I feel like I should know, but what's a cartofile? I guess you supply
"topological" information about a host, but I can't tell how this
information is used by, say, mpirun.
Dennis Luxen wrote:
In MPI, you must complete every MPI_Isend by MPI_Wait on the request
handle
(or a variant like MPI_Waitall or MPI_Test that returns TRUE). An
un-completed MPI_Isend leaves resources tied up.
Good point, but that doesn't seem to help. I augmented each MPI_Isend
with a MPI
guosong wrote:
I am new to open MPI
I am not, but I'm not real familiar with thread-safe MPI programming.
Still, I saw no other replies to your question, so I'll make an attempt
to answer.
MPI does not guarantee thread safety. E.g., see
http://www.mpi-forum.org/docs/mpi-20-html/node162.h
A. Austen wrote:
On Fri, 28 Aug 2009 10:16 -0700, "Eugene Loh"
wrote:
Big topic and actually the subject of much recent discussion. Here are
a few comments:
1) "Optimally" depends on what you're doing. A big issue is making
sure each MPI pr
Big topic and actually the subject of much recent discussion. Here are
a few comments:
1) "Optimally" depends on what you're doing. A big issue is making
sure each MPI process gets as much memory bandwidth (and cache and other
shared resources) as possible. This would argue that processes
Richard Treumann wrote:
Guess I should have kept quiet a bit longer. As I recall we had
already seen a counter example to Jeff's stronger statement and that
motivated my narrower one.
If there are no wildcard receives - every
MPI_Barrier call is semantica
Jeff Squyres wrote:
On Aug 24, 2009, at 1:03 PM, Eugene Loh wrote:
E.g., let's say P0 and P1 each send a message to P2, both using the
same tag and communicator. Let's say P2 does two receives on that
communicator and tag, using a wildcard source. So, the messages
could be r
Going back to this thread from earlier this calendar year...
Ganesh wrote:
Hi Dick,
Jeff paraphrased an unnamed source as suggesting that: "any
MPI program that relies on a barrier for correctness is an incorrect
MPI application." . That is probably too strong.
How about thi
Julia He wrote:
does the problem show up even if you turn MPI off?
My answer: Yes and No. Yes, when I used ifort compiler. But the problem
did not show up when I used gfortran, which was much slower.
Okay. So, if this is not an MPI problem, as m
Is the problem independent of the the number of MPI processes? (You
suggest this is the case.)
If so, does this problem show up even with np=1 (a single MPI process)?
If so, does the problem show up even if you turn MPI off?
If so, the problem would seem to be unrelated to the MPI implement
Lee Amy wrote:
I build a Kerrighed Clusters
Like Lenny, I'm not familiar with such clusters, but...
with 4 nodes so they look like a big SMP
machine. every node has 1 processor with dingle core.
1) Dose MPI programs could be running on such kinds of machine? If
yes, could anyone show me som
jody wrote:
But can you explain what the meaning of the max-slots entry is?
I checked the FAQs
http://www.open-mpi.org/faq/?category=running#simple-spmd-run
http://www.open-mpi.org/faq/?category=running#mpirun-scheduling
but i couldn't find any explanation. (furthermore, in the FAQ it says
"ma
amjad ali wrote:
Please tell me that if have multiple such ISEND-RECV squrntially for
sending receiving data then DO we need to declare separate
status(MPI_STATUS_SIZE) with for example status1, status2, ; or a
single declaration of it will work for all??
First of all, it really is good
amjad ali wrote:
I am parallelizing a CFD 2D code in
FORTRAN+OPENMPI. Suppose
that the grid (all triangles) is partitioned among 8 processes using
METIS.
Each process has different number of neighboring processes. Suppose
each
process has n elements/faces whose data it need
I was away on vacation for two weeks and therefore missed most of this
thread, but I'm quite interested.
Michael Di Domenico wrote:
>I'm not sure I understand what's actually happened here. I'm running
>IMB on an HP superdome, just comparing the PingPong benchmark
>
>HP-MPI v2.3
>Max ~ 700-800
Manfred Muecke wrote:
I would like to understand in more detail how much time some collective
communication calls really spend waiting for the last process to enter. I
know this can be done by logging entry times for each process, but I
wonder if there is a better and more efficient way.
"Bette
Shaun Jackman wrote:
2 calls MPI_Test. No message is waiting, so 2 decides to send.
2 sends to 0 and does not block (0 has one MPI_Irecv posted)
3 calls MPI_Test. No message is waiting, so 3 decides to send.
