ib_multifrag branch -- something Bad happened in the periodic trunk pulls on that branch making an actual merge back to the trunk effectively impossible (i.e., lots and lots of arbitrary conflicts and artifical changes). :-( == Fine-grain control of queue pair resources == Galen's fine-grain control of queue pair resources to the OpenIB BTL (thanks to Gleb for fixing broken code and providing additional functionality, Pasha for finding broken code, and Jeff for doing all the svn work and regression testing). Prior to this commit, the OpenIB BTL created two queue pairs: one for eager size fragments and one for max send size fragments. When the use of the shared receive queue (SRQ) was specified (via "-mca btl_openib_use_srq 1"), these QPs would use a shared receive queue for receive buffers instead of the default per-peer (PP) receive queues and buffers. One consequence of this design is that receive buffer utilization (the size of the data received as a percentage of the receive buffer used for the data) was quite poor for a number of applications. The new design allows multiple QPs to be specified at runtime. Each QP can be setup to use PP or SRQ receive buffers as well as giving fine-grained control over receive buffer size, number of receive buffers to post, when to replenish the receive queue (low water mark) and for SRQ QPs, the number of outstanding sends can also be specified. The following is an example of the syntax to describe QPs to the OpenIB BTL using the new MCA parameter btl_openib_receive_queues: {{{ -mca btl_openib_receive_queues \ "P,128,16,4;S,1024,256,128,32;S,4096,256,128,32;S,65536,256,128,32" }}} Each QP description is delimited by ";" (semicolon) with individual fields of the QP description delimited by "," (comma). The above example therefore describes 4 QPs. The first QP is: P,128,16,4 Meaning: per-peer receive buffer QPs are indicated by a starting field of "P"; the first QP (shown above) is therefore a per-peer based QP. The second field indicates the size of the receive buffer in bytes (128 bytes). The third field indicates the number of receive buffers to allocate to the QP (16). The fourth field indicates the low watermark for receive buffers at which time the BTL will repost receive buffers to the QP (4). The second QP is: S,1024,256,128,32 Shared receive queue based QPs are indicated by a starting field of "S"; the second QP (shown above) is therefore a shared receive queue based QP. The second, third and fourth fields are the same as in the per-peer based QP. The fifth field is the number of outstanding sends that are allowed at a given time on the QP (32). This provides a "good enough" mechanism of flow control for some regular communication patterns. QPs MUST be specified in ascending receive buffer size order. This requirement may be removed prior to 1.3 release. This commit was SVN r15474.
2007-07-18 05:15:59 +04:00
free(params);
qp++;
}
params = NULL;
/* Sanity check some sizes */
This commit brings in two major things: 1. Galen's fine-grain control of queue pair resources in the openib BTL. 1. Pasha's new implementation of asychronous HCA event handling. Pasha's new implementation doesn't take much explanation, but the new "multifrag" stuff does. Note that "svn merge" was not used to bring this new code from the /tmp/ib_multifrag branch -- something Bad happened in the periodic trunk pulls on that branch making an actual merge back to the trunk effectively impossible (i.e., lots and lots of arbitrary conflicts and artifical changes). :-( == Fine-grain control of queue pair resources == Galen's fine-grain control of queue pair resources to the OpenIB BTL (thanks to Gleb for fixing broken code and providing additional functionality, Pasha for finding broken code, and Jeff for doing all the svn work and regression testing). Prior to this commit, the OpenIB BTL created two queue pairs: one for eager size fragments and one for max send size fragments. When the use of the shared receive queue (SRQ) was specified (via "-mca btl_openib_use_srq 1"), these QPs would use a shared receive queue for receive buffers instead of the default per-peer (PP) receive queues and buffers. One consequence of this design is that receive buffer utilization (the size of the data received as a percentage of the receive buffer used for the data) was quite poor for a number of applications. The new design allows multiple QPs to be specified at runtime. Each QP can be setup to use PP or SRQ receive buffers as well as giving fine-grained control over receive buffer size, number of receive buffers to post, when to replenish the receive queue (low water mark) and for SRQ QPs, the number of outstanding sends can also be specified. The following is an example of the syntax to