2005-07-01 01:28:35 +04:00
|
|
|
/*
|
2007-03-17 02:11:45 +03:00
|
|
|
* Copyright (c) 2004-2007 The Trustees of Indiana University and Indiana
|
2005-11-05 22:57:48 +03:00
|
|
|
* 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.
|
2007-04-27 01:03:38 +04:00
|
|
|
* Copyright (c) 2004-2007 High Performance Computing Center Stuttgart,
|
2005-07-01 01:28:35 +04:00
|
|
|
* University of Stuttgart. All rights reserved.
|
|
|
|
* Copyright (c) 2004-2005 The Regents of the University of California.
|
|
|
|
* All rights reserved.
|
2007-01-13 17:22:42 +03:00
|
|
|
* Copyright (c) 2006-2007 Cisco Systems, Inc. All rights reserved.
|
2007-02-15 21:03:20 +03:00
|
|
|
* Copyright (c) 2006-2007 Mellanox Technologies. All rights reserved.
|
2005-07-01 01:28:35 +04:00
|
|
|
* $COPYRIGHT$
|
|
|
|
*
|
|
|
|
* Additional copyrights may follow
|
|
|
|
*
|
|
|
|
* $HEADER$
|
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
|
|
|
*
|
2005-07-01 01:28:35 +04:00
|
|
|
* @file
|
|
|
|
*/
|
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
|
|
|
|
2007-04-27 01:03:38 +04:00
|
|
|
#ifndef MCA_BTL_IB_H
|
|
|
|
#define MCA_BTL_IB_H
|
2005-07-01 01:28:35 +04:00
|
|
|
|
|
|
|
/* Standard system includes */
|
|
|
|
#include <sys/types.h>
|
|
|
|
#include <string.h>
|
2006-06-28 11:23:08 +04:00
|
|
|
#include <infiniband/verbs.h>
|
2005-07-01 01:28:35 +04:00
|
|
|
|
|
|
|
/* Open MPI includes */
|
2006-02-12 04:33:29 +03:00
|
|
|
#include "ompi/class/ompi_free_list.h"
|
|
|
|
#include "ompi/class/ompi_bitmap.h"
|
2006-03-26 12:30:50 +04:00
|
|
|
#include "orte/class/orte_pointer_array.h"
|
2005-07-04 03:09:55 +04:00
|
|
|
#include "opal/event/event.h"
|
2006-02-12 04:33:29 +03:00
|
|
|
#include "ompi/mca/pml/pml.h"
|
|
|
|
#include "ompi/mca/btl/btl.h"
|
2005-07-04 03:31:27 +04:00
|
|
|
#include "opal/util/output.h"
|
2006-02-12 04:33:29 +03:00
|
|
|
#include "ompi/mca/mpool/mpool.h"
|
|
|
|
#include "ompi/mca/btl/base/btl_base_error.h"
|
2005-07-01 01:28:35 +04:00
|
|
|
|
2006-02-12 04:33:29 +03:00
|
|
|
#include "ompi/mca/btl/btl.h"
|
|
|
|
#include "ompi/mca/btl/base/base.h"
|
2006-09-07 17:05:41 +04:00
|
|
|
|
|
|
|
#include "btl_openib_frag.h"
|
2005-07-01 01:28:35 +04:00
|
|
|
|
2007-06-14 05:59:25 +04:00
|
|
|
BEGIN_C_DECLS
|
2005-07-01 01:28:35 +04:00
|
|
|
|
|
|
|
#define MCA_BTL_IB_LEAVE_PINNED 1
|
2006-09-26 16:12:33 +04:00
|
|
|
#define IB_DEFAULT_GID_PREFIX 0xfe80000000000000ll
|
2005-07-01 01:28:35 +04:00
|
|
|
|
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
|
|
|
|
2005-07-01 01:28:35 +04:00
|
|
|
/**
|
|
|
|
* Infiniband (IB) BTL component.