3 sends to 1 and does not block (1 has one MPI_Irecv posted)
0 calls MPI_Test. No messag
Shaun Jackman wrote:
Eugene Loh wrote:
Shaun Jackman wrote:
For my MPI application, each process reads a file and for each line
sends a message (MPI_Send) to one of the other processes determined
by the contents of that line. Each process posts a single MPI_Irecv
and uses
Nifty Tom Mitchell wrote:
On Thu, Jun 25, 2009 at 08:37:21PM -0400, Jeff Squyres wrote:
Subject: Re: [OMPI users] 50%performance reduction due to OpenMPI v 1.3.2forcing
allMPI traffic over Ethernet instead of using Infiniband
While the previous thread on "performance r
Shaun Jackman wrote:
For my MPI application, each process reads a file and for each line
sends a message (MPI_Send) to one of the other processes determined by
the contents of that line. Each process posts a single MPI_Irecv and
uses MPI_Request_get_status to test for a received message. If a
Lenny Verkhovsky wrote:
2. if you are using Open MPI 1.3 you dont have to
use mpi_leave_pinned 1 , since it's a default value
And if you're using "-mca btl self,sm" on a single node, I think
mpi_leave_pinned is immaterial (since it's for openib).
On Thu, Jul 2, 2009 at 4:47 PM, Swamy K
Swamy Kandadai wrote:
Jeff:
I'm not Jeff, but...
Linpack has different characteristics at different problem sizes. At
small problem sizes, any number of different overheads could be the
problem. At large problem sizes, one should approach the peak
floating-point performance of the machin
it depends...
On Mon, Jun 29, 2009 at 2:28 PM, Eugene Loh wrote:
Open MPI *does* have VampirTrace in it...
Okay, yeah. You'd want to build OMPI with VT, but then need to find a
tool to read the data.
I thought I had sent e-mail out on this subject before, but cannot find it.
Ot
Jim Kusznir wrote:
Ok, trying now. This does bring up a question I've been meaning to
ask. I need to find a way to analyze the efficiency of parallel jobs,
and the only way I've known about in the past is through vampire. I
do not have the software license presently (and last I looked, it was
George Bosilca wrote:
MPI does not impose any global order on the messages. The only
requirement is that between two peers on the same communicator the
messages (or at least the part required for the matching) is
delivered in order. This make both execution traces you sent with
your origin
Mark Bolstad wrote:
I have a small test code that I've managed to duplicate the results
from a larger code. In essence, using the sm btl with ISend, I wind up
with the communication being completely serialized, i.e., all the
calls from process 1 complete, then all from 2, ...
I need to do so
Mark Bolstad wrote:
I'll post the test code if requested (this email is already long)
Yipes, how long is the test code? Short enough to send, yes? Please send.
tsi...@coas.oregonstate.edu wrote:
Thanks for the explanation. I am using GigEth + Open MPI and the
buffered MPI_BSend. I had already noticed that top behaved
differently on another cluster with Infinibandb + MPICH.
So the only option to find out how much time each process is waiting
arou
Roman Martonak wrote:
I would like to profile the MPI code using the vampir trace integrated
in openmpi-1.3.2. In order to visualize
the trace files, apart from commercial vampir, is there some free
viewer for the OTF files ?
I'm rusty on this stuff.
If you go to http://www.paratools.com/ot
HPL can "stress test" the MPI, but it is typically relatively
insensitive to MPI performance. The usual use produces a measure of the
peak floating-point performance of the system.
A broader set of system performance measurements are found in the HPCC
(HPC Challenge) tests, which include HPL.
vasilis wrote:
The original issue, still reflected by the subject heading of this e-mail,
was that a message overran its receive buffer. That was fixed by using
tags to distinguish different kinds of messages (res, jacob, row, and col).
I thought the next problem was the small (10^-
vasilis wrote:
On Wednesday 27 of May 2009 8:35:49 pm Eugene Loh wrote:
At the level of this particular e-mail thread, the issue seems to me to
be different. Results are added together in some arbitrary order and
there are variations on order of 10^-10. This is not an issue of
George Bosilca wrote:
This is a problem of numerical stability, and there is no solution
for such a problem in MPI. Usually, preconditioning the input
matrix improve the numerical stability.
At the level of this particular e-mail thread, the issue seems to me to
be different. Results are
vasilis wrote:
Rank 0 accumulates all the res_cpu values into a single array, res. It
starts with its own res_cpu and then adds all other processes. When
np=2, that means the order is prescribed. When np>2, the order is no
longer prescribed and some floating-point rounding variations
vasilis wrote:
Thank you Eugene for your suggestion. I used different tags for each variable,
and now I do not get this error.
