1
1
Граф коммитов

7657 Коммитов

Автор SHA1 Сообщение Дата
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
Brian Barrett
2cf73912e2 * fix for signal forwarding additions in bproc_orted code
This commit was SVN r10529.
2006-06-27 19:59:07 +00:00
Brian Barrett
0031e39d72 * fix for dumb memory bug introduced in romio performance fixup code
This commit was SVN r10528.
2006-06-27 19:58:18 +00:00
Brian Barrett
9a65a7ca97 * re-add -Is necessary for VPATH builds.
This commit was SVN r10524.
2006-06-27 14:10:34 +00:00
Jeff Squyres
df45221a3e Until a real fix for #142 is found, this workaround prohibits using
mpi_leave_pinned when multiple OpenIB HCA ports are found.
Specifically, if mpi_leave_pinned == 1 and ultiple HCA ports are
found, the MCA parameter btl_openib_max_btls is set to 1.  If the MCA
parameter btl_openib_warn_leave_pinned_multi_port is true, emit a
warning that this happened (having an MCA parameter to control the
warning allows users/sysadmins to turn it off instead of being nagged
for every run).

This commit was SVN r10521.
2006-06-27 10:43:03 +00:00
Gleb Natapov
012d95d195 If ompi_free_list_grow fails wait until resources are available instead of
spinning without progress.

This commit was SVN r10520.
2006-06-27 09:23:51 +00:00
Gleb Natapov
52208d7bf9 Whe don't need to register zero sized frags.
This commit was SVN r10519.
2006-06-27 08:50:12 +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
Brian Barrett
b34768962c * put event library tests into the testing infrastructure so that they can
be built without heroic effort

This commit was SVN r10517.
2006-06-26 22:28:59 +00:00
Brian Barrett
970d858f30 * Add performance code requested by LANL, per ticket #128. Must be
explicitly enabled at run-time with the mca parameter
  io_romio_enable_parallel_optimizations set to something non-zero.
  This will enable some magic flags in Panasas if the user didn't
  set them (either on or off) and do some slightly better things
  with strided collective writes.

This commit was SVN r10516.
2006-06-26 22:26:36 +00:00
Brian Barrett
b6663c64c7 * fix for bug #161 - add man page info for recently added features
This commit was SVN r10514.
2006-06-26 22:16:39 +00:00
George Bosilca
379b170a29 Update the datatype tests.
This commit was SVN r10511.
2006-06-26 20:10:27 +00:00
George Bosilca
1ab7dcc632 Cleanups.
This commit was SVN r10509.
2006-06-26 20:09:04 +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
41c886399b Don't let the user to specify flags which does not make sense. If the PUT flag is
specified check that the put function is available for the BTL. Same safe check for
the GET function. At the end make sure that at least on communication protocol is
specified, otherwise force the send flag.

This commit was SVN r10507.
2006-06-26 20:00:18 +00:00
Jeff Squyres
e9b1b63ca3 Update svn:ignore -- had to be a separate commit from when mpif.h was
removed.

This commit was SVN r10504.
2006-06-26 19:33:26 +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
Brian Barrett
86861bc1c3 * add --quiet option, and surpress a couple of the status messages in
orterun if it is actually enabled.  For ticket #129.

This commit was SVN r10497.
2006-06-26 18:21:45 +00:00
Brian Barrett
4e8abb943b * fix up signal handling code so that one function handles SIGUSR1 and
SIGUSR2.  This can be extended later if needed to include other
  signals we should forward to the user processes (TSTP and CONT,
  perhaps?)
* Since the signal handlers don't actually run in signal context, we
  can use malloc/fprintf/etc.  So clean up some of the signal handler
  code so that we don't keep message buffers around for the life of
  the process

This commit was SVN r10496.
2006-06-26 15:12:52 +00:00
Gleb Natapov
b7715395cb Return descriptor before sending credits one more time. We may need it.
This commit was SVN r10495.
2006-06-26 07:05:58 +00:00
Andrew Friedley
7bfac82ce7 Change over from lazy connection setup to setting up at initialization
time.