describe QPs to the OpenIB BTL using the new MCA parameter btl_openib_receive_queues: {{{ -mca btl_openib_receive_queues \ "P,128,16,4;S,1024,256,128,32;S,4096,256,128,32;S,65536,256,128,32" }}} Each QP description is delimited by ";" (semicolon) with individual fields of the QP description delimited by "," (comma). The above example therefore describes 4 QPs. The first QP is: P,128,16,4 Meaning: per-peer receive buffer QPs are indicated by a starting field of "P"; the first QP (shown above) is therefore a per-peer based QP. The second field indicates the size of the receive buffer in bytes (128 bytes). The third field indicates the number of receive buffers to allocate to the QP (16). The fourth field indicates the low watermark for receive buffers at which time the BTL will repost receive buffers to the QP (4). The second QP is: S,1024,256,128,32 Shared receive queue based QPs are indicated by a starting field of "S"; the second QP (shown above) is therefore a shared receive queue based QP. The second, third and fourth fields are the same as in the per-peer based QP. The fifth field is the number of outstanding sends that are allowed at a given time on the QP (32). This provides a "good enough" mechanism of flow control for some regular communication patterns. QPs MUST be specified in ascending receive buffer size order. This requirement may be removed prior to 1.3 release. This commit was SVN r15474.
2007-07-18 05:15:59 +04:00
max_qp_size = mca_btl_openib_component.qp_infos[mca_btl_openib_component.num_qps - 1].size;
This commit brings in two major things: 1. Galen's fine-grain control of queue pair resources in the openib BTL. 1. Pasha's new implementation of asychronous HCA event handling. Pasha's new implementation doesn't take much explanation, but the new "multifrag" stuff does. Note that "svn merge" was not used to bring this new code from the /tmp/ib_multifrag branch -- something Bad happened in the periodic trunk pulls on that branch making an actual merge back to the trunk effectively impossible (i.e., lots and lots of arbitrary conflicts and artifical changes). :-( == Fine-grain control of queue pair resources == Galen's fine-grain control of queue pair resources to the OpenIB BTL (thanks to Gleb for fixing broken code and providing additional functionality, Pasha for finding broken code, and Jeff for doing all the svn work and regression testing). Prior to this commit, the OpenIB BTL created two queue pairs: one for eager size fragments and one for max send size fragments. When the use of the shared receive queue (SRQ) was specified (via "-mca btl_openib_use_srq 1"), these QPs would use a shared receive queue for receive buffers instead of the default per-peer (PP) receive queues and buffers. One consequence of this design is that receive buffer utilization (the size of the data received as a percentage of the receive buffer used for the data) was quite poor for a number of applications. The new design allows multiple QPs to be specified at runtime. Each QP can be setup to use PP or SRQ receive buffers as well as giving fine-grained control over receive buffer size, number of receive buffers to post, when to replenish the receive queue (low water mark) and for SRQ QPs, the number of outstanding sends can also be specified. The following is an example of the syntax to describe QPs to the OpenIB BTL using the new MCA parameter btl_openib_receive_queues: {{{ -mca btl_openib_receive_queues \ "P,128,16,4;S,1024,256,128,32;S,4096,256,128,32;S,65536,256,128,32" }}} Each QP description is delimited by ";" (semicolon) with individual fields of the QP description delimited by "," (comma). The above example therefore describes 4 QPs. The first QP is: P,128,16,4 Meaning: per-peer receive buffer QPs are indicated by a starting field of "P"; the first QP (shown above) is therefore a per-peer based QP. The second field indicates the size of the receive buffer in bytes (128 bytes). The third field indicates the number of receive buffers to allocate to the QP (16). The fourth field indicates the low watermark for receive buffers at which time the BTL will repost receive buffers to the QP (4). The second QP is: S,1024,256,128,32 Shared receive queue based QPs are indicated by a starting field of "S"; the second QP (shown above) is therefore a shared receive queue based QP. The second, third and fourth fields are the same as in the per-peer based QP. The fifth field is the number of outstanding sends that are allowed at a given time on the QP (32). This provides a "good enough" mechanism of flow control for some regular communication patterns. QPs MUST be specified in ascending receive buffer size order. This requirement may be removed prior to 1.3 release. This commit was SVN r15474.
