Brock Palen wrote:
I was asked by mirycom to run a test using the data reliability pml.
(dr) I ran it like so:
$ mpirun --mca pml dr -np 4 ./xhpl
Is this the right format for running the dr pml?
This should be fine, yes.
I can running HPL on our test cluster to see if something is wr
Lydia Heck wrote:
Thank you very much.
I tried
mpirun -np 6 -machinefile ./myh -mca pml cm ./b_eff
What was the performance (latency and bandwidth)?
and to amuse you
mpirun -np 6 -machinefile ./myh -mca btl mx,sm,self ./b_eff
Same question here as well..
Thanks,
Galen
with myh c
m2001(120) > mpirun -np 6 -hostfile hostsfile -mca btl mx,self b_eff
This does appear to be a bug, although you are using the MX BTL. Our
higher performance path is the MX MTL. To use this try:
mpirun -np 6 -hostfile hostsfile -mca pml cm b_eff
Also, just for grins, could you try:
mpi
Galen M. Shipman wrote:
Gleb Natapov wrote:
On Mon, Oct 30, 2006 at 11:45:53AM -0700, Troy Telford wrote:
On Sun, 29 Oct 2006 01:34:06 -0700, Gleb Natapov
wrote:
If you use OB1 PML (default one) it will never recover from link down
error no matter how many other
Gleb Natapov wrote:
On Mon, Oct 30, 2006 at 11:45:53AM -0700, Troy Telford wrote:
On Sun, 29 Oct 2006 01:34:06 -0700, Gleb Natapov
wrote:
If you use OB1 PML (default one) it will never recover from link down
error no matter how many other transports you have. The reason is that
OB1
Jayanta,
What is your bus on this machine? If it is PCI-X 133 you are going to be
limited, also memory bandwidth could also be the bottleneck.
Thanks,
Galen
Jayanta Roy wrote:
Hi,
In between two nodes I have dual Gigabit ethernet full duplex links. I was
doing benchmarking using non-blo
I think the message was missing from one of the releases but is now
in the trunk
- Galen
On Aug 10, 2006, at 3:44 PM, Andrew Friedley wrote:
Donald Kerr wrote:
Hey Andrew I have one for you...
I get the following error message on a node that does not have any
IB cards
-
Durga,
Currently there are two options for porting an interconnect to Open
MPI, one would be to use the BTL interface (Byte Transfer Layer).
Another would be to use the MTL (Matching Transport Layer). The
difference is that the MTL is useful for those APIs which expose
matching and other
m an OS X
run. I'm about to test
against 1.0.3a1r10670.
Justin.
On 7/6/06, *Galen M. Shipman* < gship...@lanl.gov
<mailto:gship...@lanl.gov>> wrote:
Justin,
Is the OS X run showing the same residual failure?
peedy response,
Justin.
On 7/6/06, Galen M. Shipman < gship...@lanl.gov> wrote:
Hey Justin,
Please provide us your mca parameters (if any), these could be in a
config file, environment variables or on the command line.
Thanks,
Galen
On Jul 6, 2006, at 9:22 AM, Justin Bronder wrote:
As
Hey Justin,
Please provide us your mca parameters (if any), these could be in a
config file, environment variables or on the command line.
Thanks,
Galen
On Jul 6, 2006, at 9:22 AM, Justin Bronder wrote:
As far as the nightly builds go, I'm still seeing what I believe to be
this problem in
I'm currently working with Owen on this issue.. will continue my
findings on list..
- Galen
On Jun 29, 2006, at 7:56 AM, Jeff Squyres ((jsquyres)) wrote:
Owen --
Sorry, we all fell [way] behind on e-mail because many of us were
at an
OMPI developer's meeting last week. :-(
In the interi
Hi Brock,
You may wish to try running with the runtime option:
-mca mpi_leave_pinned 1
This turns on registration caching and such..
- Galen
On Jun 13, 2006, at 8:01 AM, Brock Palen wrote:
I ran a test using openmpi-1.0.2 on OSX vs mpich-1.2.6 from
mryicom and i get lacking results from
Hi Scott,
I believe this has been corrected on the trunk. This should hit the
1.1 release branch tonight.
Thanks,
Galen
On May 9, 2006, at 10:27 AM, Scott Weitzenkamp (sweitzen) wrote:
Pallas runs OK up to Alltoall test, then we get:
/data/software/qa/MPI/openmpi-1.1a2-rhel4-`uname -m`-
Hi Bernie,
You may specify which BTLs to use at runtime using an mca parameter:
mpirun -np 2 -mca btl self,mvapi ./my_app
This specifies to only use self (loopback) and mvapi.
You may want to also use sm (shared memory) if you have multi-core or
multi-proc.. such as:
mpirun -np 2 -mca btl s
Hi Jean,
Take a look here: http://www.open-mpi.org/faq/?category=infiniband#ib-
leave-pinned
This should improve performance for micro-benchmarks and some
applications.
Please let mw know if this doesn't solve the issue.
Thanks,
Galen
On Mar 16, 2006, at 10:34 AM, Jean Latour wrote:
Hel
Emanuel,
Thanks for the tip on this issue, we will be adding it to the FAQ
shortly.
- Galen
On Mar 15, 2006, at 4:29 PM, Emanuel Ziegler wrote:
Hi Davide!
You are using the -prefix option. I guess this is due to the fact
that You
cannot set the paths appropriately. Most likely You are
report?