|
|
|
|
*/
|
|
|
|
|
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
|
|
|
typedef enum {
|
|
|
|
MCA_BTL_OPENIB_PP_QP,
|
|
|
|
MCA_BTL_OPENIB_SRQ_QP
|
|
|
|
} mca_btl_openib_qp_type_t;
|
|
|
|
|
|
|
|
struct mca_btl_openib_pp_qp_info_t {
|
|
|
|
int32_t rd_win;
|
|
|
|
int32_t rd_rsv;
|
|
|
|
}; typedef struct mca_btl_openib_pp_qp_info_t mca_btl_openib_pp_qp_info_t;
|
|
|
|
|
|
|
|
struct mca_btl_openib_srq_qp_info_t {
|
|
|
|
int32_t sd_max;
|
|
|
|
}; typedef struct mca_btl_openib_srq_qp_info_t mca_btl_openib_srq_qp_info_t;
|
|
|
|
|
|
|
|
struct mca_btl_openib_qp_info_t {
|
|
|
|
size_t size;
|
|
|
|
int32_t rd_num;
|
|
|
|
int32_t rd_low;
|
|
|
|
mca_btl_openib_qp_type_t type;
|
|
|
|
union {
|
|
|
|
mca_btl_openib_pp_qp_info_t pp_qp;
|
|
|
|
mca_btl_openib_srq_qp_info_t srq_qp;
|
|
|
|
} u;
|
|
|
|
}; typedef struct mca_btl_openib_qp_info_t mca_btl_openib_qp_info_t;
|
|
|
|
|
|
|
|
|
2005-07-01 01:28:35 +04:00
|
|
|
struct mca_btl_openib_component_t {
|
2006-08-18 02:02:01 +04:00
|
|
|
mca_btl_base_component_1_0_1_t super; /**< base BTL component */
|
2006-03-13 20:03:21 +03:00
|
|
|
|
|
|
|
int ib_max_btls;
|
|
|
|
/**< maximum number of hcas available to the IB component */
|
2005-07-01 01:28:35 +04:00
|
|
|
|
2006-06-01 06:32:18 +04:00
|
|
|
int ib_num_btls;
|
2005-07-01 01:28:35 +04:00
|
|
|
/**< number of hcas available to the IB component */
|
|
|
|
|
2007-05-09 01:47:21 +04:00
|
|
|
struct mca_btl_openib_module_t **openib_btls;
|
2007-04-27 01:03:38 +04:00
|
|
|
/**< array of available BTLs */
|
2005-07-01 01:28:35 +04:00
|
|
|
|
|
|
|
int ib_free_list_num;
|
|
|
|
/**< initial size of free lists */
|
|
|
|
|
|
|
|
int ib_free_list_max;
|
|
|
|
/**< maximum size of free lists */
|
|
|
|
|
|
|
|
int ib_free_list_inc;
|
|
|
|
/**< number of elements to alloc when growing free lists */
|
|
|
|
|
2005-07-03 20:22:16 +04:00
|
|
|
opal_list_t ib_procs;
|
2005-07-01 01:28:35 +04:00
|
|
|
/**< list of ib proc structures */
|
|
|
|
|
2005-07-04 03:09:55 +04:00
|
|
|
opal_event_t ib_send_event;
|
2005-07-01 01:28:35 +04:00
|
|
|
/**< event structure for sends */
|
|
|
|
|
2005-07-04 03:09:55 +04:00
|
|
|
opal_event_t ib_recv_event;
|
2005-07-01 01:28:35 +04:00
|
|
|
/**< event structure for recvs */
|
|
|
|
|
2005-07-04 02:45:48 +04:00
|
|
|
opal_mutex_t ib_lock;
|
2005-07-01 01:28:35 +04:00
|
|
|
/**< lock for accessing module state */
|
|
|
|
|
|
|
|
char* ib_mpool_name;
|
|
|
|
/**< name of ib memory pool */
|
2007-04-27 01:03:38 +04:00
|
|
|
|
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
|
|
|
uint8_t num_pp_qps; /**< number of pp qp's */
|
|
|
|
uint8_t num_srq_qps; /**< number of srq qp's */
|
|
|
|
uint8_t num_qps; /**< total number of qp's */
|
|
|
|
|
|
|
|
mca_btl_openib_qp_info_t* qp_infos;
|
|
|
|
|
2007-04-27 01:03:38 +04:00
|
|
|
size_t eager_limit; /**< Eager send limit of first fragment, in Bytes */
|
|
|
|
size_t max_send_size; /**< Maximum send size, in Bytes */
|
|
|
|
uint32_t reg_mru_len; /**< Length of the registration cache most recently used list */
|
|
|
|
uint32_t use_srq; /**< Use the Shared Receive Queue (SRQ mode) */
|
2005-10-02 22:58:57 +04:00
|
|
|
|
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
|
|
|
uint32_t ib_lp_cq_size; /**< Max outstanding CQE on the CQ */
|
|
|
|
uint32_t ib_hp_cq_size; /**< Max outstanding CQE on the CQ */
|
|
|
|
|
2007-04-27 01:03:38 +04:00
|
|
|
uint32_t ib_sg_list_size; /**< Max scatter/gather descriptor entries on the WQ */
|
|
|
|
uint32_t ib_pkey_ix; /**< InfiniBand pkey index */
|
2007-04-22 14:22:12 +04:00
|
|
|
uint32_t ib_pkey_val;
|
2005-07-15 19:13:19 +04:00
|
|
|
uint32_t ib_psn;
|
|
|
|
uint32_t ib_qp_ous_rd_atom;
|
|
|
|
uint32_t ib_mtu;
|
|
|
|
uint32_t ib_min_rnr_timer;