UD is connectionless, and as long as peers are statically assigned to QPs,
there is no reason to set up the adressing information lazily.

Lots of code was axed, as endpoints no longer have state.  Removed a
number of other elements in the endpoint struct to make it as lightweight
as possible.

I was able to remove an entire function call/branch in the send path,
which I believe is the main contributor to a 2us drop in NetPIPE latency.

Some whitespace cleanups as well.

Passes IBM test suite, and all but certain intel tests that were failing
before the change, over ob1 PML.

This commit was SVN r10494.
2006-06-23 16:50:50 +00:00
Jeff Squyres
f08e54029c - Update svn:ignore
- Built to_self, but don't run it during "make check" (because it
  calls MPI_INIT, which requires a functional install)

This commit was SVN r10491.
2006-06-23 02:14:27 +00:00
Jeff Squyres
eed198f666 Update the AUTHORS list to be a) more comprehensive (some people have
been at multiple institutions) and b) more up-to-date.

This commit was SVN r10486.
2006-06-23 01:54:48 +00:00
Andrew Friedley
046f4cd4ae Enough cleanup for now.
Moved a lot of the module-specific init from the component init to the module init.

Try keeping a pointer to reduce indexing, didn't seem to help - leaving in place
for now.

This commit was SVN r10485.
2006-06-22 22:12:13 +00:00
Brian Barrett
9766c01e50 * Per discussion at quarterly meeting and bug #91, print out the bug
contact point when printing version and help strings

This commit was SVN r10484.
2006-06-22 19:48:27 +00:00
Brian Barrett
7dd1112d07 * implement missing MPI::Is_finalized() function
This commit was SVN r10482.
2006-06-22 19:40:54 +00:00
Andrew Friedley
8392ed4cac A checkpoint before I really do some cleanup.. nothing pretty here.
Playing around with OPAL_LIKELY/UNLIKELY, no real gains yet.

Reworked progress() to process many WC's at a time, as well
as immediately repost groups of receive buffers.

This commit was SVN r10481.
2006-06-22 18:06:55 +00:00
Sushant Sharma
76926756d0 variable ntid not being assigned any value was resulting in errors
This commit was SVN r10480.
2006-06-22 18:00:54 +00:00
George Bosilca
9eb023a5c2 OK my last commit was ... kind of wrong. It only worked if the element_size
was smaller than the CACHE_LINE_SIZE. Here is the version that works.

In fact this works on 2 steps. First we set the element size to something
multiple of the desired alignment. Then when we allocate memory, we compute
the total size, and we will align each of the elements (we allocate
multiple of them every time) to the CACHE_LINE_SIZE.

This commit was SVN r10479.
2006-06-22 14:47:07 +00:00
George Bosilca
c71f6c9765 All elements will be aligned to the CACHE_LINE_SIZE define (currently 128
bytes). The simplest way to make sure they are aligned is to update
the size of the basic element to a multiple of the desired alignment.
It will use a little bit more memory, but the improvements on the SM BTL
seems quite interesting.

This commit was SVN r10478.
2006-06-22 14:07:14 +00:00
Jeff Squyres
9a679644c2 Arf. Don't output the body of the WTICK or WTIME functions in the
module header if we're not doing small.

This commit was SVN r10475.
2006-06-22 13:20:01 +00:00
Jeff Squyres
67b07ba4fc AM complains if we define names with specific suffixes and the
executable name is not listed anywhere -- so just comment them out for
now.

This commit was SVN r10472.
2006-06-22 11:56:18 +00:00
Jeff Squyres
fa6b6c6098 This test calls MPI_INIT -- can't do that in the unit tests because
that assumes that OMPI has been fully installed (e.g., that may not be
valid during "make distcheck")

This commit was SVN r10470.
2006-06-22 11:47:31 +00:00
George Bosilca
90a043da16 Move the Fortran file into the nodist headers.
This commit was SVN r10465.
2006-06-21 21:28:51 +00:00
Jeff Squyres
87ec6c5384 Fix the fix -- if we're not compiling the profiling layer, then we
cannot include the PMPI_WTIME|WTICK functions in the external and
double precision statements because some compilers complain about
this.  Instead, we need to use the macro that is defined by
configure.ac (MPIF_H_PMPI_W_FUNCS).  This unfortunately means that we
need to generate mpif.h (in addition to mpif-config.h) because the
"external" statement is toxic to F90 compilers.

