Jeff,
During the original code review we found ,that by default we allocate
SRQ with size "rd_num + sd_max" but
on the SRQ we post only rd_num receive entries. It means that we do not
fill the queue completely. Looks like a bug.
Pasha
Jeff Squyres wrote:
SRQ hardware vendors -- please review
Hello all.
Jeff - thank you for your comments.
I answer you below and if you want you can see the changes in the new patch.
SRQ hardware vendors -- please review and reply...
More below.
On Dec 2, 2009, at 10:20 AM, Vasily Philipov wrote:
diff -r a5938d9dcada ompi/mca/btl/openib/btl_openi
SRQ hardware vendors -- please review and reply...
More below.
On Dec 2, 2009, at 10:20 AM, Vasily Philipov wrote:
> diff -r a5938d9dcada ompi/mca/btl/openib/btl_openib.c
> --- a/ompi/mca/btl/openib/btl_openib.cMon Nov 23 19:00:16 2009 -0800
> +++ b/ompi/mca/btl/openib/btl_openib.c
The attach patch should resolve the long pending issue that we have on
our track https://svn.open-mpi.org/trac/ompi/ticket/1912.
The issue: As process of OpenIB BTL creation we also create set of SRQs
and corresponding receive fragments are allocated and posted on all
SRQs. It mean that a pro