ports/openmpi
ports/openmpi
1
1
Форкнуть 0
Код Релизы Активность
fc686f5538
openmpi/opal/mca/btl/usnic/btl_usnic_ack.c

296 строки
9.2 KiB
C
Исходник Обычный вид История

First commit of the Cisco usNIC BTL. This BTL accesses the Cisco usNIC Linux device via the Linux verbs API via Unreliable Datagram queue pairs. A few noteworthy points: * This BTL does most of its own fragmentation; it tells the PML that it has a very high max_send_size (much higher than the network MTU). * Since UD fragments are, by definition, unreliable, the usnic BTL handles all of its own reliability via a sliding window approach using the opal_hotel construct and many tricks stolen from the corpus of knowledge surrounding efficient TCP. * There is a fun PML latency-metric based optimization for NUMA awareness of short messages. * Note that this is ''not'' a generic UD verbs BTL; it is specific to the Cisco usNIC device. This commit was SVN r28879.
2013-07-20 02:13:58 +04:00
/*
usnic: update to BTL 3.0
2015-02-07 01:59:45 +03:00
* Copyright (c) 2013-2015 Cisco Systems, Inc. All rights reserved.
First commit of the Cisco usNIC BTL. This BTL accesses the Cisco usNIC Linux device via the Linux verbs API via Unreliable Datagram queue pairs. A few noteworthy points: * This BTL does most of its own fragmentation; it tells the PML that it has a very high max_send_size (much higher than the network MTU). * Since UD fragments are, by definition, unreliable, the usnic BTL handles all of its own reliability via a sliding window approach using the opal_hotel construct and many tricks stolen from the corpus of knowledge surrounding efficient TCP. * There is a fun PML latency-metric based optimization for NUMA awareness of short messages. * Note that this is ''not'' a generic UD verbs BTL; it is specific to the Cisco usNIC device. This commit was SVN r28879.
2013-07-20 02:13:58 +04:00
* $COPYRIGHT$
*
* Additional copyrights may follow
*
* $HEADER$
*/
George did the work and deserves all the credit for it. Ralph did the merge, and deserves whatever blame results from errors in it :-) WHAT: Open our low-level communication infrastructure by moving all necessary components (btl/rcache/allocator/mpool) down in OPAL All the components required for inter-process communications are currently deeply integrated in the OMPI layer. Several groups/institutions have express interest in having a more generic communication infrastructure, without all the OMPI layer dependencies. This communication layer should be made available at a different software level, available to all layers in the Open MPI software stack. As an example, our ORTE layer could replace the current OOB and instead use the BTL directly, gaining access to more reactive network interfaces than TCP. Similarly, external software libraries could take advantage of our highly optimized AM (active message) communication layer for their own purpose. UTK with support from Sandia, developped a version of Open MPI where the entire communication infrastucture has been moved down to OPAL (btl/rcache/allocator/mpool). Most of the moved components have been updated to match the new schema, with few exceptions (mainly BTLs where I have no way of compiling/testing them). Thus, the completion of this RFC is tied to being able to completing this move for all BTLs. For this we need help from the rest of the Open MPI community, especially those supporting some of the BTLs. A non-exhaustive list of BTLs that qualify here is: mx, portals4, scif, udapl, ugni, usnic. This commit was SVN r32317.
2014-07-26 04:47:28 +04:00
#include "opal_config.h"
First commit of the Cisco usNIC BTL. This BTL accesses the Cisco usNIC Linux device via the Linux verbs API via Unreliable Datagram queue pairs. A few noteworthy points: * This BTL does most of its own fragmentation; it tells the PML that it has a very high max_send_size (much higher than the network MTU). * Since UD fragments are, by definition, unreliable, the usnic BTL handles all of its own reliability via a sliding window approach using the opal_hotel construct and many tricks stolen from the corpus of knowledge surrounding efficient TCP. * There is a fun PML latency-metric based optimization for NUMA awareness of short messages. * Note that this is ''not'' a generic UD verbs BTL; it is specific to the Cisco usNIC device. This commit was SVN r28879.
2013-07-20 02:13:58 +04:00
#include <errno.h>
#include <string.h>
#include <unistd.h>
#include "opal/util/output.h"
#include "opal/class/opal_hotel.h"
#include "btl_usnic.h"
#include "btl_usnic_frag.h"
#include "btl_usnic_endpoint.h"
#include "btl_usnic_module.h"
#include "btl_usnic_ack.h"
#include "btl_usnic_util.h"
#include "btl_usnic_send.h"
usnic: check connectivity on first communication to a peer Previously, we were only checking connectivity upon first ''send'' to a peer. But this ignores the case where the first communication to a peer is actually an ACK -- i.e., we successfully received something from the peer and we need to send an ACK back. So we need to verify that the ACK will actually get there. Specifically, certain asymmetric routing cases can lead to a hang if we don't check the connectivity in both directions. E.g., if the sender is able to get traffic to the receiver, but the receiver is unable to get traffic back to the sender because it made a different routing decision than the sender. In this case, the connectivity checker from the sender could succeed (because the connectivity checker will ACK along the same path in which the ping was received), but sending a BTL ACK could fail (because the BTL ACK will be sent back along the path chosen by the graph algorithm, which, in an erroneous asymmetric routing scenario, may be different/wrong). Hence, we want to trigger the connectivity checker at the first communication from A->B, which may either be a BTL send or an ACK. Reviewed by Dave Goodell. cmr=v1.8.2:reviewer=ompi-rm1.8 This commit was SVN r32309.
2014-07-25 01:32:56 +04:00
#include "btl_usnic_connectivity.h"
First commit of the Cisco usNIC BTL. This BTL accesses the Cisco usNIC Linux device via the Linux verbs API via Unreliable Datagram queue pairs. A few noteworthy points: * This BTL does most of its own fragmentation; it tells the PML that it has a very high max_send_size (much higher than the network MTU). * Since UD fragments are, by definition, unreliable, the usnic BTL handles all of its own reliability via a sliding window approach using the opal_hotel construct and many tricks stolen from the corpus of knowledge surrounding efficient TCP. * There is a fun PML latency-metric based optimization for NUMA awareness of short messages. * Note that this is ''not'' a generic UD verbs BTL; it is specific to the Cisco usNIC device. This commit was SVN r28879.
2013-07-20 02:13:58 +04:00
/*
* Force a retrans of a segment
*/
static void
George did the work and deserves all the credit for it. Ralph did the merge, and deserves whatever blame results from errors in it :-) WHAT: Open our low-level communication infrastructure by moving all necessary components (btl/rcache/allocator/mpool) down in OPAL All the components required for inter-process communications are currently deeply integrated in the OMPI layer. Several groups/institutions have express interest in having a more generic communication infrastructure, without all the OMPI layer dependencies. This communication layer should be made available at a different software level, available to all layers in the Open MPI software stack. As an example, our ORTE layer could replace the current OOB and instead use the BTL directly, gaining access to more reactive network interfaces than TCP. Similarly, external software libraries could take advantage of our highly optimized AM (active message) communication layer for their own purpose. UTK with support from Sandia, developped a version of Open MPI where the entire communication infrastucture has been moved down to OPAL (btl/rcache/allocator/mpool). Most of the moved components have been updated to match the new schema, with few exceptions (mainly BTLs where I have no way of compiling/testing them). Thus, the completion of this RFC is tied to being able to completing this move for all BTLs. For this we need help from the rest of the Open MPI community, especially those supporting some of the BTLs. A non-exhaustive list of BTLs that qualify here is: mx, portals4, scif, udapl, ugni, usnic. This commit was SVN r32317.
2014-07-26 04:47:28 +04:00
opal_btl_usnic_force_retrans(
opal_btl_usnic_endpoint_t *endpoint,
opal_btl_usnic_seq_t ack_seq)
First commit of the Cisco usNIC BTL. This BTL accesses the Cisco usNIC Linux device via the Linux verbs API via Unreliable Datagram queue pairs. A few noteworthy points: * This BTL does most of its own fragmentation; it tells the PML that it has a very high max_send_size (much higher than the network MTU). * Since UD fragments are, by definition, unreliable, the usnic BTL handles all of its own reliability via a sliding window approach using the opal_hotel construct and many tricks stolen from the corpus of knowledge surrounding efficient TCP. * There is a fun PML latency-metric based optimization for NUMA awareness of short messages. * Note that this is ''not'' a generic UD verbs BTL; it is specific to the Cisco usNIC device. This commit was SVN r28879.
2013-07-20 02:13:58 +04:00
{
George did the work and deserves all the credit for it. Ralph did the merge, and deserves whatever blame results from errors in it :-) WHAT: Open our low-level communication infrastructure by moving all necessary components (btl/rcache/allocator/mpool) down in OPAL All the components required for inter-process communications are currently deeply integrated in the OMPI layer. Several groups/institutions have express interest in having a more generic communication infrastructure, without all the OMPI layer dependencies. This communication layer should be made available at a different software level, available to all layers in the Open MPI software stack. As an example, our ORTE layer could replace the current OOB and instead use the BTL directly, gaining access to more reactive network interfaces than TCP. Similarly, external software libraries could take advantage of our highly optimized AM (active message) communication layer for their own purpose. UTK with support from Sandia, developped a version of Open MPI where the entire communication infrastucture has been moved down to OPAL (btl/rcache/allocator/mpool). Most of the moved components have been updated to match the new schema, with few exceptions (mainly BTLs where I have no way of compiling/testing them). Thus, the completion of this RFC is tied to being able to completing this move for all BTLs. For this we need help from the rest of the Open MPI community, especially those supporting some of the BTLs. A non-exhaustive list of BTLs that qualify here is: mx, portals4, scif, udapl, ugni, usnic. This commit was SVN r32317.
2014-07-26 04:47:28 +04:00
opal_btl_usnic_send_segment_t *sseg;
First commit of the Cisco usNIC BTL. This BTL accesses the Cisco usNIC Linux device via the Linux verbs API via Unreliable Datagram queue pairs. A few noteworthy points: * This BTL does most of its own fragmentation; it tells the PML that it has a very high max_send_size (much higher than the network MTU). * Since UD fragments are, by definition, unreliable, the usnic BTL handles all of its own reliability via a sliding window approach using the opal_hotel construct and many tricks stolen from the corpus of knowledge surrounding efficient TCP. * There is a fun PML latency-metric based optimization for NUMA awareness of short messages. * Note that this is ''not'' a generic UD verbs BTL; it is specific to the Cisco usNIC device. This commit was SVN r28879.