|
|
|
|
uint32_t ib_timeout;
|
|
|
|
uint32_t ib_retry_count;
|
|
|
|
uint32_t ib_rnr_retry;
|
|
|
|
uint32_t ib_max_rdma_dst_ops;
|
|
|
|
uint32_t ib_service_level;
|
|
|
|
uint32_t ib_static_rate;
|
2006-03-26 12:30:50 +04:00
|
|
|
uint32_t use_eager_rdma;
|
2007-04-27 01:03:38 +04:00
|
|
|
int32_t eager_rdma_threshold; /**< After this number of msg, use RDMA for short messages, always */
|
2006-03-26 12:30:50 +04:00
|
|
|
uint32_t eager_rdma_num;
|
2006-10-31 20:29:25 +03:00
|
|
|
int32_t max_eager_rdma;
|
2006-06-28 11:23:08 +04:00
|
|
|
uint32_t btls_per_lid;
|
|
|
|
uint32_t max_lmc;
|
2007-04-27 01:03:38 +04:00
|
|
|
uint32_t buffer_alignment; /**< Preferred communication buffer alignment in Bytes (must be power of two) */
|
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
|
|
|
|
int32_t fatal_counter; /**< Counts number on fatal events that we got on all hcas */
|
|
|
|
int async_pipe[2]; /**< Pipe for comunication with async event thread */
|
|
|
|
pthread_t async_thread; /**< Async thread that will handle fatal errors */
|
|
|
|
uint32_t use_async_event_thread; /**< Use the async event handler */
|
2007-05-09 01:47:21 +04:00
|
|
|
#endif
|
2007-06-14 05:59:25 +04:00
|
|
|
char *if_include;
|
|
|
|
char **if_include_list;
|
|
|
|
char *if_exclude;
|
|
|
|
char **if_exclude_list;
|
2005-07-15 19:13:19 +04:00
|
|
|
|
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
|
|
|
/** Colon-delimited list of filenames for HCA parameters */
|
|
|
|
char *hca_params_file_names;
|
|
|
|
|
|
|
|
/** Whether we're in verbose mode or not */
|
|
|
|
bool verbose;
|
|
|
|
|
|
|
|
/** Whether we want a warning if no HCA-specific parameters are
|
|
|
|
found in INI files */
|
|
|
|
bool warn_no_hca_params_found;
|
2006-09-26 16:12:33 +04:00
|
|
|
/** Whether we want a warning if non default GID prefix is not configured
|
|
|
|
on multiport setup */
|
|
|
|
bool warn_default_gid_prefix;
|
2007-06-14 05:59:25 +04:00
|
|
|
/** Whether we want a warning if the user specifies a non-existent
|
|
|
|
HCA and/or port via btl_openib_if_[in|ex]clude MCA params */
|
|
|
|
bool warn_nonexistent_if;
|
|
|
|
/** Dummy argv-style list; a copy of names from the
|
|
|
|
if_[in|ex]clude list that we use for error checking (to ensure
|
|
|
|
that they all exist) */
|
|
|
|
char **if_list;
|
2007-04-21 04:15:05 +04:00
|
|
|
#ifdef HAVE_IBV_FORK_INIT
|
|
|
|
/** Whether we want fork support or not */
|
|
|
|
int want_fork_support;
|
|
|
|
#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
|
|
|
int rdma_qp;
|
|
|
|
int eager_rdma_qp;
|
|
|
|
|
2005-07-01 01:28:35 +04:00
|
|
|
}; typedef struct mca_btl_openib_component_t mca_btl_openib_component_t;
|
|
|
|
|
2006-09-15 02:19:39 +04:00
|
|
|
OMPI_MODULE_DECLSPEC extern mca_btl_openib_component_t mca_btl_openib_component;
|
2005-07-01 01:28:35 +04:00
|
|
|
|
|
|
|
typedef mca_btl_base_recv_reg_t mca_btl_openib_recv_reg_t;
|
|
|
|
|
2006-09-07 17:05:41 +04:00
|
|
|
struct mca_btl_openib_port_info_t {
|
|
|
|
uint32_t mtu;
|
2006-12-04 23:11:42 +03:00
|
|
|
#if OMPI_ENABLE_HETEROGENEOUS_SUPPORT
|
|
|
|
uint8_t padding[4];
|
|
|
|
#endif
|
2007-01-13 01:42:20 +03:00
|
|
|
uint64_t subnet_id;
|
2006-09-07 17:05:41 +04:00
|
|
|
};
|
|
|
|
typedef struct mca_btl_openib_port_info_t mca_btl_openib_port_info_t;
|
2005-07-01 01:28:35 +04:00
|
|
|
|
2007-01-13 02:14:45 +03:00
|
|
|
#define MCA_BTL_OPENIB_PORT_INFO_NTOH(hdr) \
|
|
|
|
do { \
|
|
|
|
(hdr).mtu = ntohl((hdr).mtu); \
|
2007-01-13 17:22:42 +03:00
|
|
|
(hdr).subnet_id = ntoh64((hdr).subnet_id); \
|
2007-01-13 02:14:45 +03:00
|
|
|
} while (0)
|
|
|
|
#define MCA_BTL_OPENIB_PORT_INFO_HTON(hdr) \
|
|
|
|
do { \
|
|
|
|
(hdr).mtu = htonl((hdr).mtu); \
|
2007-01-13 17:22:42 +03:00
