ports/openmpi
ports/openmpi
1
1
Форкнуть 0
Код Релизы Активность
7 710 коммитов 3 Ветки 0 Теги 130 MiB
01a59d68da
Граф коммитов

315 Коммитов

Автор SHA1 Сообщение Дата
George Bosilca
01a59d68da Do not generate the XFER_BEGIN and XFER_END events if the length of 
the data is zero, for both the receives and the sends.

This commit was SVN r10670.
 2006-07-05 23:39:13 +00:00
George Bosilca
6265625983 Generate the XFER_CONTINUE PERUSE event (or the receive) before unpacking the data. 
This commit was SVN r10663.
 2006-07-05 19:45:00 +00:00
George Bosilca
2bdb06b549 Force the request to NULL in order to avoid complaints from the compiler. 
This commit was SVN r10647.
 2006-07-04 06:20:13 +00:00
George Bosilca
9ac1a6cdb3 Remove the warnings. Now they are ompi_free_list_item not opal_list_item_t. 
This commit was SVN r10645.
 2006-07-04 04:21:16 +00:00
Brian Barrett
7d12f9119a * make sure to include post_configure.sh in the dist tarball, so that 
direct calling the ob1 pml works properly.

This commit was SVN r10644.
 2006-07-04 04:03:58 +00:00
Brian Barrett
47725c9b02 * Add new PML (CM) and network drivers (MTL) for high speed 
interconnects that provide matching logic in the library.
  Currently includes support for MX and some support for
  Portals
* Fix overuse of proc_pml pointer on the ompi_proc structuer, 
  splitting into proc_pml for pml data and proc_bml for
  the BML endpoint data
* bug fixes in bsend init code, which wasn't being used by
  the OB1 or DR PMLs...

This commit was SVN r10642.
 2006-07-04 01:20:20 +00:00
Galen Shipman
e6cd8db0e5 DR will now checksum on a per btl basis (see MCA_BTL_FLAGS_NEED_CSUM). We 
still always send ACK's, teasing apart completion for ACK/no ACK looks like a
pain in the .. 

This commit was SVN r10530.
 2006-06-27 20:23:47 +00:00
Galen Shipman
8855e5b73a Fixes for DR as well as better diagnostic.. 
Successfully passing the intel test suite with/without induced errors/drops. 

This commit was SVN r10518.
 2006-06-26 22:29:29 +00:00
George Bosilca
940dbff0fa Add a new PERUSE macro. This is for the CONTINUE event (the one we added to the 
standard). This macro allow us to specify the length of the fragment. Now we are
able to know how the message is fragmented between the network devices or inside
the communication protocol.

This commit was SVN r10508.
 2006-06-26 20:08:33 +00:00
George Bosilca
c43b9821e7 Generate the PERUSE XFER_CONTINUE event. 
This commit was SVN r10501.
 2006-06-26 19:01:22 +00:00
George Bosilca
53a5d3df0f Remove useless lines. 
This commit was SVN r10500.
 2006-06-26 19:00:37 +00:00
George Bosilca
a514cdc068 Always limit the size of the RDMA transfer to the maximum amount supported 
by the BTL (btl_max_rdma_size). Now the PUT protocol is pipelined even if there
is just one network between the 2 peers. Unfortunately, this problem is present
the 1.1 (no pipeline for the PUT protocol).

This commit was SVN r10499.
 2006-06-26 19:00:07 +00:00
George Bosilca
8cd4718198 Generate the PERUSE PERUSE_COMM_REQ_XFER_BEGIN event only when there is some 
data to transfer.

This commit was SVN r10498.
 2006-06-26 18:57:55 +00:00
George Bosilca
31365fa799 Use the RDMA limit not the eager one when we schedule a receive (for the 
PUT protocol).

This commit was SVN r10456.
 2006-06-21 15:51:56 +00:00
George Bosilca
ec28040c58 Remove all useless assignment (now they are done inside the macro). Protect one 
call to the _UNPACK macro, in the case where the length of the received data
is zero. This might happens on the PUT protocol.

