On Wed, Jan 17, 2007 at 08:55:31AM -0700, Brian W. Barrett wrote:
>On Jan 17, 2007, at 2:39 AM, Gleb Natapov wrote:
>> On Wed, Jan 17, 2007 at 04:12:10AM -0500, Robin Humble wrote:
>>> basically I'm seeing wildly different bandwidths over InfiniBand 4x DDR
>>> when I use different kernels.
>> Try t
argh. attached.
cheers,
robin
On Thu, Jan 18, 2007 at 03:52:19AM -0500, Robin Humble wrote:
>On Wed, Jan 17, 2007 at 08:55:31AM -0700, Brian W. Barrett wrote:
>>On Jan 17, 2007, at 2:39 AM, Gleb Natapov wrote:
>>> On Wed, Jan 17, 2007 at 04:12:10AM -0500, Robin Humble wrote:
basically I'm s
On Thu, Jan 18, 2007 at 03:52:19AM -0500, Robin Humble wrote:
> On Wed, Jan 17, 2007 at 08:55:31AM -0700, Brian W. Barrett wrote:
> >On Jan 17, 2007, at 2:39 AM, Gleb Natapov wrote:
> >> On Wed, Jan 17, 2007 at 04:12:10AM -0500, Robin Humble wrote:
> >>> basically I'm seeing wildly different bandwi
On Thursday 18 January 2007 09:52, Robin Humble wrote:
...
> is ~10Gbit the best I can expect from 4x DDR IB with MPI?
> some docs @HP suggest up to 16Gbit (data rate) should be possible, and
> I've heard that 13 or 14 has been achieved before. but those might be
> verbs numbers, or maybe horsepowe
On Jan 18, 2007, at 5:05 AM, Peter Kjellstrom wrote:
On Thursday 18 January 2007 09:52, Robin Humble wrote:
...
is ~10Gbit the best I can expect from 4x DDR IB with MPI?
some docs @HP suggest up to 16Gbit (data rate) should be possible,
and
I've heard that 13 or 14 has been achieved before.
On Thu, Jan 18, 2007 at 11:08:04AM +0200, Gleb Natapov wrote:
>On Thu, Jan 18, 2007 at 03:52:19AM -0500, Robin Humble wrote:
>> On Wed, Jan 17, 2007 at 08:55:31AM -0700, Brian W. Barrett wrote:
>> >On Jan 17, 2007, at 2:39 AM, Gleb Natapov wrote:
>> >> On Wed, Jan 17, 2007 at 04:12:10AM -0500, Robi
On Thu, Jan 18, 2007 at 07:17:13AM -0500, Robin Humble wrote:
> On Thu, Jan 18, 2007 at 11:08:04AM +0200, Gleb Natapov wrote:
> >On Thu, Jan 18, 2007 at 03:52:19AM -0500, Robin Humble wrote:
> >> On Wed, Jan 17, 2007 at 08:55:31AM -0700, Brian W. Barrett wrote:
> >> >On Jan 17, 2007, at 2:39 AM, Gl
On Thursday 18 January 2007 13:08, Scott Atchley wrote:
...
> The best uni-directional performance I have heard of for PCIe 8x IB
> DDR is ~1,400 MB/s (11.2 Gb/s)
This is on par with what I have seen.
> with Lustre, which is about 55% of the
> theoretical 20 Gb/s advertised speed.
I think this
On Jan 18, 2007, at 8:11 AM, Peter Kjellstrom wrote:
with Lustre, which is about 55% of the
theoretical 20 Gb/s advertised speed.
I think this should be calculated against 16 Gbps, not 20 Gbps.
What is the advertised speed of a IB DDR card?
http://mellanox.com/products/hca_cards.php
http://
Sorry for jumping in late.
I was able to use ~128 SGE slots for my test run, with the either of the
SGE allocation rules ($fill_up or $round_robin) and -np 64 on my test
MPI program, but I wasn't able to reproduce your error though on
Solaris. Like Brian said, having the stack trace could help.
Hi List,
1. We have
HW
* 2xBladecenter H
* 2xCisco Infiniband Switch Modules
* 1xCisco Infiniband Switch
* 16x PPC64 JS21 blades each are 4 cores, with Cisco HCA
SW
* SLES 10
* OFED 1.1 w. OpenMPI 1.1.1
I am running the Intel MPI Benchmark (IMB) on the cluster as a part of
validation process f
Hi,
We tried running with 32 and 16, had some success but after a reboot of
the cluster it seems to be any DLPOLY run attempted falls over, either
interactively or through SGE. Standard benchmarks such as IMB and HPL
execute to completion.
Here's the full output of a typical error:
Signal:7 info
Are you using
-mca pml cm
for pathscale or are you using openib?
- Galen
On Jan 18, 2007, at 4:42 PM, Barry Evans wrote:
Hi,
We tried running with 32 and 16, had some success but after a
reboot of
the cluster it seems to be any DLPOLY run attempted falls over, either
interactively or
Just to answer my own question, after I explicitly specify the "--mca btl
tcp" parameter, the program works. So I will need to issue command like
this:
$ mpirun --mca btl tcp -np 2 tut01
oceanus:Hello world from 0
oceanus:Hello world from 1
Regards,
Eddie.
On 1/18/07, eddie168 wrote:
Hi Ra
Two things:
1. You will definitely need to also specify the "self" BTL.
Otherwise, if you have an MPI process that tries to send to itself,
there will be no [internal OMPI] path for that message to follow, and
an MPI exception will occur.
mpirun --mca btl tcp,self -np 2 tut01
2.
What that parameter does is turn ³off² all of the transports except tcp so
the problem you¹re seeing goes away because we no longer try to create the
shared memory file. This will somewhat hurt your performance, but it will
work.
Alternatively, you could use ³--mca btl ^sm², which would allow yo
16 matches
Mail list logo