|
|
|
(hdr).subnet_id = hton64((hdr).subnet_id); \
|
2007-01-13 02:14:45 +03:00
|
|
|
} while (0)
|
|
|
|
|
2006-06-28 11:23:08 +04:00
|
|
|
struct mca_btl_openib_hca_t {
|
|
|
|
struct ibv_device *ib_dev; /* the ib device */
|
2006-11-02 19:15:21 +03:00
|
|
|
#if OMPI_ENABLE_PROGRESS_THREADS == 1
|
|
|
|
struct ibv_comp_channel *ib_channel; /* Channel event for the HCA */
|
|
|
|
opal_thread_t thread; /* Progress thread */
|
|
|
|
volatile bool progress; /* Progress status */
|
|
|
|
#endif
|
|
|
|
opal_mutex_t hca_lock; /* hca level lock */
|
2006-06-28 11:23:08 +04:00
|
|
|
struct ibv_context *ib_dev_context;
|
|
|
|
struct ibv_device_attr ib_dev_attr;
|
|
|
|
struct ibv_pd *ib_pd;
|
|
|
|
mca_mpool_base_module_t *mpool;
|
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
|
|
|
/* MTU for this HCA */
|
|
|
|
uint32_t mtu;
|
2006-12-14 18:52:13 +03:00
|
|
|
/* Whether this HCA supports eager RDMA */
|
|
|
|
uint8_t use_eager_rdma;
|
2006-06-28 11:23:08 +04:00
|
|
|
uint8_t btls; /** < number of btls using this HCA */
|
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
|
|
|
|
volatile bool got_fatal_event;
|
2007-05-09 01:47:21 +04:00
|
|
|
#endif
|
2006-06-28 11:23:08 +04:00
|
|
|
};
|
|
|
|
typedef struct mca_btl_openib_hca_t mca_btl_openib_hca_t;
|
2007-04-27 01:03:38 +04:00
|
|
|
|
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
|
|
|
struct mca_btl_openib_module_pp_qp_t {
|
|
|
|
int32_t dummy;
|
|
|
|
}; typedef struct mca_btl_openib_module_pp_qp_t mca_btl_openib_module_pp_qp_t;
|
|
|
|
|
|
|
|
struct mca_btl_openib_module_srq_qp_t {
|
|
|
|
struct ibv_srq *srq;
|
|
|
|
int32_t rd_posted;
|
|
|
|
int32_t sd_credits; /* the max number of outstanding sends on a QP when using SRQ */
|
|
|
|
/* i.e. the number of frags that can be outstanding (down counter) */
|
|
|
|
opal_list_t pending_frags; /**< list of pending frags */
|
|
|
|
|
|
|
|
|
|
|
|
}; typedef struct mca_btl_openib_module_srq_qp_t mca_btl_openib_module_srq_qp_t;
|
|
|
|
|
|
|
|
struct mca_btl_openib_module_qp_t {
|
|
|
|
ompi_free_list_t send_free; /**< free lists of send buffer descriptors */
|
|
|
|
ompi_free_list_t recv_free; /**< free lists of receive buffer descriptors */
|
|
|
|
mca_btl_openib_qp_type_t type;
|
|
|
|
union {
|
|
|
|
mca_btl_openib_module_pp_qp_t pp_qp;
|
|
|
|
mca_btl_openib_module_srq_qp_t srq_qp;
|
|
|
|
} u;
|
|
|
|
}; typedef struct mca_btl_openib_module_qp_t mca_btl_openib_module_qp_t;
|
|
|
|
|
2005-07-01 01:28:35 +04:00
|
|
|
/**
|
2007-04-27 01:03:38 +04:00
|
|
|
* IB BTL Interface
|
2005-07-01 01:28:35 +04:00
|
|
|
*/
|
|
|
|
struct mca_btl_openib_module_t {
|
2007-04-27 01:03:38 +04:00
|
|
|
mca_btl_base_module_t super; /**< base BTL interface */
|
2005-07-01 01:28:35 +04:00
|
|
|
bool btl_inited;
|
|
|
|
mca_btl_openib_recv_reg_t ib_reg[256];
|
2007-01-13 01:42:20 +03:00
|
|
|
mca_btl_openib_port_info_t port_info; /* contains only the subnet id right now */
|
2006-06-28 11:23:08 +04:00
|
|
|
mca_btl_openib_hca_t *hca;
|
2007-04-27 01:03:38 +04:00
|
|
|
uint8_t port_num; /**< ID of the PORT */
|
2007-04-22 14:22:12 +04:00
|
|
|
uint16_t pkey_index;
|
2006-09-12 13:17:59 +04:00
|
|
|
struct ibv_cq *ib_cq[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
|
|
|
uint32_t cq_users[2];
|
2006-06-28 11:23:08 +04:00
|
|
|
struct ibv_port_attr ib_port_attr;
|
|
|
|
uint16_t lid; /**< lid that is actually used (for LMC) */
|
|
|
|
uint8_t src_path_bits; /**< offset from base lid (for LMC) */
|
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
|
|
|
|
|
|
|
int32_t num_peers;
|
2005-07-01 01:28:35 +04:00
|
|
|
|
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
|
|
|
ompi_free_list_t send_user_free; /**< free list of frags only...