2007-07-18 05:15:59 +04:00
max_size_needed = (mca_btl_openib_module.super.btl_eager_limit >
mca_btl_openib_module.super.btl_max_send_size) ?
This commit brings in two major things: 1. Galen's fine-grain control of queue pair resources in the openib BTL. 1. Pasha's new implementation of asychronous HCA event handling. Pasha's new implementation doesn't take much explanation, but the new "multifrag" stuff does. Note that "svn merge" was not used to bring this new code from the /tmp/ib_multifrag branch -- something Bad happened in the periodic trunk pulls on that branch making an actual merge back to the trunk effectively impossible (i.e., lots and lots of arbitrary conflicts and artifical changes). :-( == Fine-grain control of queue pair resources == Galen's fine-grain control of queue pair resources to the OpenIB BTL (thanks to Gleb for fixing broken code and providing additional functionality, Pasha for finding broken code, and Jeff for doing all the svn work and regression testing). Prior to this commit, the OpenIB BTL created two queue pairs: one for eager size fragments and one for max send size fragments. When the use of the shared receive queue (SRQ) was specified (via "-mca btl_openib_use_srq 1"), these QPs would use a shared receive queue for receive buffers instead of the default per-peer (PP) receive queues and buffers. One consequence of this design is that receive buffer utilization (the size of the data received as a percentage of the receive buffer used for the data) was quite poor for a number of applications. The new design allows multiple QPs to be specified at runtime. Each QP can be setup to use PP or SRQ receive buffers as well as giving fine-grained control over receive buffer size, number of receive buffers to post, when to replenish the receive queue (low water mark) and for SRQ QPs, the number of outstanding sends can also be specified. The following is an example of the syntax to describe QPs to the OpenIB BTL using the new MCA parameter btl_openib_receive_queues: {{{ -mca btl_openib_receive_queues \ "P,128,16,4;S,1024,256,128,32;S,4096,256,128,32;S,65536,256,128,32" }}} Each QP description is delimited by ";" (semicolon) with individual fields of the QP description delimited by "," (comma). The above example therefore describes 4 QPs. The first QP is: P,128,16,4 Meaning: per-peer receive buffer QPs are indicated by a starting field of "P"; the first QP (shown above) is therefore a per-peer based QP. The second field indicates the size of the receive buffer in bytes (128 bytes). The third field indicates the number of receive buffers to allocate to the QP (16). The fourth field indicates the low watermark for receive buffers at which time the BTL will repost receive buffers to the QP (4). The second QP is: S,1024,256,128,32 Shared receive queue based QPs are indicated by a starting field of "S"; the second QP (shown above) is therefore a shared receive queue based QP. The second, third and fourth fields are the same as in the per-peer based QP. The fifth field is the number of outstanding sends that are allowed at a given time on the QP (32). This provides a "good enough" mechanism of flow control for some regular communication patterns. QPs MUST be specified in ascending receive buffer size order. This requirement may be removed prior to 1.3 release. This commit was SVN r15474.