This commit was SVN r10464.
2006-06-21 21:24:01 +00:00
George Bosilca
cde42e68e8 As now the MPI_Wtime and MPI_Wtick are functions do not export the
profiling prototype by default.

This commit was SVN r10463.
2006-06-21 20:05:16 +00:00
Jeff Squyres
723b6e50a9 George suggested a better way to make WTICK and WTIME -- be consistent
with the other methodology even if there are no choice buffers and no
special constants.  But it keeps the Makefile.am simple and the
methodology consistent.

This commit was SVN r10462.
2006-06-21 19:07:09 +00:00
George Bosilca
efb987f156 Output the right message.
This commit was SVN r10457.
2006-06-21 16:25:02 +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
f27591444a Remove one of the internal variable to make things more clear and more
similar with the other pack/unpack functions.

This commit was SVN r10455.
2006-06-21 14:49:41 +00:00
George Bosilca
710a49ce79 Correctly update the flags when we build data-types. Play nicely with the NO_GAP flag.
This commit was SVN r10454.
2006-06-21 14:46:10 +00:00
Jeff Squyres
48e9a72c47 Add the missing files -- they're svn:ignored because of all the
generated files.

This commit was SVN r10451.
2006-06-21 14:11:12 +00:00
George Bosilca
820f103cd9 Remove one of the optimizations, as it lead to non correct data description.
This commit was SVN r10450.
2006-06-21 14:06:52 +00:00
George Bosilca
382a0209f7 Correctly play with the flags. Ported from the 1.1 branch.
This commit was SVN r10449.
2006-06-21 14:05:09 +00:00
Jeff Squyres
720f38efc5 Fix for MPI_WTICK / MPI_WTIME F90 bindings issue. The previous hope
was that declaring the type of MPI_WTICK and MPI_TIME in mpif-common.h
would allow the F90 bindings to call through to the back end f77
function and have the right return type.  But upon reflection, that's
silly -- we were just declaring the variables MPI_WTICK and MPI_WTIME
that were of type double precision.  Duh.

So add some fixed (non-generated) wrapper F90 functions to call the
back-end *C* MPI_WTICK and MPI_TIME functions (vs. the back end *F77*
functions).  We have to call the back-end C functions because there's
a name conflict if we try to call the back-end F77 functions -- for
the same reasons that we can't "implicitly" define MPI_WTIME and
MPI_WTICK in the f90 module, we can't call such an implicitly-defined
function.  So we had to add new back-end C functions that are directly
callable from Fortran, the easiest implementation of which was to
provide 4 one-line functions for each (rather than muck around with
weak symbols).

This commit was SVN r10448.
2006-06-21 13:44:20 +00:00
Andrew Friedley
365c81d6e9 Fix a few issues reported by Terry Dontje:
1. ompi/mca/btl/udapl/btl_udapl_proc.c should be including
btl_udapl_endpoint.h for mca_btl_udapl_proc_insert function.

2. btl_udapl_endpoint.c it looks like you are using
&endpoint->endpoint_lock when you should use &ep->endpoint_lock in a
OPAL_THREAD_LOCK call.

3. btl_udapl_frag.h has a couple opal_list_item_t's that should be
ompi_free_list_item_t in the _FRAG_ALLOC_{EAGER,MAX} macros.

This commit was SVN r10442.
2006-06-20 17:13:44 +00:00
Brian Barrett
03cc22842c fix error message typo and correct the flag that should be provided to gcc
to force v8+ ABI

This commit was SVN r10438.
2006-06-20 16:25:31 +00:00