2013-07-20 02:13:58 +04:00
int is;
is = WINDOW_SIZE_MOD(ack_seq+1);
sseg = endpoint->endpoint_sent_segs[is];
if (sseg == NULL || sseg->ss_hotel_room == -1) {
return;
}
/* cancel retrans timer */
opal_hotel_checkout(&endpoint->endpoint_hotel, sseg->ss_hotel_room);
sseg->ss_hotel_room = -1;
usnic: properly destroy embedded small send frag Without this, an `--enable-debug` build would hit an assertion in the list code when run under valgrind with `--malloc-fill=0xff` or any other case where malloc returned non-zeroed buffers. Also allow the normal OBJ_ machinery to handle the constructor invocation ordering for us instead of doing it by hand (which could have led to future bugs). Reviewed-by: jsquyres@cisco.com cmr=v1.7.4 Depends on trunk functionality in r29095 and r29096. Refs trac:3740,#3741. This commit was SVN r29127. The following SVN revision numbers were found above: r29095 --> open-mpi/ompi@d1b5940e97ebb18535dce75e0c2b97d869a18d73 r29096 --> open-mpi/ompi@a552921171e596c500551c63e8fc6d00c8b2dcd5 The following Trac tickets were found above: Ticket 3740 --> https://svn.open-mpi.org/trac/ompi/ticket/3740
2013-09-05 00:59:12 +04:00
/* Queue up this segment to be resent */
First commit of the Cisco usNIC BTL. This BTL accesses the Cisco usNIC Linux device via the Linux verbs API via Unreliable Datagram queue pairs. A few noteworthy points: * This BTL does most of its own fragmentation; it tells the PML that it has a very high max_send_size (much higher than the network MTU). * Since UD fragments are, by definition, unreliable, the usnic BTL handles all of its own reliability via a sliding window approach using the opal_hotel construct and many tricks stolen from the corpus of knowledge surrounding efficient TCP. * There is a fun PML latency-metric based optimization for NUMA awareness of short messages. * Note that this is ''not'' a generic UD verbs BTL; it is specific to the Cisco usNIC device. This commit was SVN r28879.
2013-07-20 02:13:58 +04:00
opal_list_append(&(endpoint->endpoint_module->pending_resend_segs),
&(sseg->ss_base.us_list.super));
Move all usNIC stats to _stats.c|h and export them as MPI_T pvars. This commit moves all the module stats into their own struct so that the stats only need to appear as a single line in the module_t definition, and then moves all the logic for reporting the stats into btl_usnic_stats.c|h. Further, the stats are now exported as MPI_T_BIND_NO_OBJECT entities (i.e., not bound to any particular MPI handle), and are marked as READONLY and CONTINUOUS. They currently all default to verbose level 5 ("Application tuner / detailed", according to https://svn.open-mpi.org/trac/ompi/wiki/MCAParamLevels). Most of the statistics are counters, but a small number are high watermark values. Due to how counters are reported via MPI_T, none of the counters are exported through MPI_T if the MCA param btl_usnic_stats_relative=1 (i.e., the module resets the stats back to zero at a given frequency). When MPI_T_pvar_handle_alloc() is invoked on any of these pvars, it will return a count that is equal to the number of active usnic BTL modules. The values returned for any given pvar (e.g., num_total_sends) are an array containing one value for each active usnic BTL module. The ordering of values in the array is both consistent across all usnic pvars and stable throughout a single job: array slot 0 corresponds to module X, array slot 1 corresponds to module Y, etc. Mapping which array slot corresponds to which underlying Linux usnic_X device works as follows: * The btl_usnic_devices MPI_T state pvar is associated with a btl_usnic_device MPI_T enum, and be obtained via MPI_T_pvar_get_info(). * If all usNIC pvars are of length N, the values [0,N) in the btl_usnic_device enum are associated with strings of the corresponding underlying Linux device. For exampe, to look up which Linux device is reported in all usNIC pvars' array slot 1, look up the int value 1 in the btl_usnic_devices enum. Its corresponding string value is underlying Linux device name (e.g., "usnic_1"). cmr=v1.7.4:subject="usnic BTL MPI_T pvars" This commit was SVN r29545.
2013-10-29 02:23:08 +04:00
++endpoint->endpoint_module->stats.num_fast_retrans;
First commit of the Cisco usNIC BTL. This BTL accesses the Cisco usNIC Linux device via the Linux verbs API via Unreliable Datagram queue pairs. A few noteworthy points: * This BTL does most of its own fragmentation; it tells the PML that it has a very high max_send_size (much higher than the network MTU). * Since UD fragments are, by definition, unreliable, the usnic BTL handles all of its own reliability via a sliding window approach using the opal_hotel construct and many tricks stolen from the corpus of knowledge surrounding efficient TCP. * There is a fun PML latency-metric based optimization for NUMA awareness of short messages. * Note that this is ''not'' a generic UD verbs BTL; it is specific to the Cisco usNIC device. This commit was SVN r28879.
2013-07-20 02:13:58 +04:00
}
/*
* We have received an ACK for a given sequence number (either standalone
* or via piggy-back on a regular send)
*/
void
George did the work and deserves all the credit for it. Ralph did the merge, and deserves whatever blame results from errors in it :-) WHAT: Open our low-level communication infrastructure by moving all necessary components (btl/rcache/allocator/mpool) down in OPAL All the components required for inter-process communications are currently deeply integrated in the OMPI layer. Several groups/institutions have express interest in having a more generic communication infrastructure, without all the OMPI layer dependencies. This communication layer should be made available at a different software level, available to all layers in the Open MPI software stack. As an example, our ORTE layer could replace the current OOB and instead use the BTL directly, gaining access to more reactive network interfaces than TCP. Similarly, external software libraries could take advantage of our highly optimized AM (active message) communication layer for their own purpose. UTK with support from Sandia, developped a version of Open MPI where the entire communication infrastucture has been moved down to OPAL (btl/rcache/allocator/mpool). Most of the moved components have been updated to match the new schema, with few exceptions (mainly BTLs where I have no way of compiling/testing them). Thus, the completion of this RFC is tied to being able to completing this move for all BTLs. For this we need help from the rest of the Open MPI community, especially those supporting some of the BTLs. A non-exhaustive list of BTLs that qualify here is: mx, portals4, scif, udapl, ugni, usnic. This commit was SVN r32317.
2014-07-26 04:47:28 +04:00
opal_btl_usnic_handle_ack(
opal_btl_usnic_endpoint_t *endpoint,
opal_btl_usnic_seq_t ack_seq)
First commit of the Cisco usNIC BTL. This BTL accesses the Cisco usNIC Linux device via the Linux verbs API via Unreliable Datagram queue pairs. A few noteworthy points: * This BTL does most of its own fragmentation; it tells the PML that it has a very high max_send_size (much higher than the network MTU). * Since UD fragments are, by definition, unreliable, the usnic BTL handles all of its own reliability via a sliding window approach using the opal_hotel construct and many tricks stolen from the corpus of knowledge surrounding efficient TCP. * There is a fun PML latency-metric based optimization for NUMA awareness of short messages. * Note that this is ''not'' a generic UD verbs BTL; it is specific to the Cisco usNIC device. This commit was SVN r28879.
2013-07-20 02:13:58 +04:00
{
George did the work and deserves all the credit for it. Ralph did the merge, and deserves whatever blame results from errors in it :-) WHAT: Open our low-level communication infrastructure by moving all necessary components (btl/rcache/allocator/mpool) down in OPAL All the components required for inter-process communications are currently deeply integrated in the OMPI layer. Several groups/institutions have express interest in having a more generic communication infrastructure, without all the OMPI layer dependencies. This communication layer should be made available at a different software level, available to all layers in the Open MPI software stack. As an example, our ORTE layer could replace the current OOB and instead use the BTL directly, gaining access to more reactive network interfaces than TCP. Similarly, external software libraries could take advantage of our highly optimized AM (active message) communication layer for their own purpose. UTK with support from Sandia, developped a version of Open MPI where the entire communication infrastucture has been moved down to OPAL (btl/rcache/allocator/mpool). Most of the moved components have been updated to match the new schema, with few exceptions (mainly BTLs where I have no way of compiling/testing them). Thus, the completion of this RFC is tied to being able to completing this move for all BTLs. For this we need help from the rest of the Open MPI community, especially those supporting some of the BTLs. A non-exhaustive list of BTLs that qualify here is: mx, portals4, scif, udapl, ugni, usnic. This commit was SVN r32317.
2014-07-26 04:47:28 +04:00
opal_btl_usnic_seq_t is;
opal_btl_usnic_send_segment_t *sseg;
opal_btl_usnic_send_frag_t *frag;
opal_btl_usnic_module_t *module;
First commit of the Cisco usNIC BTL. This BTL accesses the Cisco usNIC Linux device via the Linux verbs API via Unreliable Datagram queue pairs. A few noteworthy points: * This BTL does most of its own fragmentation; it tells the PML that it has a very high max_send_size (much higher than the network MTU). * Since UD fragments are, by definition, unreliable, the usnic BTL handles all of its own reliability via a sliding window approach using the opal_hotel construct and many tricks stolen from the corpus of knowledge surrounding efficient TCP. * There is a fun PML latency-metric based optimization for NUMA awareness of short messages. * Note that this is ''not'' a generic UD verbs BTL; it is specific to the Cisco usNIC device. This commit was SVN r28879.
2013-07-20 02:13:58 +04:00
uint32_t bytes_acked;
module = endpoint->endpoint_module;
/* ignore if this is an old ACK */
usnic: Shrink sequence numbers to 16 bits Authored-by: Reese Faucette <rfaucett@cisco.com> Reviewed-by: Jeff Squyres <jsquyres@cisco.com> cmr=v1.7.5:ticket=trac:4253 This commit was SVN r30834. The following Trac tickets were found above: Ticket 4253 --> https://svn.open-mpi.org/trac/ompi/ticket/4253
2014-02-26 11:40:10 +04:00
if (SEQ_LT(ack_seq, endpoint->endpoint_ack_seq_rcvd)) {
Clean up debugging logging a bit. MSGDEBUG2 now means "print a one-liner for all PML calls into BTL, and also when BTL calls PML with a recv completion (not send completions)" MSGDEBUG1 means print more internal gory detail MSGDEBUG is gone, replaced by MSGDEBUG1 In the process also found that PUT_DEST style fragments could potentially be leaked in usnic_free() since send_fragment tests were being applied to see if it was eligible to be freed. This commit was SVN r29185.
2013-09-17 11:29:40 +04:00
#if MSGDEBUG1
opal_output(0, "Got OLD DUP ACK seq %"UDSEQ" < %"UDSEQ"\n",
First commit of the Cisco usNIC BTL. This BTL accesses the Cisco usNIC Linux device via the Linux verbs API via Unreliable Datagram queue pairs. A few noteworthy points: * This BTL does most of its own fragmentation; it tells the PML that it has a very high max_send_size (much higher than the network MTU). * Since UD fragments are, by definition, unreliable, the usnic BTL handles all of its own reliability via a sliding window approach using the opal_hotel construct and many tricks stolen from the corpus of knowledge surrounding efficient TCP. * There is a fun PML latency-metric based optimization for NUMA awareness of short messages. * Note that this is ''not'' a generic UD verbs BTL; it is specific to the Cisco usNIC device. This commit was SVN r28879.