This commit was SVN r10431.
 2006-06-20 14:16:52 +00:00
George Bosilca
f38480f1d1 Set the recv_bytes value in all the cases. Somehow the PERUSE macro 
contained an error, so now it hould be back again.

This commit was SVN r10430.
 2006-06-20 14:14:04 +00:00
George Bosilca
dee2a7a08d On this branch the rdma_offset should be set. The send_offset is *anyway* 
already set in the _START macro.

This commit was SVN r10429.
 2006-06-20 14:12:32 +00:00
Brian Barrett
c9e8dbc10e * fix for multi-nic case with put protocol -- index will be 1 for the first 
put request if we have more than one nic

This commit was SVN r10397.
 2006-06-16 22:25:04 +00:00
George Bosilca
27000ef7d6 More compact and readable code. Otherwise, no big difference with the 
previous version.

This commit was SVN r10389.
 2006-06-16 03:07:42 +00:00
George Bosilca
3f96f39e46 If the goal of this code was to copy the iovec and skip the first offset 
bytes then it was not correct.

This commit was SVN r10388.
 2006-06-16 03:06:30 +00:00
George Bosilca
93afe59226 It is not required to initialize the csum. 
This commit was SVN r10387.
 2006-06-16 03:05:20 +00:00
George Bosilca
1f96768b76 For zero length persistent request do not reposition the convertor as 
it is not initialized.

This commit was SVN r10386.
 2006-06-16 03:04:41 +00:00
Galen Shipman
5d71c149c2 Another fix for PML request completion when local network completion can occur 
out of order.. 

Reviewed by Brian.. needs to hit 1.1 

This commit was SVN r10353.
 2006-06-14 16:55:35 +00:00
George Bosilca
3727fa2ae6 Nothing relevant. I add some more output in the case we have a checksum error. 
Just to be able to know more information about the failure.

This commit was SVN r10337.
 2006-06-13 19:36:38 +00:00
Galen Shipman
0eddad6849 Handle out of order completion/receives when marking completion... 
this is a fix for #107... needs to go to the 1.1 branch.. 

This commit was SVN r10331.
 2006-06-13 16:57:41 +00:00
Galen Shipman
218a438509 finished the ompi_free_list_t class nightmare.. 
This commit was SVN r10314.
 2006-06-12 22:09:03 +00:00
Galen Shipman
18dda70fd0 make ompi_free_list_item_t a class.. 
This will go to the 1.1 branch but will probably require a few changes as
ompi_free_list_t is different in the branch.. 

This commit was SVN r10306.
 2006-06-12 16:44:00 +00:00
Gleb Natapov
48d348b577 Don't complete send request before we've got completion on the first rndv packet. 
Sender can receive and complete PUT request before it gets completion on the first rndv packet. senreq struct may be reused for the next MPI_Send and unexpected completion mess up the things. I sometimes got SEGV and sometimes data corruption.

This commit was SVN r10301.
 2006-06-12 14:00:43 +00:00
George Bosilca
7d1feffbf7 The real solution. If the sendreq->req_send.req_bytes_packed is zero then there 
is no data to be trasfered. And this is the condition which lead to a non
initialized convertor.

This commit was SVN r10299.
 2006-06-12 06:18:18 +00:00
George Bosilca
c959c2f214 Don't reset the convertor's position if it wasn't initialized before. This can 
only happens for zero byte persistent requests.

This commit was SVN r10298.
 2006-06-12 06:14:35 +00:00
Galen Shipman
84479d0b5a potential fix for iprobe test,, tested with openib.. will have andy try ud.. 
This commit was SVN r10232.
 2006-06-06 22:10:41 +00:00
Brian Barrett
c70fff6ed0 * Fix for bug #44 for the trunk -- remove a bunch of warnings from the DR 
PML when compiling on Solaris.  Patch won't apply cleanly to the v1.1
  branch, so a diff for that is coming up soon.

This commit was SVN r10173.
 2006-06-01 18:58:38 +00:00
George Bosilca
95d0395578 I'm skeptical about the ability of the compiler to correctly optimize the 
loop local variables.