2007-07-18 05:15:59 +04:00
mca_btl_openib_module.super.btl_eager_limit :
mca_btl_openib_module.super.btl_max_send_size;
if (max_qp_size < max_size_needed) {
opal_show_help("help-mpi-btl-openib.txt",
"biggest qp size is too small", true,
orte_system_info.nodename, max_qp_size,
max_size_needed);
This commit brings in two major things: 1. Galen's fine-grain control of queue pair resources in the openib BTL. 1. Pasha's new implementation of asychronous HCA event handling. Pasha's new implementation doesn't take much explanation, but the new "multifrag" stuff does. Note that "svn merge" was not used to bring this new code from the /tmp/ib_multifrag branch -- something Bad happened in the periodic trunk pulls on that branch making an actual merge back to the trunk effectively impossible (i.e., lots and lots of arbitrary conflicts and artifical changes). :-( == Fine-grain control of queue pair resources == Galen's fine-grain control of queue pair resources to the OpenIB BTL (thanks to Gleb for fixing broken code and providing additional functionality, Pasha for finding broken code, and Jeff for doing all the svn work and regression testing). Prior to this commit, the OpenIB BTL created two queue pairs: one for eager size fragments and one for max send size fragments. When the use of the shared receive queue (SRQ) was specified (via "-mca btl_openib_use_srq 1"), these QPs would use a shared receive queue for receive buffers instead of the default per-peer (PP) receive queues and buffers. One consequence of this design is that receive buffer utilization (the size of the data received as a percentage of the receive buffer used for the data) was quite poor for a number of applications. The new design allows multiple QPs to be specified at runtime. Each QP can be setup to use PP or SRQ receive buffers as well as giving fine-grained control over receive buffer size, number of receive buffers to post, when to replenish the receive queue (low water mark) and for SRQ QPs, the number of outstanding sends can also be specified. The following is an example of the syntax to describe QPs to the OpenIB BTL using the new MCA parameter btl_openib_receive_queues: {{{ -mca btl_openib_receive_queues \ "P,128,16,4;S,1024,256,128,32;S,4096,256,128,32;S,65536,256,128,32" }}} Each QP description is delimited by ";" (semicolon) with individual fields of the QP description delimited by "," (comma). The above example therefore describes 4 QPs. The first QP is: P,128,16,4 Meaning: per-peer receive buffer QPs are indicated by a starting field of "P"; the first QP (shown above) is therefore a per-peer based QP. The second field indicates the size of the receive buffer in bytes (128 bytes). The third field indicates the number of receive buffers to allocate to the QP (16). The fourth field indicates the low watermark for receive buffers at which time the BTL will repost receive buffers to the QP (4). The second QP is: S,1024,256,128,32 Shared receive queue based QPs are indicated by a starting field of "S"; the second QP (shown above) is therefore a shared receive queue based QP. The second, third and fourth fields are the same as in the per-peer based QP. The fifth field is the number of outstanding sends that are allowed at a given time on the QP (32). This provides a "good enough" mechanism of flow control for some regular communication patterns. QPs MUST be specified in ascending receive buffer size order. This requirement may be removed prior to 1.3 release. This commit was SVN r15474.