2013-07-20 02:13:58 +04:00
ack_seq, endpoint->endpoint_ack_seq_rcvd);
#endif
Move all usNIC stats to _stats.c|h and export them as MPI_T pvars. This commit moves all the module stats into their own struct so that the stats only need to appear as a single line in the module_t definition, and then moves all the logic for reporting the stats into btl_usnic_stats.c|h. Further, the stats are now exported as MPI_T_BIND_NO_OBJECT entities (i.e., not bound to any particular MPI handle), and are marked as READONLY and CONTINUOUS. They currently all default to verbose level 5 ("Application tuner / detailed", according to https://svn.open-mpi.org/trac/ompi/wiki/MCAParamLevels). Most of the statistics are counters, but a small number are high watermark values. Due to how counters are reported via MPI_T, none of the counters are exported through MPI_T if the MCA param btl_usnic_stats_relative=1 (i.e., the module resets the stats back to zero at a given frequency). When MPI_T_pvar_handle_alloc() is invoked on any of these pvars, it will return a count that is equal to the number of active usnic BTL modules. The values returned for any given pvar (e.g., num_total_sends) are an array containing one value for each active usnic BTL module. The ordering of values in the array is both consistent across all usnic pvars and stable throughout a single job: array slot 0 corresponds to module X, array slot 1 corresponds to module Y, etc. Mapping which array slot corresponds to which underlying Linux usnic_X device works as follows: * The btl_usnic_devices MPI_T state pvar is associated with a btl_usnic_device MPI_T enum, and be obtained via MPI_T_pvar_get_info(). * If all usNIC pvars are of length N, the values [0,N) in the btl_usnic_device enum are associated with strings of the corresponding underlying Linux device. For exampe, to look up which Linux device is reported in all usNIC pvars' array slot 1, look up the int value 1 in the btl_usnic_devices enum. Its corresponding string value is underlying Linux device name (e.g., "usnic_1"). cmr=v1.7.4:subject="usnic BTL MPI_T pvars" This commit was SVN r29545.
2013-10-29 02:23:08 +04:00
++module->stats.num_old_dup_acks;
First commit of the Cisco usNIC BTL. This BTL accesses the Cisco usNIC Linux device via the Linux verbs API via Unreliable Datagram queue pairs. A few noteworthy points: * This BTL does most of its own fragmentation; it tells the PML that it has a very high max_send_size (much higher than the network MTU). * Since UD fragments are, by definition, unreliable, the usnic BTL handles all of its own reliability via a sliding window approach using the opal_hotel construct and many tricks stolen from the corpus of knowledge surrounding efficient TCP. * There is a fun PML latency-metric based optimization for NUMA awareness of short messages. * Note that this is ''not'' a generic UD verbs BTL; it is specific to the Cisco usNIC device. This commit was SVN r28879.
2013-07-20 02:13:58 +04:00
return;
/* A duplicate ACK means next seg was lost */
} else if (ack_seq == endpoint->endpoint_ack_seq_rcvd) {
Move all usNIC stats to _stats.c|h and export them as MPI_T pvars. This commit moves all the module stats into their own struct so that the stats only need to appear as a single line in the module_t definition, and then moves all the logic for reporting the stats into btl_usnic_stats.c|h. Further, the stats are now exported as MPI_T_BIND_NO_OBJECT entities (i.e., not bound to any particular MPI handle), and are marked as READONLY and CONTINUOUS. They currently all default to verbose level 5 ("Application tuner / detailed", according to https://svn.open-mpi.org/trac/ompi/wiki/MCAParamLevels). Most of the statistics are counters, but a small number are high watermark values. Due to how counters are reported via MPI_T, none of the counters are exported through MPI_T if the MCA param btl_usnic_stats_relative=1 (i.e., the module resets the stats back to zero at a given frequency). When MPI_T_pvar_handle_alloc() is invoked on any of these pvars, it will return a count that is equal to the number of active usnic BTL modules. The values returned for any given pvar (e.g., num_total_sends) are an array containing one value for each active usnic BTL module. The ordering of values in the array is both consistent across all usnic pvars and stable throughout a single job: array slot 0 corresponds to module X, array slot 1 corresponds to module Y, etc. Mapping which array slot corresponds to which underlying Linux usnic_X device works as follows: * The btl_usnic_devices MPI_T state pvar is associated with a btl_usnic_device MPI_T enum, and be obtained via MPI_T_pvar_get_info(). * If all usNIC pvars are of length N, the values [0,N) in the btl_usnic_device enum are associated with strings of the corresponding underlying Linux device. For exampe, to look up which Linux device is reported in all usNIC pvars' array slot 1, look up the int value 1 in the btl_usnic_devices enum. Its corresponding string value is underlying Linux device name (e.g., "usnic_1"). cmr=v1.7.4:subject="usnic BTL MPI_T pvars" This commit was SVN r29545.
2013-10-29 02:23:08 +04:00
++module->stats.num_dup_acks;
First commit of the Cisco usNIC BTL. This BTL accesses the Cisco usNIC Linux device via the Linux verbs API via Unreliable Datagram queue pairs. A few noteworthy points: * This BTL does most of its own fragmentation; it tells the PML that it has a very high max_send_size (much higher than the network MTU). * Since UD fragments are, by definition, unreliable, the usnic BTL handles all of its own reliability via a sliding window approach using the opal_hotel construct and many tricks stolen from the corpus of knowledge surrounding efficient TCP. * There is a fun PML latency-metric based optimization for NUMA awareness of short messages. * Note that this is ''not'' a generic UD verbs BTL; it is specific to the Cisco usNIC device. This commit was SVN r28879.
2013-07-20 02:13:58 +04:00
George did the work and deserves all the credit for it. Ralph did the merge, and deserves whatever blame results from errors in it :-) WHAT: Open our low-level communication infrastructure by moving all necessary components (btl/rcache/allocator/mpool) down in OPAL All the components required for inter-process communications are currently deeply integrated in the OMPI layer. Several groups/institutions have express interest in having a more generic communication infrastructure, without all the OMPI layer dependencies. This communication layer should be made available at a different software level, available to all layers in the Open MPI software stack. As an example, our ORTE layer could replace the current OOB and instead use the BTL directly, gaining access to more reactive network interfaces than TCP. Similarly, external software libraries could take advantage of our highly optimized AM (active message) communication layer for their own purpose. UTK with support from Sandia, developped a version of Open MPI where the entire communication infrastucture has been moved down to OPAL (btl/rcache/allocator/mpool). Most of the moved components have been updated to match the new schema, with few exceptions (mainly BTLs where I have no way of compiling/testing them). Thus, the completion of this RFC is tied to being able to completing this move for all BTLs. For this we need help from the rest of the Open MPI community, especially those supporting some of the BTLs. A non-exhaustive list of BTLs that qualify here is: mx, portals4, scif, udapl, ugni, usnic. This commit was SVN r32317.
2014-07-26 04:47:28 +04:00
opal_btl_usnic_force_retrans(endpoint, ack_seq);
First commit of the Cisco usNIC BTL. This BTL accesses the Cisco usNIC Linux device via the Linux verbs API via Unreliable Datagram queue pairs. A few noteworthy points: * This BTL does most of its own fragmentation; it tells the PML that it has a very high max_send_size (much higher than the network MTU). * Since UD fragments are, by definition, unreliable, the usnic BTL handles all of its own reliability via a sliding window approach using the opal_hotel construct and many tricks stolen from the corpus of knowledge surrounding efficient TCP. * There is a fun PML latency-metric based optimization for NUMA awareness of short messages. * Note that this is ''not'' a generic UD verbs BTL; it is specific to the Cisco usNIC device. This commit was SVN r28879.
2013-07-20 02:13:58 +04:00
return;
}
/* Does this ACK have a new sequence number that we haven't
seen before? */
usnic: Shrink sequence numbers to 16 bits Authored-by: Reese Faucette <rfaucett@cisco.com> Reviewed-by: Jeff Squyres <jsquyres@cisco.com> cmr=v1.7.5:ticket=trac:4253 This commit was SVN r30834. The following Trac tickets were found above: Ticket 4253 --> https://svn.open-mpi.org/trac/ompi/ticket/4253
2014-02-26 11:40:10 +04:00
for (is = endpoint->endpoint_ack_seq_rcvd + 1; SEQ_LE(is, ack_seq); ++is) {
First commit of the Cisco usNIC BTL. This BTL accesses the Cisco usNIC Linux device via the Linux verbs API via Unreliable Datagram queue pairs. A few noteworthy points: * This BTL does most of its own fragmentation; it tells the PML that it has a very high max_send_size (much higher than the network MTU). * Since UD fragments are, by definition, unreliable, the usnic BTL handles all of its own reliability via a sliding window approach using the opal_hotel construct and many tricks stolen from the corpus of knowledge surrounding efficient TCP. * There is a fun PML latency-metric based optimization for NUMA awareness of short messages. * Note that this is ''not'' a generic UD verbs BTL; it is specific to the Cisco usNIC device. This commit was SVN r28879.
2013-07-20 02:13:58 +04:00
sseg = endpoint->endpoint_sent_segs[WINDOW_SIZE_MOD(is)];
#if MSGDEBUG1
opal_output(0, " Checking ACK/sent_segs window %p, index %lu, seq %lu, occupied=%p, seg_room=%d",
(void*) endpoint->endpoint_sent_segs,
usnic: stomp format specifier warnings The usnic BTL now builds cleanly under `--enable-picky` when `MSGDEBUG1` is set. Reviewed-by: jsquyres cmr=v1.7.4:reviewer=jsquyres This commit was SVN r29097.
2013-08-30 03:24:14 +04:00
WINDOW_SIZE_MOD(is), is, (void*)sseg, (sseg?sseg->ss_hotel_room:-2));
First commit of the Cisco usNIC BTL. This BTL accesses the Cisco usNIC Linux device via the Linux verbs API via Unreliable Datagram queue pairs. A few noteworthy points: * This BTL does most of its own fragmentation; it tells the PML that it has a very high max_send_size (much higher than the network MTU). * Since UD fragments are, by definition, unreliable, the usnic BTL handles all of its own reliability via a sliding window approach using the opal_hotel construct and many tricks stolen from the corpus of knowledge surrounding efficient TCP. * There is a fun PML latency-metric based optimization for NUMA awareness of short messages. * Note that this is ''not'' a generic UD verbs BTL; it is specific to the Cisco usNIC device. This commit was SVN r28879.
2013-07-20 02:13:58 +04:00
#endif
assert(sseg != NULL);
usnic: Shrink sequence numbers to 16 bits Authored-by: Reese Faucette <rfaucett@cisco.com> Reviewed-by: Jeff Squyres <jsquyres@cisco.com> cmr=v1.7.5:ticket=trac:4253 This commit was SVN r30834. The following Trac tickets were found above: Ticket 4253 --> https://svn.open-mpi.org/trac/ompi/ticket/4253
2014-02-26 11:40:10 +04:00
assert(sseg->ss_base.us_btl_header->pkt_seq == is);
Clean up debugging logging a bit. MSGDEBUG2 now means "print a one-liner for all PML calls into BTL, and also when BTL calls PML with a recv completion (not send completions)" MSGDEBUG1 means print more internal gory detail MSGDEBUG is gone, replaced by MSGDEBUG1 In the process also found that PUT_DEST style fragments could potentially be leaked in usnic_free() since send_fragment tests were being applied to see if it was eligible to be freed. This commit was SVN r29185.
2013-09-17 11:29:40 +04:00
#if MSGDEBUG1
First commit of the Cisco usNIC BTL. This BTL accesses the Cisco usNIC Linux device via the Linux verbs API via Unreliable Datagram queue pairs. A few noteworthy points: * This BTL does most of its own fragmentation; it tells the PML that it has a very high max_send_size (much higher than the network MTU). * Since UD fragments are, by definition, unreliable, the usnic BTL handles all of its own reliability via a sliding window approach using the opal_hotel construct and many tricks stolen from the corpus of knowledge surrounding efficient TCP. * There is a fun PML latency-metric based optimization for NUMA awareness of short messages. * Note that this is ''not'' a generic UD verbs BTL; it is specific to the Cisco usNIC device. This commit was SVN r28879.