This commit was SVN r10019.
 2006-05-23 03:21:15 +00:00
George Bosilca
b8ef0cc749 Minor cleanups. 
This commit was SVN r10001.
 2006-05-21 05:55:21 +00:00
Galen Shipman
9165882c07 fixes for failover... 
This commit was SVN r9998.
 2006-05-20 02:39:05 +00:00
Brian Barrett
db03ca0cc0 rip out a bunch of code that didn't work and really sucked and was only there 
to try to get some numbers that I couldn't actually get.  So back to the
restart point.

This commit was SVN r9914.
 2006-05-14 00:59:40 +00:00
Brian Barrett
91086cf2a4 * we want to unlink match entries when we unlink memory descriptors, but 
I want to be lazy and not do it by hand, so set the match entries to
  PTL_UNLINK.

This commit was SVN r9861.
 2006-05-09 14:20:51 +00:00
Brian Barrett
8397a1d71f still running into issues, but... 
- change MASK behavior for tags - we need the upper bit to be whether
  the tag is reseved or not.  MPI_ANY_TAG should not pull off any
  reserved tag communication
- some other random debugging output to try to get some idea what is
  spewing out of here.

This commit was SVN r9844.
 2006-05-08 09:23:09 +00:00
George Bosilca
e658557d52 Move the convertor creation out of th critical path. If we expect a 
message from a known peer (not MPI_ANY_SOURCE) then we can attach the
remote proc and initialize the convertor as soon as we know the data-type,
and the count (so basically in the _INIT macro). If it's not the case, then
create them in the _MATCHED macro (as in the original version). Of course,
beforeinitializing the convertor we check that there will be some data
in the message.

This commit, plus the convertor improvements from few days ago, lower the
latency for my test case environment (mvapi) by 0.1 microseconds. The convertor
now is as slim as it can be, I don't think there is anything else to
remove/improve. 

This commit was SVN r9843.
 2006-05-07 21:03:12 +00:00
George Bosilca
a7542824ed Generic length computation (moved from the endpoint.h). 
This commit was SVN r9842.
 2006-05-07 20:54:44 +00:00
George Bosilca
569b88e093 The endpoint include is not required. 
This commit was SVN r9841.
 2006-05-07 20:52:55 +00:00
George Bosilca
e63c1dc242 The last commit wans't supposed to bring this function in. It's not yet 
ready for primetime...

This commit was SVN r9840.
 2006-05-07 20:51:43 +00:00
George Bosilca
33aa65f894 Remove useless include. 
This commit was SVN r9839.
 2006-05-07 20:49:45 +00:00
Rainer Keller
0f9b10ff8e - Update test dup MPI_COMM_WORLD -- so that we may 
have additional Barriers for output.

This commit was SVN r9831.
 2006-05-05 07:42:33 +00:00
Rainer Keller
71d328c086 - Add the PERUSE_COMM_REQ_XFER_CONTINUE for recv. 
This commit was SVN r9820.
 2006-05-04 19:31:33 +00:00
Tim Woodall
d8ff8010f3 track wether the vfrag is being retransmitted 
This commit was SVN r9817.
 2006-05-04 17:30:58 +00:00
Tim Woodall
1b26caa95b first cut at btl failover - seems to be working for simple test case 
This commit was SVN r9816.
 2006-05-04 16:16:26 +00:00
Brian Barrett
d101e91b97 * fix matching logic - since tag might be negative, need to mask the proper bits 
or the bit-wise or changes all the high bits, which is bad
* push convertor creation to init to save a bit of time
* make debugging use macros so that it can go bye-bye

This commit was SVN r9810.
 2006-05-04 13:48:32 +00:00
Brian Barrett
6db0f2a027 * couple of corrections to compile on Red Storm 
This commit was SVN r9801.
 2006-05-03 13:13:59 +00:00
Brian Barrett
4add400f7d * properly start with the memory descriptor inactive 
This commit was SVN r9787.
 2006-05-01 20:23:38 +00:00