2007-07-18 05:15:59 +04:00
ret = OMPI_ERROR;
goto error;
} else if (max_qp_size > max_size_needed) {
opal_show_help("help-mpi-btl-openib.txt",
"biggest qp size is too big", true,
orte_system_info.nodename, max_qp_size,
max_size_needed);
This commit brings in two major things: 1. Galen's fine-grain control of queue pair resources in the openib BTL. 1. Pasha's new implementation of asychronous HCA event handling. Pasha's new implementation doesn't take much explanation, but the new "multifrag" stuff does. Note that "svn merge" was not used to bring this new code from the /tmp/ib_multifrag branch -- something Bad happened in the periodic trunk pulls on that branch making an actual merge back to the trunk effectively impossible (i.e., lots and lots of arbitrary conflicts and artifical changes). :-( == Fine-grain control of queue pair resources == Galen's fine-grain control of queue pair resources to the OpenIB BTL (thanks to Gleb for fixing broken code and providing additional functionality, Pasha for finding broken code, and Jeff for doing all the svn work and regression testing). Prior to this commit, the OpenIB BTL created two queue pairs: one for eager size fragments and one for max send size fragments. When the use of the shared receive queue (SRQ) was specified (via "-mca btl_openib_use_srq 1"), these QPs would use a shared receive queue for receive buffers instead of the default per-peer (PP) receive queues and buffers. One consequence of this design is that receive buffer utilization (the size of the data received as a percentage of the receive buffer used for the data) was quite poor for a number of applications. The new design allows multiple QPs to be specified at runtime. Each QP can be setup to use PP or SRQ receive buffers as well as giving fine-grained control over receive buffer size, number of receive buffers to post, when to replenish the receive queue (low water mark) and for SRQ QPs, the number of outstanding sends can also be specified. The following is an example of the syntax to describe QPs to the OpenIB BTL using the new MCA parameter btl_openib_receive_queues: {{{ -mca btl_openib_receive_queues \ "P,128,16,4;S,1024,256,128,32;S,4096,256,128,32;S,65536,256,128,32" }}} Each QP description is delimited by ";" (semicolon) with individual fields of the QP description delimited by "," (comma). The above example therefore describes 4 QPs. The first QP is: P,128,16,4 Meaning: per-peer receive buffer QPs are indicated by a starting field of "P"; the first QP (shown above) is therefore a per-peer based QP. The second field indicates the size of the receive buffer in bytes (128 bytes). The third field indicates the number of receive buffers to allocate to the QP (16). The fourth field indicates the low watermark for receive buffers at which time the BTL will repost receive buffers to the QP (4). The second QP is: S,1024,256,128,32 Shared receive queue based QPs are indicated by a starting field of "S"; the second QP (shown above) is therefore a shared receive queue based QP. The second, third and fourth fields are the same as in the per-peer based QP. The fifth field is the number of outstanding sends that are allowed at a given time on the QP (32). This provides a "good enough" mechanism of flow control for some regular communication patterns. QPs MUST be specified in ascending receive buffer size order. This requirement may be removed prior to 1.3 release. This commit was SVN r15474.
2007-07-18 05:15:59 +04:00
opal_output(0, "The biggest QP size is bigger than maximum send size. "
"This is not optimal configuration as memory will be waisted.\n");
}
if (mca_btl_openib_component.ib_free_list_max > 0 &&
min_freelist_size > mca_btl_openib_component.ib_free_list_max) {
opal_show_help("help-mpi-btl-openib.txt", "freelist too small", true,
orte_system_info.nodename,
mca_btl_openib_component.ib_free_list_max,
min_freelist_size);
goto error;
}
This commit brings in two major things: 1. Galen's fine-grain control of queue pair resources in the openib BTL. 1. Pasha's new implementation of asychronous HCA event handling. Pasha's new implementation doesn't take much explanation, but the new "multifrag" stuff does. Note that "svn merge" was not used to bring this new code from the /tmp/ib_multifrag branch -- something Bad happened in the periodic trunk pulls on that branch making an actual merge back to the trunk effectively impossible (i.e., lots and lots of arbitrary conflicts and artifical changes). :-( == Fine-grain control of queue pair resources == Galen's fine-grain control of queue pair resources to the OpenIB BTL (thanks to Gleb for fixing broken code and providing additional functionality, Pasha for finding broken code, and Jeff for doing all the svn work and regression testing). Prior to this commit, the OpenIB BTL created two queue pairs: one for eager size fragments and one for max send size fragments. When the use of the shared receive queue (SRQ) was specified (via "-mca btl_openib_use_srq 1"), these QPs would use a shared receive queue for receive buffers instead of the default per-peer (PP) receive queues and buffers. One consequence of this design is that receive buffer utilization (the size of the data received as a percentage of the receive