2013-07-20 02:13:58 +04:00
if (sseg->ss_hotel_room == -1) {
opal_output(0, "=== ACKed frag in sent_frags array is not in hotel/enqueued, module %p, endpoint %p, seg %p, seq %" UDSEQ ", slot %lu",
(void*) module, (void*) endpoint,
(void*) sseg, is, WINDOW_SIZE_MOD(is));
}
#endif
/* Check the sending segment out from the hotel. NOTE: The
segment might not actually be in a hotel room if it has
already been evicted and queued for resend.
If it's not in the hotel, don't check it out! */
if (OPAL_LIKELY(sseg->ss_hotel_room != -1)) {
opal_hotel_checkout(&endpoint->endpoint_hotel, sseg->ss_hotel_room);
sseg->ss_hotel_room = -1;
/* hotel_room == -1 means queued for resend, remove it */
} else {
opal_list_remove_item((&module->pending_resend_segs),
&sseg->ss_base.us_list.super);
}
/* update the owning fragment */
bytes_acked = sseg->ss_base.us_btl_header->payload_len;
frag = sseg->ss_parent_frag;
#if MSGDEBUG1
usnic: pack via convertor on the fly If we need to use a convertor, go back to stashing that convertor in the frag and populating segments "on the fly" (in ompi_btl_usnic_module_progress_sends). Previously we would pack into a chain of chunk segments at prepare_src time, unnecessarily consuming additional memory. Reviewed-by: Jeff Squyres <jsquyres@cisco.com> Reviewed-by: Reese Faucette <rfaucett@cisco.com> This commit was SVN r29592.
2013-11-05 02:52:03 +04:00
opal_output(0, " ACKED seg %p frag %p ack_bytes=%"PRIu32" left=%zd dst_seg[0].seg_addr=%p des_flags=0x%x\n",
usnic: SEGV in OSU benchmarks Prevent frag from being freed out from under us in the case the PML callback routine calls usnic_free(). We accomplish this by delaying decrement of sf_bytes_to_ack until after the callback is performed, since sf_bytes_to_ack == 0 is condition of freeing the frag. Fixes Cisco bug CSCuj45094. Authored-by: Reese Faucette <rfaucett@cisco.com> cmr=v1.7.3 This commit was SVN r29264.
2013-09-27 01:48:04 +04:00
(void*)sseg, (void*)frag, bytes_acked,
usnic: convert from verbs to libfabric (yay!) This commit represents the conversion of the usnic BTL from verbs to libfabric. For the moment, libfabric is embedded in Open MPI (currently in the usnic BTL). This is because the libfabric API is still changing, and also has not yet been released. Ultimately, this embedded copy of libfabric will likely disappear and the usnic BTL will rely on an external installation of libfabric. New configure options: * --with-libfabric: will cause configure to fail if libfabric support cannot be built * --without-libfabric: will prevent libfabric support from being built * --with-libfabric=DIR: use an external libfabric installation * --with-libfabric-libdir=LIBDIR: when paired with --with-libfabric=DIR, use LIBDIR for the libfabric installation library dir The --with-libnl3[-libdir] arguments are now gone.
2014-12-03 00:09:46 +03:00
frag->sf_ack_bytes_left - bytes_acked,
frag->sf_base.uf_local_seg[0].seg_addr.pval,
usnic: pack via convertor on the fly If we need to use a convertor, go back to stashing that convertor in the frag and populating segments "on the fly" (in ompi_btl_usnic_module_progress_sends). Previously we would pack into a chain of chunk segments at prepare_src time, unnecessarily consuming additional memory. Reviewed-by: Jeff Squyres <jsquyres@cisco.com> Reviewed-by: Reese Faucette <rfaucett@cisco.com> This commit was SVN r29592.
2013-11-05 02:52:03 +04:00
frag->sf_base.uf_base.des_flags);
First commit of the Cisco usNIC BTL. This BTL accesses the Cisco usNIC Linux device via the Linux verbs API via Unreliable Datagram queue pairs. A few noteworthy points: * This BTL does most of its own fragmentation; it tells the PML that it has a very high max_send_size (much higher than the network MTU). * Since UD fragments are, by definition, unreliable, the usnic BTL handles all of its own reliability via a sliding window approach using the opal_hotel construct and many tricks stolen from the corpus of knowledge surrounding efficient TCP. * There is a fun PML latency-metric based optimization for NUMA awareness of short messages. * Note that this is ''not'' a generic UD verbs BTL; it is specific to the Cisco usNIC device. This commit was SVN r28879.
2013-07-20 02:13:58 +04:00
#endif
Cisco CSCuj12520: various problems running c_fence_put_1 - tag needs to be sent in *our* header, not the PML header - usnic_alloc() should return smaller value if too much data requested - be careful about callbacks vs removing items from lists (we need to remove from outr lists *before* the callback) - improve send callback handling - add some more MSGDEBUG2 logging and cleanup This commit was SVN r29181.
2013-09-17 11:20:44 +04:00
/* If all ACKs received, and this is a put or a regular send
* that needs a callback, perform the callback now
usnic: SEGV in OSU benchmarks Prevent frag from being freed out from under us in the case the PML callback routine calls usnic_free(). We accomplish this by delaying decrement of sf_bytes_to_ack until after the callback is performed, since sf_bytes_to_ack == 0 is condition of freeing the frag. Fixes Cisco bug CSCuj45094. Authored-by: Reese Faucette <rfaucett@cisco.com> cmr=v1.7.3 This commit was SVN r29264.
2013-09-27 01:48:04 +04:00
*
usnic: remove all trailing whitespace Style cleanup only; no code changes. This commit was SVN r32366.
2014-07-31 00:56:15 +04:00
* NOTE on sf_ack_bytes_left - here we check for
usnic: SEGV in OSU benchmarks Prevent frag from being freed out from under us in the case the PML callback routine calls usnic_free(). We accomplish this by delaying decrement of sf_bytes_to_ack until after the callback is performed, since sf_bytes_to_ack == 0 is condition of freeing the frag. Fixes Cisco bug CSCuj45094. Authored-by: Reese Faucette <rfaucett@cisco.com> cmr=v1.7.3 This commit was SVN r29264.
2013-09-27 01:48:04 +04:00
* sf_ack_bytes_left == bytes_acked
* as opposed to adjusting sf_ack_bytes_left and checking for 0 because
* if we don't, the callback function may call usnic_free() and free
* the fragment out from under us which we do not want. If the
* fragment really needs to be freed, we'll take care of it in a few
* lines below.
Cisco CSCuj12520: various problems running c_fence_put_1 - tag needs to be sent in *our* header, not the PML header - usnic_alloc() should return smaller value if too much data requested - be careful about callbacks vs removing items from lists (we need to remove from outr lists *before* the callback) - improve send callback handling - add some more MSGDEBUG2 logging and cleanup This commit was SVN r29181.
2013-09-17 11:20:44 +04:00
*/
usnic: update to BTL 3.0
2015-02-07 01:59:45 +03:00
if (frag->sf_ack_bytes_left == bytes_acked) {
usnic: restore compatibilty between master and v1.8 Add the functions that changed between BTL 2.0 and 3.0 into compat.h and compat.c: * module.btl_prepare_src: the signature and body of this method changed between 2.0 and 3.0. However, the functions that this method calls did *not* need to change, so they are copied over wholesale (with the exception that they no longer accept the unused `registration` parameter). * module.btl_prepare_dst: this method does not exist in BTL 3.0. * module.btl_put: the signature and body of this method changed between 2.0 and 3.0.
2015-02-07 04:41:50 +03:00
#if BTL_VERSION == 30
usnic: update to BTL 3.0
2015-02-07 01:59:45 +03:00
if (frag->sf_base.uf_remote_seg[0].seg_addr.pval != NULL) {
OPAL_BTL_USNIC_DO_PUT_FRAG_CB(module, frag, "put completion");
} else if (frag->sf_base.uf_base.des_flags &
MCA_BTL_DES_SEND_ALWAYS_CALLBACK) {
OPAL_BTL_USNIC_DO_SEND_FRAG_CB(module, frag, "send completion");
}
usnic: restore compatibilty between master and v1.8 Add the functions that changed between BTL 2.0 and 3.0 into compat.h and compat.c: * module.btl_prepare_src: the signature and body of this method changed between 2.0 and 3.0. However, the functions that this method calls did *not* need to change, so they are copied over wholesale (with the exception that they no longer accept the unused `registration` parameter). * module.btl_prepare_dst: this method does not exist in BTL 3.0. * module.btl_put: the signature and body of this method changed between 2.0 and 3.0.
2015-02-07 04:41:50 +03:00
#else
if ((frag->sf_base.uf_remote_seg[0].seg_addr.pval != NULL) ||
(frag->sf_base.uf_base.des_flags &
MCA_BTL_DES_SEND_ALWAYS_CALLBACK)) {
OPAL_BTL_USNIC_DO_SEND_FRAG_CB(module, frag, "send completion");
}
#endif
First commit of the Cisco usNIC BTL. This BTL accesses the Cisco usNIC Linux device via the Linux verbs API via Unreliable Datagram queue pairs. A few noteworthy points: * This BTL does most of its own fragmentation; it tells the PML that it has a very high max_send_size (much higher than the network MTU). * Since UD fragments are, by definition, unreliable, the usnic BTL handles all of its own reliability via a sliding window approach using the opal_hotel construct and many tricks stolen from the corpus of knowledge surrounding efficient TCP. * There is a fun PML latency-metric based optimization for NUMA awareness of short messages. * Note that this is ''not'' a generic UD verbs BTL; it is specific to the Cisco usNIC device. This commit was SVN r28879.
2013-07-20 02:13:58 +04:00
}
/* free this segment */
sseg->ss_ack_pending = false;
Cisco CSCuj22803: fixes for Bsend changes required to support MPI_Bsend(). Introduces concept of attaching a buffer to a large segment that the PML can scribble into and we will send from. The reason we don't use a pinned buffer and send directly from that is that usnic_verbs does not (yes) support num_sge>1 for regular sends. This means the data gets copied twice, but that is unavoidable. changed the logic in handle_large_send to be more sensible Incorporated David's review comments This commit was SVN r29184.
2013-09-17 11:27:39 +04:00
if (sseg->ss_send_posted == 0) {
George did the work and deserves all the credit for it. Ralph did the merge, and deserves whatever blame results from errors in it :-) WHAT: Open our low-level communication infrastructure by moving all necessary components (btl/rcache/allocator/mpool) down in OPAL All the components required for inter-process communications are currently deeply integrated in the OMPI layer. Several groups/institutions have express interest in having a more generic communication infrastructure, without all the OMPI layer dependencies. This communication layer should be made available at a different software level, available to all layers in the Open MPI software stack. As an example, our ORTE layer could replace the current OOB and instead use the BTL directly, gaining access to more reactive network interfaces than TCP. Similarly, external software libraries could take advantage of our highly optimized AM (active message) communication layer for their own purpose. UTK with support from Sandia, developped a version of Open MPI where the entire communication infrastucture has been moved down to OPAL (btl/rcache/allocator/mpool). Most of the moved components have been updated to match the new schema, with few exceptions (mainly BTLs where I have no way of compiling/testing them). Thus, the completion of this RFC is tied to being able to completing this move for all BTLs. For this we need help from the rest of the Open MPI community, especially those supporting some of the BTLs. A non-exhaustive list of BTLs that qualify here is: mx, portals4, scif, udapl, ugni, usnic. This commit was SVN r32317.