Galen
mpirun -v -d -hostfile hostfile -np 2 -prefix /usr/local mpi_test
Thank you so much for your help!
Davide
Galen M. Shipman wrote:
Hi Davide,
Are you able to run this as root? This would tell me if this is a
permissions issue.
Also what options are you specifying to mpirun?
Tha
Hi Davide,
Are you able to run this as root? This would tell me if this is a
permissions issue.
Also what options are you specifying to mpirun?
Thanks,
Galen
On Mar 15, 2006, at 1:41 PM, Davide Bergamasco wrote:
Hello,
I'm trying to run OpenMPI on top of OpenIB but I'm running into a
wei
I've been working for the MVAPICH project for around three years.
Since
this thread is discussing MVAPICH, I thought I should post to this
thread. Galen's description of MVAPICH is not accurate. MVAPICH uses
RDMA for short message to deliver performance benefits to the
applications. However, i
This is very helpfull, I will try to obtain a system wired for dual port
in order to correct this.
Thanks,
Galen
On Tue, 7 Feb 2006, Jean-Christophe Hugly wrote:
On Thu, 2006-02-02 at 21:49 -0700, Galen M. Shipman wrote:
I suspect the problem may be in the bcast
Hi Konstantin,
MPI_Alltoall_Isend_Irecv
This is a very unscalable algorithm in skampi as it simply posts N
MPI_Irecv's and MPI_Isend's and then does a Waitall. We shouldn't
have an issue though on 8 procs but in general I would expect the
performance of this algorithm to degrade quite q
Hello Konstantin,
By using coll_basic_crossover 8 you are forcing all of your
benchmarks to use the basic collectives, which offer poor
performance. I ran the skampi Alltoall benchmark with the tuned
collectives I get the following results which seem to scale quite
well, when I have a bit
isolate the problem.
Thanks,
Galen
On Feb 2, 2006, at 7:04 PM, Jean-Christophe Hugly wrote:
On Thu, 2006-02-02 at 15:19 -0700, Galen M. Shipman wrote:
Is it possible for you to get a stack trace where this is hanging?
You might try:
mpirun -prefix /opt/ompi -wdir `pwd` -machinefile /root
Hi Jean,
I just noticed that you are running Quad proc nodes and are using:
bench1 slots=4 max-slots=4
in your machine file and you are running the benchmark using only 2
processes via:
mpirun -prefix /opt/ompi -wdir `pwd` -machinefile /root/machines -
np 2 PMB-MPI1
By using slots=4 y
Jean,
I am not able to reproduce this problem on a non-threaded build, can
you try taking a fresh src package and configuring without thread
support. I am wondering if this is simply a threading issue. I did
note that you said you configured both with and without threads but
try the conf
Hi Daryl,
Sounds like this might be a ulimit issue, what do you get when you
run ulimit -l?
Also, check out: http://www.open-mpi.org/faq/?category=infiniband
Thanks,
Galen
On Dec 6, 2005, at 10:46 AM, Daryl W. Grunau wrote:
Hi, I'm running OMPI 1.1a1r8378 on 2.6.14 + recent OpenIB stack a
Bernard,
This code is using MPI_Alloc_mem, which is good. Do you have an idea of
approx. how much memory has been allocated via MPI_Alloc_mem at the time
of failure?
Thanks,
Galen
On Mon, 21 Nov 2005, Borenstein, Bernard S wrote:
Things have improved alot since I ran the code using the e
The bad:
OpenIB frequently crashes with the error:
***
[0,1,2][btl_openib_endpoint.c:
135:mca_btl_openib_endpoint_post_send] error posting send request
errno says Operation now in progress[0,1,2d
[0,1,3][btl_openib_endpoint.c:
135:mca_btl_openib_endpoint_post_send] error posting s
On Nov 8, 2005, at 6:10 PM, Troy Telford wrote:
I decided to try OpenMPI using the 'openib' module, rather than
'mvapi'; however I'm having a bit of difficulty:
The test hardware is the same as in my earlier posts, the only
software difference is:
Linux 2.6.14 (OpenIB 2nd gen IB drivers)
On Oct 31, 2005, at 8:50 AM, Mike Houston wrote:
When only sending a few messages, we get reasonably good IB
performance,
~500MB/s (MVAPICH is 850MB/s).
What is your message size? Are you using the leave pinned option? If
not, specify -mca mpi_leave_pinned 1 option to mpirun. This tells
O
Correction: HPL_NO_DATATYPE should be: HPL_NO_MPI_DATATYPE.
- Galen
On Oct 25, 2005, at 10:13 AM, Galen M. Shipman wrote:
Hi Troy,
Sorry for the delay, I am now able to reproduce this behavior when I
do not specify HPL_NO_DATATYPE. If I do specify HPL_NO_DATATYPE the
run completes. We will
Hi Troy,
Sorry for the delay, I am now able to reproduce this behavior when I
do not specify HPL_NO_DATATYPE. If I do specify HPL_NO_DATATYPE the
run completes. We will be looking into this now.
Thanks,
Galen
On Oct 21, 2005, at 5:03 PM, Troy Telford wrote:
I've been trying out the RC4
Hi Peter,
I'm a little surprised that tcp was used -- OMPI should "prefer" the
low latency interconnects (such as mvapi) to tcp and automatically
use
them.
This is a small issue and should be fixed in the next day or two. In
the meantime to run using mvapi only use:
mpirun -np 2 -mca
34 matches
Mail list logo