buffer used for the data) was quite poor for a number of applications. The new design allows multiple QPs to be specified at runtime. Each QP can be setup to use PP or SRQ receive buffers as well as giving fine-grained control over receive buffer size, number of receive buffers to post, when to replenish the receive queue (low water mark) and for SRQ QPs, the number of outstanding sends can also be specified. The following is an example of the syntax to describe QPs to the OpenIB BTL using the new MCA parameter btl_openib_receive_queues: {{{ -mca btl_openib_receive_queues \ "P,128,16,4;S,1024,256,128,32;S,4096,256,128,32;S,65536,256,128,32" }}} Each QP description is delimited by ";" (semicolon) with individual fields of the QP description delimited by "," (comma). The above example therefore describes 4 QPs. The first QP is: P,128,16,4 Meaning: per-peer receive buffer QPs are indicated by a starting field of "P"; the first QP (shown above) is therefore a per-peer based QP. The second field indicates the size of the receive buffer in bytes (128 bytes). The third field indicates the number of receive buffers to allocate to the QP (16). The fourth field indicates the low watermark for receive buffers at which time the BTL will repost receive buffers to the QP (4). The second QP is: S,1024,256,128,32 Shared receive queue based QPs are indicated by a starting field of "S"; the second QP (shown above) is therefore a shared receive queue based QP. The second, third and fourth fields are the same as in the per-peer based QP. The fifth field is the number of outstanding sends that are allowed at a given time on the QP (32). This provides a "good enough" mechanism of flow control for some regular communication patterns. QPs MUST be specified in ascending receive buffer size order. This requirement may be removed prior to 1.3 release. This commit was SVN r15474.
2007-07-18 05:15:59 +04:00
mca_btl_openib_component.rdma_qp = mca_btl_openib_component.num_qps - 1;
mca_btl_openib_component.credits_qp = smallest_pp_qp;
This commit brings in two major things: 1. Galen's fine-grain control of queue pair resources in the openib BTL. 1. Pasha's new implementation of asychronous HCA event handling. Pasha's new implementation doesn't take much explanation, but the new "multifrag" stuff does. Note that "svn merge" was not used to bring this new code from the /tmp/ib_multifrag branch -- something Bad happened in the periodic trunk pulls on that branch making an actual merge back to the trunk effectively impossible (i.e., lots and lots of arbitrary conflicts and artifical changes). :-( == Fine-grain control of queue pair resources == Galen's fine-grain control of queue pair resources to the OpenIB BTL (thanks to Gleb for fixing broken code and providing additional functionality, Pasha for finding broken code, and Jeff for doing all the svn work and regression testing). Prior to this commit, the OpenIB BTL created two queue pairs: one for eager size fragments and one for max send size fragments. When the use of the shared receive queue (SRQ) was specified (via "-mca btl_openib_use_srq 1"), these QPs would use a shared receive queue for receive buffers instead of the default per-peer (PP) receive queues and buffers. One consequence of this design is that receive buffer utilization (the size of the data received as a percentage of the receive buffer used for the data) was quite poor for a number of applications. The new design allows multiple QPs to be specified at runtime. Each QP can be setup to use PP or SRQ receive buffers as well as giving fine-grained control over receive buffer size, number of receive buffers to post, when to replenish the receive queue (low water mark) and for SRQ QPs, the number of outstanding sends can also be specified. The following is an example of the syntax to describe QPs to the OpenIB BTL using the new MCA parameter btl_openib_receive_queues: {{{ -mca btl_openib_receive_queues \ "P,128,16,4;S,1024,256,128,32;S,4096,256,128,32;S,65536,256,128,32" }}} Each QP description is delimited by ";" (semicolon) with individual fields of the QP description delimited by "," (comma). The above example therefore describes 4 QPs. The first QP is: P,128,16,4 Meaning: per-peer receive buffer QPs are indicated by a starting field of "P"; the first QP (shown above) is therefore a per-peer based QP. The second field indicates the size of the receive buffer in bytes (128 bytes). The third field indicates the number of receive buffers to allocate to the QP (16). The fourth field indicates the low watermark for receive buffers at which time the BTL will repost receive buffers to the QP (4). The second QP is: S,1024,256,128,32 Shared receive queue based QPs are indicated by a starting field of "S"; the second QP (shown above) is therefore a shared receive queue based QP. The second, third and fourth fields are the same as in the per-peer based QP. The fifth field is the number of outstanding sends that are allowed at a given time on the QP (32). This provides a "good enough" mechanism of flow control for some regular communication patterns. QPs MUST be specified in ascending receive buffer size order. This requirement may be removed prior to 1.3 release. This commit was SVN r15474.
