Fixed in https://svn.open-mpi.org/trac/ompi/changeset/15850. I also
added much more descriptive error messages. FWIW: I've refrained
from adding text to the FAQ yet about the receive_queues MCA
parameter because we will likely be changing some of the parameters
due to changes in credit
#$@#$ I see what happened. The help string for the MCA parameter
said "colon" (and that's the string that I sent out last week and
pasted into the code yesterday), but the parser was actually checking
for semicolons.
On Aug 14, 2007, at 9:57 AM, Jeff Squyres wrote:
Looks like the
Looks like the parsing is wrong for the QP specification.
@#!@#$@!#$ I swear I checked this before I committed, but apparently
I must have goofed. I'll go fix now...
On Aug 14, 2007, at 9:44 AM, Sven Stork wrote:
On Tuesday 14 August 2007 15:23, Tim Prins wrote:
This might be breaking
On Tuesday 14 August 2007 15:23, Tim Prins wrote:
> This might be breaking things on odin. All our 64 bit openib mtt tests
> have the following output:
>
> [odin003.cs.indiana.edu:30971] Wrong QP specification (QP 0
> "P,128,256,128,16:S,1024,256,128,32:S,4096,256,128,32:S,65536,256,128,32").
This might be breaking things on odin. All our 64 bit openib mtt tests
have the following output:
[odin003.cs.indiana.edu:30971] Wrong QP specification (QP 0
"P,128,256,128,16:S,1024,256,128,32:S,4096,256,128,32:S,65536,256,128,32").
Point-to-point QP get 1-5 parameters
However, on my debug