|
|
|
|
* used for pining user memory */
|
|
|
|
|
|
|
|
ompi_free_list_t recv_user_free; /**< free list of frags only...
|
|
|
|
* used for pining user memory */
|
2005-07-01 01:28:35 +04:00
|
|
|
|
2006-07-27 18:09:30 +04:00
|
|
|
ompi_free_list_t send_free_control; /**< frags for control massages */
|
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
|
|
|
|
2007-04-27 01:03:38 +04:00
|
|
|
opal_mutex_t ib_lock; /**< module level lock */
|
2005-07-01 01:28:35 +04:00
|
|
|
|
|
|
|
size_t ib_inline_max; /**< max size of inline send*/
|
2005-07-20 01:04:22 +04:00
|
|
|
bool poll_cq;
|
2005-07-15 19:13:19 +04:00
|
|
|
|
2005-10-21 06:21:45 +04:00
|
|
|
|
2007-04-27 01:03:38 +04:00
|
|
|
size_t eager_rdma_frag_size; /**< length of eager frag */
|
|
|
|
orte_pointer_array_t *eager_rdma_buffers; /**< RDMA buffers to poll */
|
|
|
|
volatile int32_t eager_rdma_buffers_count; /**< number of RDMA buffers */
|
2006-08-17 00:21:38 +04:00
|
|
|
|
|
|
|
mca_btl_base_module_error_cb_fn_t error_cb; /**< error handler */
|
2006-09-05 20:04:04 +04:00
|
|
|
|
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_qp_t * qps;
|
|
|
|
|
2006-09-05 20:04:04 +04:00
|
|
|
orte_pointer_array_t *endpoints;
|
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
|
|
|
|
2007-04-27 01:03:38 +04:00
|
|
|
};
|
|
|
|
typedef struct mca_btl_openib_module_t mca_btl_openib_module_t;
|
2006-12-17 15:26:41 +03:00
|
|
|
|
2005-07-01 01:28:35 +04:00
|
|
|
extern mca_btl_openib_module_t mca_btl_openib_module;
|
|
|
|
|
2006-12-17 15:26:41 +03:00
|
|
|
struct mca_btl_openib_reg_t {
|
|
|
|
mca_mpool_base_registration_t base;
|
|
|
|
struct ibv_mr *mr;
|
|
|
|
};
|
|
|
|
typedef struct mca_btl_openib_reg_t mca_btl_openib_reg_t;
|
|
|
|
|
2006-11-02 19:15:21 +03:00
|
|
|
#if OMPI_ENABLE_PROGRESS_THREADS == 1
|
|
|
|
extern void* mca_btl_openib_progress_thread(opal_object_t*);
|
|
|
|
#endif
|
2007-04-27 01:03:38 +04:00
|
|
|
|
2005-07-01 01:28:35 +04:00
|
|
|
/**
|
|
|
|
* Register a callback function that is called on receipt
|
|
|
|
* of a fragment.
|
|
|
|
*
|
|
|
|
* @param btl (IN) BTL module
|
|
|
|
* @return Status indicating if cleanup was successful
|
|
|
|
*
|
|
|
|
* When the process list changes, the PML notifies the BTL of the
|
|
|
|
* change, to provide the opportunity to cleanup or release any
|
|
|
|
* resources associated with the peer.
|
|
|
|
*/
|
|
|
|
|
|
|
|
int mca_btl_openib_register(
|
|
|
|
struct mca_btl_base_module_t* btl,
|
|
|
|
mca_btl_base_tag_t tag,
|
|
|
|
mca_btl_base_module_recv_cb_fn_t cbfunc,
|
|
|
|
void* cbdata
|
|
|
|
);
|
2006-08-17 00:21:38 +04:00
|
|
|
|
|
|
|
|
|
|
|
/**
|
|
|
|
* Register a callback function that is called on error..
|
|
|
|
*
|
|
|
|
* @param btl (IN) BTL module
|
|
|
|
* @return Status indicating if cleanup was successful
|
|
|
|
*/
|
|
|
|
|
|
|
|
int mca_btl_openib_register_error_cb(
|
|
|
|
struct mca_btl_base_module_t* btl,
|
|
|
|
mca_btl_base_module_error_cb_fn_t cbfunc
|
|
|
|
);
|
2005-07-01 01:28:35 +04:00
|
|
|
|
|
|
|
|
|
|
|
/**
|
|
|
|
* Cleanup any resources held by the BTL.
|
|
|
|
*
|
|
|
|
* @param btl BTL instance.
|
|
|
|
* @return OMPI_SUCCESS or error status on failure.