2014-07-26 04:47:28 +04:00
opal_btl_usnic_release_send_segment(module, frag, sseg);
First commit of the Cisco usNIC BTL. This BTL accesses the Cisco usNIC Linux device via the Linux verbs API via Unreliable Datagram queue pairs. A few noteworthy points: * This BTL does most of its own fragmentation; it tells the PML that it has a very high max_send_size (much higher than the network MTU). * Since UD fragments are, by definition, unreliable, the usnic BTL handles all of its own reliability via a sliding window approach using the opal_hotel construct and many tricks stolen from the corpus of knowledge surrounding efficient TCP. * There is a fun PML latency-metric based optimization for NUMA awareness of short messages. * Note that this is ''not'' a generic UD verbs BTL; it is specific to the Cisco usNIC device. This commit was SVN r28879.
2013-07-20 02:13:58 +04:00
}
usnic: SEGV in OSU benchmarks Prevent frag from being freed out from under us in the case the PML callback routine calls usnic_free(). We accomplish this by delaying decrement of sf_bytes_to_ack until after the callback is performed, since sf_bytes_to_ack == 0 is condition of freeing the frag. Fixes Cisco bug CSCuj45094. Authored-by: Reese Faucette <rfaucett@cisco.com> cmr=v1.7.3 This commit was SVN r29264.
2013-09-27 01:48:04 +04:00
/* when no bytes left to ACK, fragment send is truly done */
/* see note above on why this is done here as opposed to earlier */
frag->sf_ack_bytes_left -= bytes_acked;
Cisco CSCuj12520: various problems running c_fence_put_1 - tag needs to be sent in *our* header, not the PML header - usnic_alloc() should return smaller value if too much data requested - be careful about callbacks vs removing items from lists (we need to remove from outr lists *before* the callback) - improve send callback handling - add some more MSGDEBUG2 logging and cleanup This commit was SVN r29181.
2013-09-17 11:20:44 +04:00
/* OK to return this fragment? */
George did the work and deserves all the credit for it. Ralph did the merge, and deserves whatever blame results from errors in it :-) WHAT: Open our low-level communication infrastructure by moving all necessary components (btl/rcache/allocator/mpool) down in OPAL All the components required for inter-process communications are currently deeply integrated in the OMPI layer. Several groups/institutions have express interest in having a more generic communication infrastructure, without all the OMPI layer dependencies. This communication layer should be made available at a different software level, available to all layers in the Open MPI software stack. As an example, our ORTE layer could replace the current OOB and instead use the BTL directly, gaining access to more reactive network interfaces than TCP. Similarly, external software libraries could take advantage of our highly optimized AM (active message) communication layer for their own purpose. UTK with support from Sandia, developped a version of Open MPI where the entire communication infrastucture has been moved down to OPAL (btl/rcache/allocator/mpool). Most of the moved components have been updated to match the new schema, with few exceptions (mainly BTLs where I have no way of compiling/testing them). Thus, the completion of this RFC is tied to being able to completing this move for all BTLs. For this we need help from the rest of the Open MPI community, especially those supporting some of the BTLs. A non-exhaustive list of BTLs that qualify here is: mx, portals4, scif, udapl, ugni, usnic. This commit was SVN r32317.
2014-07-26 04:47:28 +04:00
opal_btl_usnic_send_frag_return_cond(module, frag);
Cisco CSCuj12520: various problems running c_fence_put_1 - tag needs to be sent in *our* header, not the PML header - usnic_alloc() should return smaller value if too much data requested - be careful about callbacks vs removing items from lists (we need to remove from outr lists *before* the callback) - improve send callback handling - add some more MSGDEBUG2 logging and cleanup This commit was SVN r29181.
2013-09-17 11:20:44 +04:00
First commit of the Cisco usNIC BTL. This BTL accesses the Cisco usNIC Linux device via the Linux verbs API via Unreliable Datagram queue pairs. A few noteworthy points: * This BTL does most of its own fragmentation; it tells the PML that it has a very high max_send_size (much higher than the network MTU). * Since UD fragments are, by definition, unreliable, the usnic BTL handles all of its own reliability via a sliding window approach using the opal_hotel construct and many tricks stolen from the corpus of knowledge surrounding efficient TCP. * There is a fun PML latency-metric based optimization for NUMA awareness of short messages. * Note that this is ''not'' a generic UD verbs BTL; it is specific to the Cisco usNIC device. This commit was SVN r28879.
2013-07-20 02:13:58 +04:00
/* indicate this segment has been ACKed */
endpoint->endpoint_sent_segs[WINDOW_SIZE_MOD(is)] = NULL;
}
/* update ACK received */
endpoint->endpoint_ack_seq_rcvd = ack_seq;
/* send window may have opened, possibly make endpoint ready-to-send */
George did the work and deserves all the credit for it. Ralph did the merge, and deserves whatever blame results from errors in it :-) WHAT: Open our low-level communication infrastructure by moving all necessary components (btl/rcache/allocator/mpool) down in OPAL All the components required for inter-process communications are currently deeply integrated in the OMPI layer. Several groups/institutions have express interest in having a more generic communication infrastructure, without all the OMPI layer dependencies. This communication layer should be made available at a different software level, available to all layers in the Open MPI software stack. As an example, our ORTE layer could replace the current OOB and instead use the BTL directly, gaining access to more reactive network interfaces than TCP. Similarly, external software libraries could take advantage of our highly optimized AM (active message) communication layer for their own purpose. UTK with support from Sandia, developped a version of Open MPI where the entire communication infrastucture has been moved down to OPAL (btl/rcache/allocator/mpool). Most of the moved components have been updated to match the new schema, with few exceptions (mainly BTLs where I have no way of compiling/testing them). Thus, the completion of this RFC is tied to being able to completing this move for all BTLs. For this we need help from the rest of the Open MPI community, especially those supporting some of the BTLs. A non-exhaustive list of BTLs that qualify here is: mx, portals4, scif, udapl, ugni, usnic. This commit was SVN r32317.
2014-07-26 04:47:28 +04:00
opal_btl_usnic_check_rts(endpoint);
First commit of the Cisco usNIC BTL. This BTL accesses the Cisco usNIC Linux device via the Linux verbs API via Unreliable Datagram queue pairs. A few noteworthy points: * This BTL does most of its own fragmentation; it tells the PML that it has a very high max_send_size (much higher than the network MTU). * Since UD fragments are, by definition, unreliable, the usnic BTL handles all of its own reliability via a sliding window approach using the opal_hotel construct and many tricks stolen from the corpus of knowledge surrounding efficient TCP. * There is a fun PML latency-metric based optimization for NUMA awareness of short messages. * Note that this is ''not'' a generic UD verbs BTL; it is specific to the Cisco usNIC device. This commit was SVN r28879.
2013-07-20 02:13:58 +04:00
}
/*
* Send an ACK
*/
void
George did the work and deserves all the credit for it. Ralph did the merge, and deserves whatever blame results from errors in it :-) WHAT: Open our low-level communication infrastructure by moving all necessary components (btl/rcache/allocator/mpool) down in OPAL All the components required for inter-process communications are currently deeply integrated in the OMPI layer. Several groups/institutions have express interest in having a more generic communication infrastructure, without all the OMPI layer dependencies. This communication layer should be made available at a different software level, available to all layers in the Open MPI software stack. As an example, our ORTE layer could replace the current OOB and instead use the BTL directly, gaining access to more reactive network interfaces than TCP. Similarly, external software libraries could take advantage of our highly optimized AM (active message) communication layer for their own purpose. UTK with support from Sandia, developped a version of Open MPI where the entire communication infrastucture has been moved down to OPAL (btl/rcache/allocator/mpool). Most of the moved components have been updated to match the new schema, with few exceptions (mainly BTLs where I have no way of compiling/testing them). Thus, the completion of this RFC is tied to being able to completing this move for all BTLs. For this we need help from the rest of the Open MPI community, especially those supporting some of the BTLs. A non-exhaustive list of BTLs that qualify here is: mx, portals4, scif, udapl, ugni, usnic. This commit was SVN r32317.
2014-07-26 04:47:28 +04:00
opal_btl_usnic_ack_send(
opal_btl_usnic_module_t *module,
opal_btl_usnic_endpoint_t *endpoint)
First commit of the Cisco usNIC BTL. This BTL accesses the Cisco usNIC Linux device via the Linux verbs API via Unreliable Datagram queue pairs. A few noteworthy points: * This BTL does most of its own fragmentation; it tells the PML that it has a very high max_send_size (much higher than the network MTU). * Since UD fragments are, by definition, unreliable, the usnic BTL handles all of its own reliability via a sliding window approach using the opal_hotel construct and many tricks stolen from the corpus of knowledge surrounding efficient TCP. * There is a fun PML latency-metric based optimization for NUMA awareness of short messages. * Note that this is ''not'' a generic UD verbs BTL; it is specific to the Cisco usNIC device. This commit was SVN r28879.
2013-07-20 02:13:58 +04:00
{
George did the work and deserves all the credit for it. Ralph did the merge, and deserves whatever blame results from errors in it :-) WHAT: Open our low-level communication infrastructure by moving all necessary components (btl/rcache/allocator/mpool) down in OPAL All the components required for inter-process communications are currently deeply integrated in the OMPI layer. Several groups/institutions have express interest in having a more generic communication infrastructure, without all the OMPI layer dependencies. This communication layer should be made available at a different software level, available to all layers in the Open MPI software stack. As an example, our ORTE layer could replace the current OOB and instead use the BTL directly, gaining access to more reactive network interfaces than TCP. Similarly, external software libraries could take advantage of our highly optimized AM (active message) communication layer for their own purpose. UTK with support from Sandia, developped a version of Open MPI where the entire communication infrastucture has been moved down to OPAL (btl/rcache/allocator/mpool). Most of the moved components have been updated to match the new schema, with few exceptions (mainly BTLs where I have no way of compiling/testing them). Thus, the completion of this RFC is tied to being able to completing this move for all BTLs. For this we need help from the rest of the Open MPI community, especially those supporting some of the BTLs. A non-exhaustive list of BTLs that qualify here is: mx, portals4, scif, udapl, ugni, usnic. This commit was SVN r32317.
2014-07-26 04:47:28 +04:00
opal_btl_usnic_ack_segment_t *ack;
First commit of the Cisco usNIC BTL. This BTL accesses the Cisco usNIC Linux device via the Linux verbs API via Unreliable Datagram queue pairs. A few noteworthy points: * This BTL does most of its own fragmentation; it tells the PML that it has a very high max_send_size (much higher than the network MTU). * Since UD fragments are, by definition, unreliable, the usnic BTL handles all of its own reliability via a sliding window approach using the opal_hotel construct and many tricks stolen from the corpus of knowledge surrounding efficient TCP. * There is a fun PML latency-metric based optimization for NUMA awareness of short messages. * Note that this is ''not'' a generic UD verbs BTL; it is specific to the Cisco usNIC device. This commit was SVN r28879.