2007-07-18 05:15:59 +04:00
/* Register any MCA params for the connect pseudo-components */
if (OMPI_SUCCESS != ompi_btl_openib_connect_base_open())
goto error;
ret = OMPI_SUCCESS;
This commit brings in two major things: 1. Galen's fine-grain control of queue pair resources in the openib BTL. 1. Pasha's new implementation of asychronous HCA event handling. Pasha's new implementation doesn't take much explanation, but the new "multifrag" stuff does. Note that "svn merge" was not used to bring this new code from the /tmp/ib_multifrag branch -- something Bad happened in the periodic trunk pulls on that branch making an actual merge back to the trunk effectively impossible (i.e., lots and lots of arbitrary conflicts and artifical changes). :-( == Fine-grain control of queue pair resources == Galen's fine-grain control of queue pair resources to the OpenIB BTL (thanks to Gleb for fixing broken code and providing additional functionality, Pasha for finding broken code, and Jeff for doing all the svn work and regression testing). Prior to this commit, the OpenIB BTL created two queue pairs: one for eager size fragments and one for max send size fragments. When the use of the shared receive queue (SRQ) was specified (via "-mca btl_openib_use_srq 1"), these QPs would use a shared receive queue for receive buffers instead of the default per-peer (PP) receive queues and buffers. One consequence of this design is that receive buffer utilization (the size of the data received as a percentage of the receive buffer used for the data) was quite poor for a number of applications. The new design allows multiple QPs to be specified at runtime. Each QP can be setup to use PP or SRQ receive buffers as well as giving fine-grained control over receive buffer size, number of receive buffers to post, when to replenish the receive queue (low water mark) and for SRQ QPs, the number of outstanding sends can also be specified. The following is an example of the syntax to describe QPs to the OpenIB BTL using the new MCA parameter btl_openib_receive_queues: {{{ -mca btl_openib_receive_queues \ "P,128,16,4;S,1024,256,128,32;S,4096,256,128,32;S,65536,256,128,32" }}} Each QP description is delimited by ";" (semicolon) with individual fields of the QP description delimited by "," (comma). The above example therefore describes 4 QPs. The first QP is: P,128,16,4 Meaning: per-peer receive buffer QPs are indicated by a starting field of "P"; the first QP (shown above) is therefore a per-peer based QP. The second field indicates the size of the receive buffer in bytes (128 bytes). The third field indicates the number of receive buffers to allocate to the QP (16). The fourth field indicates the low watermark for receive buffers at which time the BTL will repost receive buffers to the QP (4). The second QP is: S,1024,256,128,32 Shared receive queue based QPs are indicated by a starting field of "S"; the second QP (shown above) is therefore a shared receive queue based QP. The second, third and fourth fields are the same as in the per-peer based QP. The fifth field is the number of outstanding sends that are allowed at a given time on the QP (32). This provides a "good enough" mechanism of flow control for some regular communication patterns. QPs MUST be specified in ascending receive buffer size order. This requirement may be removed prior to 1.3 release. This commit was SVN r15474.
2007-07-18 05:15:59 +04:00
error:
if(params) {
qp = 0;
while(params[qp] != NULL)
free(params[qp++]);
free(params);
}
if(queues) {
qp = 0;
while(queues[qp] != NULL)
free(queues[qp++]);
free(queues);
}
return ret;
}