|
|
|
|
*/
|
|
|
|
|
|
|
|
extern int mca_btl_openib_finalize(
|
|
|
|
struct mca_btl_base_module_t* btl
|
|
|
|
);
|
|
|
|
|
|
|
|
|
|
|
|
/**
|
|
|
|
* PML->BTL notification of change in the process list.
|
|
|
|
*
|
2007-04-27 01:03:38 +04:00
|
|
|
* @param btl (IN) BTL module
|
|
|
|
* @param nprocs (IN) Number of processes
|
|
|
|
* @param procs (IN) Set of processes
|
|
|
|
* @param peers (OUT) Set of (optional) peer addressing info.
|
|
|
|
* @param reachable (IN/OUT) Set of processes that are reachable via this BTL.
|
2005-07-01 01:28:35 +04:00
|
|
|
* @return OMPI_SUCCESS or error status on failure.
|
|
|
|
*
|
|
|
|
*/
|
|
|
|
|
|
|
|
extern int mca_btl_openib_add_procs(
|
|
|
|
struct mca_btl_base_module_t* btl,
|
|
|
|
size_t nprocs,
|
|
|
|
struct ompi_proc_t **procs,
|
|
|
|
struct mca_btl_base_endpoint_t** peers,
|
|
|
|
ompi_bitmap_t* reachable
|
|
|
|
);
|
|
|
|
|
|
|
|
/**
|
|
|
|
* PML->BTL notification of change in the process list.
|
|
|
|
*
|
|
|
|
* @param btl (IN) BTL instance
|
|
|
|
* @param nproc (IN) Number of processes.
|
|
|
|
* @param procs (IN) Set of processes.
|
|
|
|
* @param peers (IN) Set of peer data structures.
|
|
|
|
* @return Status indicating if cleanup was successful
|
|
|
|
*
|
|
|
|
*/
|
|
|
|
extern int mca_btl_openib_del_procs(
|
|
|
|
struct mca_btl_base_module_t* btl,
|
|
|
|
size_t nprocs,
|
|
|
|
struct ompi_proc_t **procs,
|
|
|
|
struct mca_btl_base_endpoint_t** peers
|
|
|
|
);
|
|
|
|
|
|
|
|
|
|
|
|
/**
|
|
|
|
* PML->BTL Initiate a send of the specified size.
|
|
|
|
*
|
|
|
|
* @param btl (IN) BTL instance
|
2007-04-27 01:03:38 +04:00
|
|
|
* @param btl_peer (IN) BTL peer addressing
|
|
|
|
* @param descriptor (IN) Descriptor of data to be transmitted.
|
|
|
|
* @param tag (IN) Tag.
|
2005-07-01 01:28:35 +04:00
|
|
|
*/
|
|
|
|
extern int mca_btl_openib_send(
|
|
|
|
struct mca_btl_base_module_t* btl,
|
|
|
|
struct mca_btl_base_endpoint_t* btl_peer,
|
|
|
|
struct mca_btl_base_descriptor_t* descriptor,
|
|
|
|
mca_btl_base_tag_t tag
|
|
|
|
);
|
|
|
|
|
|
|
|
/**
|
|
|
|
* PML->BTL Initiate a put of the specified size.
|
|
|
|
*
|
|
|
|
* @param btl (IN) BTL instance
|
2007-04-27 01:03:38 +04:00
|
|
|
* @param btl_peer (IN) BTL peer addressing
|
|
|
|
* @param descriptor (IN) Descriptor of data to be transmitted.
|
2005-07-01 01:28:35 +04:00
|
|
|
*/
|
|
|
|
extern int mca_btl_openib_put(
|
|
|
|
struct mca_btl_base_module_t* btl,
|
|
|
|
struct mca_btl_base_endpoint_t* btl_peer,
|
2007-04-27 01:03:38 +04:00
|
|
|
struct mca_btl_base_descriptor_t* descriptor
|
2005-08-18 21:08:27 +04:00
|
|
|
);
|
|
|
|
|
|
|
|
/**
|
|
|
|
* PML->BTL Initiate a get of the specified size.
|
|
|
|
*
|
|
|
|
* @param btl (IN) BTL instance
|
|
|
|
* @param btl_base_peer (IN) BTL peer addressing
|
2007-04-27 01:03:38 +04:00
|
|
|
* @param descriptor (IN) Descriptor of data to be transmitted.
|
2005-08-18 21:08:27 +04:00
|
|
|
*/
|
|
|
|
extern int mca_btl_openib_get(
|
|
|
|
struct mca_btl_base_module_t* btl,
|
|
|
|
struct mca_btl_base_endpoint_t* btl_peer,
|
2007-04-27 01:03:38 +04:00
|
|
|
struct mca_btl_base_descriptor_t* descriptor
|
2005-08-18 21:08:27 +04:00
|
|
|
);
|
|
|
|
|
2005-07-01 01:28:35 +04:00
|
|
|
|
|
|
|
/**
|
|
|
|
* Allocate a descriptor.