2013-07-20 02:13:58 +04:00
/* Get an ACK frag. If we don't get one, just discard this ACK. */
George did the work and deserves all the credit for it. Ralph did the merge, and deserves whatever blame results from errors in it :-) WHAT: Open our low-level communication infrastructure by moving all necessary components (btl/rcache/allocator/mpool) down in OPAL All the components required for inter-process communications are currently deeply integrated in the OMPI layer. Several groups/institutions have express interest in having a more generic communication infrastructure, without all the OMPI layer dependencies. This communication layer should be made available at a different software level, available to all layers in the Open MPI software stack. As an example, our ORTE layer could replace the current OOB and instead use the BTL directly, gaining access to more reactive network interfaces than TCP. Similarly, external software libraries could take advantage of our highly optimized AM (active message) communication layer for their own purpose. UTK with support from Sandia, developped a version of Open MPI where the entire communication infrastucture has been moved down to OPAL (btl/rcache/allocator/mpool). Most of the moved components have been updated to match the new schema, with few exceptions (mainly BTLs where I have no way of compiling/testing them). Thus, the completion of this RFC is tied to being able to completing this move for all BTLs. For this we need help from the rest of the Open MPI community, especially those supporting some of the BTLs. A non-exhaustive list of BTLs that qualify here is: mx, portals4, scif, udapl, ugni, usnic. This commit was SVN r32317.
2014-07-26 04:47:28 +04:00
ack = opal_btl_usnic_ack_segment_alloc(module);
First commit of the Cisco usNIC BTL. This BTL accesses the Cisco usNIC Linux device via the Linux verbs API via Unreliable Datagram queue pairs. A few noteworthy points: * This BTL does most of its own fragmentation; it tells the PML that it has a very high max_send_size (much higher than the network MTU). * Since UD fragments are, by definition, unreliable, the usnic BTL handles all of its own reliability via a sliding window approach using the opal_hotel construct and many tricks stolen from the corpus of knowledge surrounding efficient TCP. * There is a fun PML latency-metric based optimization for NUMA awareness of short messages. * Note that this is ''not'' a generic UD verbs BTL; it is specific to the Cisco usNIC device. This commit was SVN r28879.
2013-07-20 02:13:58 +04:00
if (OPAL_UNLIKELY(NULL == ack)) {
opal_output(0, "====================== No frag for sending the ACK -- skipped");
abort();
}
/* send the seq of the lowest item in the window that
we've received */
ack->ss_base.us_btl_header->ack_seq =
endpoint->endpoint_next_contig_seq_to_recv - 1;
usnic: convert from verbs to libfabric (yay!) This commit represents the conversion of the usnic BTL from verbs to libfabric. For the moment, libfabric is embedded in Open MPI (currently in the usnic BTL). This is because the libfabric API is still changing, and also has not yet been released. Ultimately, this embedded copy of libfabric will likely disappear and the usnic BTL will rely on an external installation of libfabric. New configure options: * --with-libfabric: will cause configure to fail if libfabric support cannot be built * --without-libfabric: will prevent libfabric support from being built * --with-libfabric=DIR: use an external libfabric installation * --with-libfabric-libdir=LIBDIR: when paired with --with-libfabric=DIR, use LIBDIR for the libfabric installation library dir The --with-libnl3[-libdir] arguments are now gone.
2014-12-03 00:09:46 +03:00
ack->ss_len = sizeof(opal_btl_usnic_btl_header_t);
First commit of the Cisco usNIC BTL. This BTL accesses the Cisco usNIC Linux device via the Linux verbs API via Unreliable Datagram queue pairs. A few noteworthy points: * This BTL does most of its own fragmentation; it tells the PML that it has a very high max_send_size (much higher than the network MTU). * Since UD fragments are, by definition, unreliable, the usnic BTL handles all of its own reliability via a sliding window approach using the opal_hotel construct and many tricks stolen from the corpus of knowledge surrounding efficient TCP. * There is a fun PML latency-metric based optimization for NUMA awareness of short messages. * Note that this is ''not'' a generic UD verbs BTL; it is specific to the Cisco usNIC device. This commit was SVN r28879.
2013-07-20 02:13:58 +04:00
#if MSGDEBUG1
usnic: minor cleanup This commit was SVN r32367.
2014-07-31 00:56:49 +04:00
{
usnic: convert from verbs to libfabric (yay!) This commit represents the conversion of the usnic BTL from verbs to libfabric. For the moment, libfabric is embedded in Open MPI (currently in the usnic BTL). This is because the libfabric API is still changing, and also has not yet been released. Ultimately, this embedded copy of libfabric will likely disappear and the usnic BTL will rely on an external installation of libfabric. New configure options: * --with-libfabric: will cause configure to fail if libfabric support cannot be built * --without-libfabric: will prevent libfabric support from being built * --with-libfabric=DIR: use an external libfabric installation * --with-libfabric-libdir=LIBDIR: when paired with --with-libfabric=DIR, use LIBDIR for the libfabric installation library dir The --with-libnl3[-libdir] arguments are now gone.
2014-12-03 00:09:46 +03:00
char remote_ip[IPV4STRADDRLEN];
struct opal_btl_usnic_modex_t *modex =
&endpoint->endpoint_remote_modex;
opal_btl_usnic_snprintf_ipv4_addr(remote_ip, sizeof(remote_ip),
modex->ipv4_addr,
modex->netmask);
opal_output(0, "--> Sending ACK, length %d, seq %" UDSEQ " to %s, port %u",
ack->ss_len,
ack->ss_base.us_btl_header->ack_seq,
remote_ip,
modex->ports[ack->ss_channel]);
usnic: minor cleanup This commit was SVN r32367.
2014-07-31 00:56:49 +04:00
}
First commit of the Cisco usNIC BTL. This BTL accesses the Cisco usNIC Linux device via the Linux verbs API via Unreliable Datagram queue pairs. A few noteworthy points: * This BTL does most of its own fragmentation; it tells the PML that it has a very high max_send_size (much higher than the network MTU). * Since UD fragments are, by definition, unreliable, the usnic BTL handles all of its own reliability via a sliding window approach using the opal_hotel construct and many tricks stolen from the corpus of knowledge surrounding efficient TCP. * There is a fun PML latency-metric based optimization for NUMA awareness of short messages. * Note that this is ''not'' a generic UD verbs BTL; it is specific to the Cisco usNIC device. This commit was SVN r28879.
2013-07-20 02:13:58 +04:00
#endif
usnic: check connectivity on first communication to a peer Previously, we were only checking connectivity upon first ''send'' to a peer. But this ignores the case where the first communication to a peer is actually an ACK -- i.e., we successfully received something from the peer and we need to send an ACK back. So we need to verify that the ACK will actually get there. Specifically, certain asymmetric routing cases can lead to a hang if we don't check the connectivity in both directions. E.g., if the sender is able to get traffic to the receiver, but the receiver is unable to get traffic back to the sender because it made a different routing decision than the sender. In this case, the connectivity checker from the sender could succeed (because the connectivity checker will ACK along the same path in which the ping was received), but sending a BTL ACK could fail (because the BTL ACK will be sent back along the path chosen by the graph algorithm, which, in an erroneous asymmetric routing scenario, may be different/wrong). Hence, we want to trigger the connectivity checker at the first communication from A->B, which may either be a BTL send or an ACK. Reviewed by Dave Goodell. cmr=v1.8.2:reviewer=ompi-rm1.8 This commit was SVN r32309.
2014-07-25 01:32:56 +04:00
/* Do we need to check the connectivity? If enabled, we'll check
the connectivity at either first send to peer X or first ACK to
peer X. */
George did the work and deserves all the credit for it. Ralph did the merge, and deserves whatever blame results from errors in it :-) WHAT: Open our low-level communication infrastructure by moving all necessary components (btl/rcache/allocator/mpool) down in OPAL All the components required for inter-process communications are currently deeply integrated in the OMPI layer. Several groups/institutions have express interest in having a more generic communication infrastructure, without all the OMPI layer dependencies. This communication layer should be made available at a different software level, available to all layers in the Open MPI software stack. As an example, our ORTE layer could replace the current OOB and instead use the BTL directly, gaining access to more reactive network interfaces than TCP. Similarly, external software libraries could take advantage of our highly optimized AM (active message) communication layer for their own purpose. UTK with support from Sandia, developped a version of Open MPI where the entire communication infrastucture has been moved down to OPAL (btl/rcache/allocator/mpool). Most of the moved components have been updated to match the new schema, with few exceptions (mainly BTLs where I have no way of compiling/testing them). Thus, the completion of this RFC is tied to being able to completing this move for all BTLs. For this we need help from the rest of the Open MPI community, especially those supporting some of the BTLs. A non-exhaustive list of BTLs that qualify here is: mx, portals4, scif, udapl, ugni, usnic. This commit was SVN r32317.
2014-07-26 04:47:28 +04:00
opal_btl_usnic_check_connectivity(module, endpoint);
usnic: check connectivity on first communication to a peer Previously, we were only checking connectivity upon first ''send'' to a peer. But this ignores the case where the first communication to a peer is actually an ACK -- i.e., we successfully received something from the peer and we need to send an ACK back. So we need to verify that the ACK will actually get there. Specifically, certain asymmetric routing cases can lead to a hang if we don't check the connectivity in both directions. E.g., if the sender is able to get traffic to the receiver, but the receiver is unable to get traffic back to the sender because it made a different routing decision than the sender. In this case, the connectivity checker from the sender could succeed (because the connectivity checker will ACK along the same path in which the ping was received), but sending a BTL ACK could fail (because the BTL ACK will be sent back along the path chosen by the graph algorithm, which, in an erroneous asymmetric routing scenario, may be different/wrong). Hence, we want to trigger the connectivity checker at the first communication from A->B, which may either be a BTL send or an ACK. Reviewed by Dave Goodell. cmr=v1.8.2:reviewer=ompi-rm1.8 This commit was SVN r32309.
2014-07-25 01:32:56 +04:00
First commit of the Cisco usNIC BTL. This BTL accesses the Cisco usNIC Linux device via the Linux verbs API via Unreliable Datagram queue pairs. A few noteworthy points: * This BTL does most of its own fragmentation; it tells the PML that it has a very high max_send_size (much higher than the network MTU). * Since UD fragments are, by definition, unreliable, the usnic BTL handles all of its own reliability via a sliding window approach using the opal_hotel construct and many tricks stolen from the corpus of knowledge surrounding efficient TCP. * There is a fun PML latency-metric based optimization for NUMA awareness of short messages. * Note that this is ''not'' a generic UD verbs BTL; it is specific to the Cisco usNIC device. This commit was SVN r28879.
2013-07-20 02:13:58 +04:00
/* send the ACK */
usnic: convert from verbs to libfabric (yay!) This commit represents the conversion of the usnic BTL from verbs to libfabric. For the moment, libfabric is embedded in Open MPI (currently in the usnic BTL). This is because the libfabric API is still changing, and also has not yet been released. Ultimately, this embedded copy of libfabric will likely disappear and the usnic BTL will rely on an external installation of libfabric. New configure options: * --with-libfabric: will cause configure to fail if libfabric support cannot be built * --without-libfabric: will prevent libfabric support from being built * --with-libfabric=DIR: use an external libfabric installation * --with-libfabric-libdir=LIBDIR: when paired with --with-libfabric=DIR, use LIBDIR for the libfabric installation library dir The --with-libnl3[-libdir] arguments are now gone.
2014-12-03 00:09:46 +03:00
opal_btl_usnic_post_ack(module, endpoint, ack);
First commit of the Cisco usNIC BTL. This BTL accesses the Cisco usNIC Linux device via the Linux verbs API via Unreliable Datagram queue pairs. A few noteworthy points: * This BTL does most of its own fragmentation; it tells the PML that it has a very high max_send_size (much higher than the network MTU). * Since UD fragments are, by definition, unreliable, the usnic BTL handles all of its own reliability via a sliding window approach using the opal_hotel construct and many tricks stolen from the corpus of knowledge surrounding efficient TCP. * There is a fun PML latency-metric based optimization for NUMA awareness of short messages. * Note that this is ''not'' a generic UD verbs BTL; it is specific to the Cisco usNIC device. This commit was SVN r28879.
2013-07-20 02:13:58 +04:00
/* Stats */
Move all usNIC stats to _stats.c|h and export them as MPI_T pvars. This commit moves all the module stats into their own struct so that the stats only need to appear as a single line in the module_t definition, and then moves all the logic for reporting the stats into btl_usnic_stats.c|h. Further, the stats are now exported as MPI_T_BIND_NO_OBJECT entities (i.e., not bound to any particular MPI handle), and are marked as READONLY and CONTINUOUS. They currently all default to verbose level 5 ("Application tuner / detailed", according to https://svn.open-mpi.org/trac/ompi/wiki/MCAParamLevels). Most of the statistics are counters, but a small number are high watermark values. Due to how counters are reported via MPI_T, none of the counters are exported through MPI_T if the MCA param btl_usnic_stats_relative=1 (i.e., the module resets the stats back to zero at a given frequency). When MPI_T_pvar_handle_alloc() is invoked on any of these pvars, it will return a count that is equal to the number of active usnic BTL modules. The values returned for any given pvar (e.g., num_total_sends) are an array containing one value for each active usnic BTL module. The ordering of values in the array is both consistent across all usnic pvars and stable throughout a single job: array slot 0 corresponds to module X, array slot 1 corresponds to module Y, etc. Mapping which array slot corresponds to which underlying Linux usnic_X device works as follows: * The btl_usnic_devices MPI_T state pvar is associated with a btl_usnic_device MPI_T enum, and be obtained via MPI_T_pvar_get_info(). * If all usNIC pvars are of length N, the values [0,N) in the btl_usnic_device enum are associated with strings of the corresponding underlying Linux device. For exampe, to look up which Linux device is reported in all usNIC pvars' array slot 1, look up the int value 1 in the btl_usnic_devices enum. Its corresponding string value is underlying Linux device name (e.g., "usnic_1"). cmr=v1.7.4:subject="usnic BTL MPI_T pvars" This commit was SVN r29545.
2013-10-29 02:23:08 +04:00
++module->stats.num_ack_sends;
First commit of the Cisco usNIC BTL. This BTL accesses the Cisco usNIC Linux device via the Linux verbs API via Unreliable Datagram queue pairs. A few noteworthy points: * This BTL does most of its own fragmentation; it tells the PML that it has a very high max_send_size (much higher than the network MTU). * Since UD fragments are, by definition, unreliable, the usnic BTL handles all of its own reliability via a sliding window approach using the opal_hotel construct and many tricks stolen from the corpus of knowledge surrounding efficient TCP. * There is a fun PML latency-metric based optimization for NUMA awareness of short messages. * Note that this is ''not'' a generic UD verbs BTL; it is specific to the Cisco usNIC device. This commit was SVN r28879.
2013-07-20 02:13:58 +04:00
return;
}
/*
* Sending an ACK has completed, return the segment to the free list
*/
void
George did the work and deserves all the credit for it. Ralph did the merge, and deserves whatever blame results from errors in it :-) WHAT: Open our low-level communication infrastructure by moving all necessary components (btl/rcache/allocator/mpool) down in OPAL All the components required for inter-process communications are currently deeply integrated in the OMPI layer. Several groups/institutions have express interest in having a more generic communication infrastructure, without all the OMPI layer dependencies. This communication layer should be made available at a different software level, available to all layers in the Open MPI software stack. As an example, our ORTE layer could replace the current OOB and instead use the BTL directly, gaining access to more reactive network interfaces than TCP. Similarly, external software libraries could take advantage of our highly optimized AM (active message) communication layer for their own purpose. UTK with support from Sandia, developped a version of Open MPI where the entire communication infrastucture has been moved down to OPAL (btl/rcache/allocator/mpool). Most of the moved components have been updated to match the new schema, with few exceptions (mainly BTLs where I have no way of compiling/testing them). Thus, the completion of this RFC is tied to being able to completing this move for all BTLs. For this we need help from the rest of the Open MPI community, especially those supporting some of the BTLs. A non-exhaustive list of BTLs that qualify here is: mx, portals4, scif, udapl, ugni, usnic. This commit was SVN r32317.
2014-07-26 04:47:28 +04:00
opal_btl_usnic_ack_complete(opal_btl_usnic_module_t *module,
opal_btl_usnic_ack_segment_t *ack)
First commit of the Cisco usNIC BTL. This BTL accesses the Cisco usNIC Linux device via the Linux verbs API via Unreliable Datagram queue pairs. A few noteworthy points: * This BTL does most of its own fragmentation; it tells the PML that it has a very high max_send_size (much higher than the network MTU). * Since UD fragments are, by definition, unreliable, the usnic BTL handles all of its own reliability via a sliding window approach using the opal_hotel construct and many tricks stolen from the corpus of knowledge surrounding efficient TCP. * There is a fun PML latency-metric based optimization for NUMA awareness of short messages. * Note that this is ''not'' a generic UD verbs BTL; it is specific to the Cisco usNIC device. This commit was SVN r28879.
2013-07-20 02:13:58 +04:00
{
George did the work and deserves all the credit for it. Ralph did the merge, and deserves whatever blame results from errors in it :-) WHAT: Open our low-level communication infrastructure by moving all necessary components (btl/rcache/allocator/mpool) down in OPAL All the components required for inter-process communications are currently deeply integrated in the OMPI layer. Several groups/institutions have express interest in having a more generic communication infrastructure, without all the OMPI layer dependencies. This communication layer should be made available at a different software level, available to all layers in the Open MPI software stack. As an example, our ORTE layer could replace the current OOB and instead use the BTL directly, gaining access to more reactive network interfaces than TCP. Similarly, external software libraries could take advantage of our highly optimized AM (active message) communication layer for their own purpose. UTK with support from Sandia, developped a version of Open MPI where the entire communication infrastucture has been moved down to OPAL (btl/rcache/allocator/mpool). Most of the moved components have been updated to match the new schema, with few exceptions (mainly BTLs where I have no way of compiling/testing them). Thus, the completion of this RFC is tied to being able to completing this move for all BTLs. For this we need help from the rest of the Open MPI community, especially those supporting some of the BTLs. A non-exhaustive list of BTLs that qualify here is: mx, portals4, scif, udapl, ugni, usnic. This commit was SVN r32317.
2014-07-26 04:47:28 +04:00
opal_btl_usnic_ack_segment_return(module, ack);
First commit of the Cisco usNIC BTL. This BTL accesses the Cisco usNIC Linux device via the Linux verbs API via Unreliable Datagram queue pairs. A few noteworthy points: * This BTL does most of its own fragmentation; it tells the PML that it has a very high max_send_size (much higher than the network MTU). * Since UD fragments are, by definition, unreliable, the usnic BTL handles all of its own reliability via a sliding window approach using the opal_hotel construct and many tricks stolen from the corpus of knowledge surrounding efficient TCP. * There is a fun PML latency-metric based optimization for NUMA awareness of short messages. * Note that this is ''not'' a generic UD verbs BTL; it is specific to the Cisco usNIC device. This commit was SVN r28879.
2013-07-20 02:13:58 +04:00
}
/*****************************************************************************/
/*
* Callback for when a send times out without receiving a
* corresponding ACK.
*/
void
George did the work and deserves all the credit for it. Ralph did the merge, and deserves whatever blame results from errors in it :-) WHAT: Open our low-level communication infrastructure by moving all necessary components (btl/rcache/allocator/mpool) down in OPAL All the components required for inter-process communications are currently deeply integrated in the OMPI layer. Several groups/institutions have express interest in having a more generic communication infrastructure, without all the OMPI layer dependencies. This communication layer should be made available at a different software level, available to all layers in the Open MPI software stack. As an example, our ORTE layer could replace the current OOB and instead use the BTL directly, gaining access to more reactive network interfaces than TCP. Similarly, external software libraries could take advantage of our highly optimized AM (active message) communication layer for their own purpose. UTK with support from Sandia, developped a version of Open MPI where the entire communication infrastucture has been moved down to OPAL (btl/rcache/allocator/mpool). Most of the moved components have been updated to match the new schema, with few exceptions (mainly BTLs where I have no way of compiling/testing them). Thus, the completion of this RFC is tied to being able to completing this move for all BTLs. For this we need help from the rest of the Open MPI community, especially those supporting some of the BTLs. A non-exhaustive list of BTLs that qualify here is: mx, portals4, scif, udapl, ugni, usnic. This commit was SVN r32317.
2014-07-26 04:47:28 +04:00
opal_btl_usnic_ack_timeout(
First commit of the Cisco usNIC BTL. This BTL accesses the Cisco usNIC Linux device via the Linux verbs API via Unreliable Datagram queue pairs. A few noteworthy points: * This BTL does most of its own fragmentation; it tells the PML that it has a very high max_send_size (much higher than the network MTU). * Since UD fragments are, by definition, unreliable, the usnic BTL handles all of its own reliability via a sliding window approach using the opal_hotel construct and many tricks stolen from the corpus of knowledge surrounding efficient TCP. * There is a fun PML latency-metric based optimization for NUMA awareness of short messages. * Note that this is ''not'' a generic UD verbs BTL; it is specific to the Cisco usNIC device. This commit was SVN r28879.
2013-07-20 02:13:58 +04:00
opal_hotel_t *hotel,
usnic: remove all trailing whitespace Style cleanup only; no code changes. This commit was SVN r32366.
2014-07-31 00:56:15 +04:00
int room_num,
First commit of the Cisco usNIC BTL. This BTL accesses the Cisco usNIC Linux device via the Linux verbs API via Unreliable Datagram queue pairs. A few noteworthy points: * This BTL does most of its own fragmentation; it tells the PML that it has a very high max_send_size (much higher than the network MTU). * Since UD fragments are, by definition, unreliable, the usnic BTL handles all of its own reliability via a sliding window approach using the opal_hotel construct and many tricks stolen from the corpus of knowledge surrounding efficient TCP. * There is a fun PML latency-metric based optimization for NUMA awareness of short messages. * Note that this is ''not'' a generic UD verbs BTL; it is specific to the Cisco usNIC device. This commit was SVN r28879.