|
|
|
|
*
|
|
|
|
* @param btl (IN) BTL module
|
|
|
|
* @param size (IN) Requested descriptor size.
|
|
|
|
*/
|
|
|
|
extern mca_btl_base_descriptor_t* mca_btl_openib_alloc(
|
2005-07-12 17:38:54 +04:00
|
|
|
struct mca_btl_base_module_t* btl,
|
2007-05-24 23:51:26 +04:00
|
|
|
uint8_t order,
|
2005-07-12 17:38:54 +04:00
|
|
|
size_t size);
|
|
|
|
|
2005-07-01 01:28:35 +04:00
|
|
|
|
|
|
|
/**
|
|
|
|
* Return a segment allocated by this BTL.
|
|
|
|
*
|
|
|
|
* @param btl (IN) BTL module
|
|
|
|
* @param descriptor (IN) Allocated descriptor.
|
|
|
|
*/
|
|
|
|
extern int mca_btl_openib_free(
|
2005-07-12 17:38:54 +04:00
|
|
|
struct mca_btl_base_module_t* btl,
|
|
|
|
mca_btl_base_descriptor_t* des);
|
2005-07-01 01:28:35 +04:00
|
|
|
|
|
|
|
|
|
|
|
/**
|
|
|
|
* Pack data and return a descriptor that can be
|
|
|
|
* used for send/put.
|
|
|
|
*
|
|
|
|
* @param btl (IN) BTL module
|
|
|
|
* @param peer (IN) BTL peer addressing
|
|
|
|
*/
|
|
|
|
mca_btl_base_descriptor_t* mca_btl_openib_prepare_src(
|
2005-07-12 17:38:54 +04:00
|
|
|
struct mca_btl_base_module_t* btl,
|
|
|
|
struct mca_btl_base_endpoint_t* peer,
|
|
|
|
mca_mpool_base_registration_t* registration,
|
|
|
|
struct ompi_convertor_t* convertor,
|
2007-05-24 23:51:26 +04:00
|
|
|
uint8_t order,
|
2005-07-12 17:38:54 +04:00
|
|
|
size_t reserve,
|
|
|
|
size_t* size
|
|
|
|
);
|
2005-07-01 01:28:35 +04:00
|
|
|
|
|
|
|
/**
|
|
|
|
* Allocate a descriptor initialized for RDMA write.
|
|
|
|
*
|
|
|
|
* @param btl (IN) BTL module
|
|
|
|
* @param peer (IN) BTL peer addressing
|
|
|
|
*/
|
|
|
|
extern mca_btl_base_descriptor_t* mca_btl_openib_prepare_dst(
|
2005-07-12 17:38:54 +04:00
|
|
|
struct mca_btl_base_module_t* btl,
|
|
|
|
struct mca_btl_base_endpoint_t* peer,
|
|
|
|
mca_mpool_base_registration_t* registration,
|
|
|
|
struct ompi_convertor_t* convertor,
|
2007-05-24 23:51:26 +04:00
|
|
|
uint8_t order,
|
2005-07-12 17:38:54 +04:00
|
|
|
size_t reserve,
|
|
|
|
size_t* size);
|
2005-07-01 01:28:35 +04:00
|
|
|
/**
|
|
|
|
* Return a send fragment to the modules free list.
|
|
|
|
*
|
2007-04-27 01:03:38 +04:00
|
|
|
* @param btl (IN) BTL module
|
2005-07-01 01:28:35 +04:00
|
|
|
* @param frag (IN) IB send fragment
|
|
|
|
*
|
|
|
|
*/
|
2006-12-17 15:26:41 +03:00
|
|
|
extern void mca_btl_openib_send_frag_return(mca_btl_base_module_t* btl,
|
2007-04-27 01:03:38 +04:00
|
|
|
mca_btl_openib_frag_t* frag);
|
2005-07-01 01:28:35 +04:00
|
|
|
|
|
|
|
|
2007-03-17 02:11:45 +03:00
|
|
|
/**
|
|
|
|
* Fault Tolerance Event Notification Function
|
2007-04-27 01:03:38 +04:00
|
|
|
*
|
|
|
|
* @param state (IN) Checkpoint State
|
2007-03-17 02:11:45 +03:00
|
|
|
* @return OMPI_SUCCESS or failure status
|
|
|
|
*/
|
2007-04-27 01:03:38 +04:00
|
|
|
extern int mca_btl_openib_ft_event(int state);
|
|
|
|
|
2007-03-17 02:11:45 +03:00
|
|
|
|
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
|
|
|
#define BTL_OPENIB_HP_CQ 0
|
|
|
|
#define BTL_OPENIB_LP_CQ 1
|
|
|
|
|
2006-09-05 20:00:18 +04:00
|
|
|
|
2007-04-27 01:03:38 +04:00
|
|
|
/**
|
|
|
|
* Post to Shared Receive Queue with certain priority
|
|
|
|
*
|
|
|
|
* @param openib_btl (IN) BTL module
|
|
|
|
* @param additional (IN) Additional Bytes to reserve
|
|
|
|
* @param prio (IN) Priority (either BTL_OPENIB_HP_QP or BTL_OPENIB_LP_QP)
|
|
|
|
* @return OMPI_SUCCESS or failure status
|
|
|
|
*/
|
|
|
|
|
2006-09-07 17:31:50 +04:00
|
|
|
static inline int mca_btl_openib_post_srr(mca_btl_openib_module_t* openib_btl,
|
2007-04-27 01:03:38 +04:00
|
|
|
const int additional,
|
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
|
|
|
const int qp)
|
2006-09-07 17:05:41 +04:00
|
|
|
{
|
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
|
|
|
assert(MCA_BTL_OPENIB_SRQ_QP == openib_btl->qps[qp].type);
|
2006-09-07 17:05:41 +04:00
|
|
|
OPAL_THREAD_LOCK(&openib_btl->ib_lock);
|
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(openib_btl->qps[qp].u.srq_qp.rd_posted <=
|
|
|
|
mca_btl_openib_component.qp_infos[qp].rd_low + additional &&
|
|
|
|
openib_btl->qps[qp].u.srq_qp.rd_posted <
|
|
|
|
mca_btl_openib_component.qp_infos[qp].rd_num) {
|
2007-04-27 01:03:38 +04:00
|
|
|
int rc;
|
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
|
|
|
int32_t i, num_post = mca_btl_openib_component.qp_infos[qp].rd_num -
|
|
|
|
openib_btl->qps[qp].u.srq_qp.rd_posted;
|
2006-09-07 17:05:41 +04:00
|
|
|
struct ibv_recv_wr *bad_wr;
|
|
|
|
ompi_free_list_t *free_list;
|
|
|
|
|
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_list = &openib_btl->qps[qp].recv_free;
|
2006-09-07 17:05:41 +04:00
|
|
|
|
|
|
|
for(i = 0; i < num_post; i++) {
|
2007-04-27 01:03:38 +04:00
|
|
|
ompi_free_list_item_t* item;
|
|
|
|
mca_btl_openib_frag_t* frag;
|
2006-09-07 17:05:41 +04:00
|
|
|
OMPI_FREE_LIST_WAIT(free_list, item, rc);
|
|
|
|
frag = (mca_btl_openib_frag_t*)item;
|
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
|
|
|
frag->base.order = qp;
|
|
|
|
if(ibv_post_srq_recv(openib_btl->qps[qp].u.srq_qp.srq,
|
|
|
|
&frag->wr_desc.rd_desc,
|
|
|
|
&bad_wr)) {
|
2006-09-07 17:05:41 +04:00
|
|
|
BTL_ERROR(("error posting receive descriptors to shared "
|
2007-04-27 01:03:38 +04:00
|
|
|
"receive queue: %s", strerror(errno)));
|
2007-04-26 17:33:02 +04:00
|
|
|
OPAL_THREAD_UNLOCK(&openib_btl->ib_lock);
|
2006-09-07 17:05:41 +04:00
|
|
|
return OMPI_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
|
|
|
OPAL_THREAD_ADD32(&openib_btl->qps[qp].u.srq_qp.rd_posted, num_post);
|
2006-09-07 17:05:41 +04:00
|
|
|
}
|
|
|
|
OPAL_THREAD_UNLOCK(&openib_btl->ib_lock);
|
2006-09-07 17:31:50 +04:00
|
|
|
|
|
|
|
return OMPI_SUCCESS;
|
2006-09-07 17:05:41 +04:00
|
|
|
}
|
|
|
|
|
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 inline int mca_btl_openib_post_srr_all(mca_btl_openib_module_t *openib_btl,
|
|
|
|
const int additional)
|
|
|
|
{
|
|
|
|
int qp;
|
|
|
|
for(qp = 0; qp < mca_btl_openib_component.num_srq_qps; qp++){
|
|
|
|
if(MCA_BTL_OPENIB_SRQ_QP == openib_btl->qps[qp].type) {
|
|
|
|
mca_btl_openib_post_srr(openib_btl, additional, qp);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
return OMPI_SUCCESS;
|
|
|
|
}
|
|
|
|
|
|
|
|
#define BTL_OPENIB_EAGER_RDMA_QP(QP) \
|
|
|
|
((QP) == mca_btl_openib_component.eager_rdma_qp)
|
|
|
|
|
|
|
|
#define BTL_OPENIB_RDMA_QP(QP) \
|
|
|
|
((QP) == mca_btl_openib_component.rdma_qp)
|
|
|
|
|
|
|
|
#if defined(c_plusplus) || defined(__cplusplus)
|
|
|
|
}
|
|
|
|
#endif
|
2007-06-14 05:59:25 +04:00
|
|
|
END_C_DECLS
|
|
|
|
|
2007-04-27 01:03:38 +04:00
|
|
|
#endif /* MCA_BTL_IB_H */
|