2013-07-20 02:13:58 +04:00
void *occupant)
{
George did the work and deserves all the credit for it. Ralph did the merge, and deserves whatever blame results from errors in it :-) WHAT: Open our low-level communication infrastructure by moving all necessary components (btl/rcache/allocator/mpool) down in OPAL All the components required for inter-process communications are currently deeply integrated in the OMPI layer. Several groups/institutions have express interest in having a more generic communication infrastructure, without all the OMPI layer dependencies. This communication layer should be made available at a different software level, available to all layers in the Open MPI software stack. As an example, our ORTE layer could replace the current OOB and instead use the BTL directly, gaining access to more reactive network interfaces than TCP. Similarly, external software libraries could take advantage of our highly optimized AM (active message) communication layer for their own purpose. UTK with support from Sandia, developped a version of Open MPI where the entire communication infrastucture has been moved down to OPAL (btl/rcache/allocator/mpool). Most of the moved components have been updated to match the new schema, with few exceptions (mainly BTLs where I have no way of compiling/testing them). Thus, the completion of this RFC is tied to being able to completing this move for all BTLs. For this we need help from the rest of the Open MPI community, especially those supporting some of the BTLs. A non-exhaustive list of BTLs that qualify here is: mx, portals4, scif, udapl, ugni, usnic. This commit was SVN r32317.
2014-07-26 04:47:28 +04:00
opal_btl_usnic_send_segment_t *seg;
opal_btl_usnic_endpoint_t *endpoint;
opal_btl_usnic_module_t *module;
First commit of the Cisco usNIC BTL. This BTL accesses the Cisco usNIC Linux device via the Linux verbs API via Unreliable Datagram queue pairs. A few noteworthy points: * This BTL does most of its own fragmentation; it tells the PML that it has a very high max_send_size (much higher than the network MTU). * Since UD fragments are, by definition, unreliable, the usnic BTL handles all of its own reliability via a sliding window approach using the opal_hotel construct and many tricks stolen from the corpus of knowledge surrounding efficient TCP. * There is a fun PML latency-metric based optimization for NUMA awareness of short messages. * Note that this is ''not'' a generic UD verbs BTL; it is specific to the Cisco usNIC device. This commit was SVN r28879.
2013-07-20 02:13:58 +04:00
George did the work and deserves all the credit for it. Ralph did the merge, and deserves whatever blame results from errors in it :-) WHAT: Open our low-level communication infrastructure by moving all necessary components (btl/rcache/allocator/mpool) down in OPAL All the components required for inter-process communications are currently deeply integrated in the OMPI layer. Several groups/institutions have express interest in having a more generic communication infrastructure, without all the OMPI layer dependencies. This communication layer should be made available at a different software level, available to all layers in the Open MPI software stack. As an example, our ORTE layer could replace the current OOB and instead use the BTL directly, gaining access to more reactive network interfaces than TCP. Similarly, external software libraries could take advantage of our highly optimized AM (active message) communication layer for their own purpose. UTK with support from Sandia, developped a version of Open MPI where the entire communication infrastucture has been moved down to OPAL (btl/rcache/allocator/mpool). Most of the moved components have been updated to match the new schema, with few exceptions (mainly BTLs where I have no way of compiling/testing them). Thus, the completion of this RFC is tied to being able to completing this move for all BTLs. For this we need help from the rest of the Open MPI community, especially those supporting some of the BTLs. A non-exhaustive list of BTLs that qualify here is: mx, portals4, scif, udapl, ugni, usnic. This commit was SVN r32317.
2014-07-26 04:47:28 +04:00
seg = (opal_btl_usnic_send_segment_t*) occupant;
First commit of the Cisco usNIC BTL. This BTL accesses the Cisco usNIC Linux device via the Linux verbs API via Unreliable Datagram queue pairs. A few noteworthy points: * This BTL does most of its own fragmentation; it tells the PML that it has a very high max_send_size (much higher than the network MTU). * Since UD fragments are, by definition, unreliable, the usnic BTL handles all of its own reliability via a sliding window approach using the opal_hotel construct and many tricks stolen from the corpus of knowledge surrounding efficient TCP. * There is a fun PML latency-metric based optimization for NUMA awareness of short messages. * Note that this is ''not'' a generic UD verbs BTL; it is specific to the Cisco usNIC device. This commit was SVN r28879.
2013-07-20 02:13:58 +04:00
endpoint = seg->ss_parent_frag->sf_endpoint;
module = endpoint->endpoint_module;
Clean up debugging logging a bit. MSGDEBUG2 now means "print a one-liner for all PML calls into BTL, and also when BTL calls PML with a recv completion (not send completions)" MSGDEBUG1 means print more internal gory detail MSGDEBUG is gone, replaced by MSGDEBUG1 In the process also found that PUT_DEST style fragments could potentially be leaked in usnic_free() since send_fragment tests were being applied to see if it was eligible to be freed. This commit was SVN r29185.
2013-09-17 11:29:40 +04:00
#if MSGDEBUG1
First commit of the Cisco usNIC BTL. This BTL accesses the Cisco usNIC Linux device via the Linux verbs API via Unreliable Datagram queue pairs. A few noteworthy points: * This BTL does most of its own fragmentation; it tells the PML that it has a very high max_send_size (much higher than the network MTU). * Since UD fragments are, by definition, unreliable, the usnic BTL handles all of its own reliability via a sliding window approach using the opal_hotel construct and many tricks stolen from the corpus of knowledge surrounding efficient TCP. * There is a fun PML latency-metric based optimization for NUMA awareness of short messages. * Note that this is ''not'' a generic UD verbs BTL; it is specific to the Cisco usNIC device. This commit was SVN r28879.
2013-07-20 02:13:58 +04:00
{
opal_output(0, "Send timeout! seg %p, room %d, seq %" UDSEQ "\n",
usnic: stomp format specifier warnings The usnic BTL now builds cleanly under `--enable-picky` when `MSGDEBUG1` is set. Reviewed-by: jsquyres cmr=v1.7.4:reviewer=jsquyres This commit was SVN r29097.
2013-08-30 03:24:14 +04:00
(void*)seg, seg->ss_hotel_room,
usnic: Shrink sequence numbers to 16 bits Authored-by: Reese Faucette <rfaucett@cisco.com> Reviewed-by: Jeff Squyres <jsquyres@cisco.com> cmr=v1.7.5:ticket=trac:4253 This commit was SVN r30834. The following Trac tickets were found above: Ticket 4253 --> https://svn.open-mpi.org/trac/ompi/ticket/4253
2014-02-26 11:40:10 +04:00
seg->ss_base.us_btl_header->pkt_seq);
First commit of the Cisco usNIC BTL. This BTL accesses the Cisco usNIC Linux device via the Linux verbs API via Unreliable Datagram queue pairs. A few noteworthy points: * This BTL does most of its own fragmentation; it tells the PML that it has a very high max_send_size (much higher than the network MTU). * Since UD fragments are, by definition, unreliable, the usnic BTL handles all of its own reliability via a sliding window approach using the opal_hotel construct and many tricks stolen from the corpus of knowledge surrounding efficient TCP. * There is a fun PML latency-metric based optimization for NUMA awareness of short messages. * Note that this is ''not'' a generic UD verbs BTL; it is specific to the Cisco usNIC device. This commit was SVN r28879.
2013-07-20 02:13:58 +04:00
}
#endif
/* timeout checks us out, note this */
seg->ss_hotel_room = -1;
/* Queue up this frag to be resent */
usnic: remove all trailing whitespace Style cleanup only; no code changes. This commit was SVN r32366.
2014-07-31 00:56:15 +04:00
opal_list_append(&(module->pending_resend_segs),
First commit of the Cisco usNIC BTL. This BTL accesses the Cisco usNIC Linux device via the Linux verbs API via Unreliable Datagram queue pairs. A few noteworthy points: * This BTL does most of its own fragmentation; it tells the PML that it has a very high max_send_size (much higher than the network MTU). * Since UD fragments are, by definition, unreliable, the usnic BTL handles all of its own reliability via a sliding window approach using the opal_hotel construct and many tricks stolen from the corpus of knowledge surrounding efficient TCP. * There is a fun PML latency-metric based optimization for NUMA awareness of short messages. * Note that this is ''not'' a generic UD verbs BTL; it is specific to the Cisco usNIC device. This commit was SVN r28879.
2013-07-20 02:13:58 +04:00
&(seg->ss_base.us_list.super));
/* Stats */
Move all usNIC stats to _stats.c|h and export them as MPI_T pvars. This commit moves all the module stats into their own struct so that the stats only need to appear as a single line in the module_t definition, and then moves all the logic for reporting the stats into btl_usnic_stats.c|h. Further, the stats are now exported as MPI_T_BIND_NO_OBJECT entities (i.e., not bound to any particular MPI handle), and are marked as READONLY and CONTINUOUS. They currently all default to verbose level 5 ("Application tuner / detailed", according to https://svn.open-mpi.org/trac/ompi/wiki/MCAParamLevels). Most of the statistics are counters, but a small number are high watermark values. Due to how counters are reported via MPI_T, none of the counters are exported through MPI_T if the MCA param btl_usnic_stats_relative=1 (i.e., the module resets the stats back to zero at a given frequency). When MPI_T_pvar_handle_alloc() is invoked on any of these pvars, it will return a count that is equal to the number of active usnic BTL modules. The values returned for any given pvar (e.g., num_total_sends) are an array containing one value for each active usnic BTL module. The ordering of values in the array is both consistent across all usnic pvars and stable throughout a single job: array slot 0 corresponds to module X, array slot 1 corresponds to module Y, etc. Mapping which array slot corresponds to which underlying Linux usnic_X device works as follows: * The btl_usnic_devices MPI_T state pvar is associated with a btl_usnic_device MPI_T enum, and be obtained via MPI_T_pvar_get_info(). * If all usNIC pvars are of length N, the values [0,N) in the btl_usnic_device enum are associated with strings of the corresponding underlying Linux device. For exampe, to look up which Linux device is reported in all usNIC pvars' array slot 1, look up the int value 1 in the btl_usnic_devices enum. Its corresponding string value is underlying Linux device name (e.g., "usnic_1"). cmr=v1.7.4:subject="usnic BTL MPI_T pvars" This commit was SVN r29545.
2013-10-29 02:23:08 +04:00
++module->stats.num_timeout_retrans;
First commit of the Cisco usNIC BTL. This BTL accesses the Cisco usNIC Linux device via the Linux verbs API via Unreliable Datagram queue pairs. A few noteworthy points: * This BTL does most of its own fragmentation; it tells the PML that it has a very high max_send_size (much higher than the network MTU). * Since UD fragments are, by definition, unreliable, the usnic BTL handles all of its own reliability via a sliding window approach using the opal_hotel construct and many tricks stolen from the corpus of knowledge surrounding efficient TCP. * There is a fun PML latency-metric based optimization for NUMA awareness of short messages. * Note that this is ''not'' a generic UD verbs BTL; it is specific to the Cisco usNIC device. This commit was SVN r28879.
2013-07-20 02:13:58 +04:00
}
Копировать постоянную ссылку