The problem now is that I am getting a different solution when I use more than
2 CPUs. I checked the matrices and I found that they differ by a very small
amount of
mtcreekm...@broncs.utpa.edu wrote:
I and new to OpenMPI and have looked for this, but not getting
anywhere.
I got an example program off this site:
The compute PI in Fortran example.
http://www.mcs.anl.gov/research/projects/mpi/usingmpi/examples/simplempi/main
vasilis wrote:
Dear openMpi users,
I am trying to develop a code that runs in parallel mode with openMPI (1.3.2
version). The code is written in Fortran 90, and I am running on a cluster
If I use 2 CPU the program runs fine, but for a larger number of CPUs I get the
following error:
[com
Simone Pellegrini wrote:
sorry for the delay but I did some additional experiments to found out
whether the problem was openmpi or gcc!
The program just hangs... and never terminates! I am running on a SMP
machine with 32 cores, actually it is a Sun Fire X4600 X2. (8
quad-core Barcelona AMD
François PELLEGRINI wrote:
users-requ...@open-mpi.org wrote:
Date: Thu, 14 May 2009 17:06:07 -0700
From: Eugene Loh
Subject: Re: [OMPI users] OpenMPI deadlocks and race conditions ?
To: Open MPI Users
Fran?ois PELLEGRINI wrote:
I sometimes run into deadlocks in
François PELLEGRINI wrote:
I sometimes run into deadlocks in OpenMPI (1.3.3a1r21206), when
running my MPI+threaded PT-Scotch software.
So, are there multiple threads per process that perform message-passing
operations?
Other comments below.
Luckily, the case
is very small, with 4 procs onl
Eugene Loh wrote:
Put more strongly, the "correct" (subjective term) way for an MPI
implementation to bind processes is upon process creation and waiting
until MPI_Init is "wrong". This point of view has nothing to do with
asking the MPI implementation to support binding
Ralph Castain wrote:
On May 5, 2009, at 3:37 AM, Geoffroy Pignot wrote:
The result is : everything works fine with MPI executables : logical !!!
What I was trying to do , was to run non MPI exes thanks to mpirun.
There , openmpi is not able to bind these processes to a particular CPU.
My
?
*) How large is N?
Eugene Loh wrote:
Simone Pellegrini wrote:
Dear all,
I have successfully compiled and installed openmpi 1.3.2 on a 8
socket quad-core machine from Sun.
I have used both Gcc-4.4 and Gcc-4.3.3 during the compilation phase
but when I try to run simple MPI programs proce
I'm responsible for some sm changes in 1.3.2, so I can try looking at
this. Some questions below:
Simone Pellegrini wrote:
Dear all,
I have successfully compiled and installed openmpi 1.3.2 on a 8 socket
quad-core machine from Sun.
I have used both Gcc-4.4 and Gcc-4.3.3 during the compilat
Barnabas Debreczeni wrote:
I am using PGAPack as a GA library, and it uses MPI to parallelize
optimization runs. This is how I got to Open MPI.
Let me see if I understand the underlying premise. You want to
parallelize, but there are some large shared tables. There are many
different para
Second cluster is almost the same features with the previous one
From: Eugene Loh
To: Open MPI Users
Sent: Friday, April 24, 2009 1:26:14 AM
Subject: Re: [OMPI users] MPI_Bcast from OpenMPI
So, the remaining mystery is the 6x or so spike at 128 Mbyte. Dunno.
How important is it to resolve that mystery?
Okay. So, going back to Jeff's second surprise, we have 256 Mbyte/2.5
sec = 100 Mbyte/sec = 1 Gbit/sec (sloppy math). So, without getting
into details of what we're measuring/reporting here, there doesn't on
the face of it appear to be anything wrong with the baseline
performance. Jeff was r
jody wrote:
If i get vprof or one of the other tools running, i'll write something up -
perhaps the profiling subject would be worthy for a FAQ entry...
Yes, perhaps. Note that there really are lots of options out there.
Spend a few minutes googling. There are tools, surveys of tools, et
shan axida wrote:
It
have read that FAQ.
Terrific. If you want to offer feedback for its improvement, that
feedback would be welcome.
Does it mean shared memory communication is used when send
messages
between the processes in same node in default?
Yes.
No
need an
memory among processes which are runing in a node?
From: Eugene Loh
To: Open MPI Users
Sent: Thursday, April 23, 2009 1:20:05 PM
Subject: Re: [OMPI users] SHARED Memory
Just to clarify (since "send to self" strikes me as confusing
Just to clarify (since "send to self" strikes me as confusing)...
If you're talking about using shared memory for point-to-point MPI
message passing, OMPI typically uses it automatically between two
processes on the same node. It is *not* used for a process sending to
itself. There is a well
Ankush Kaul wrote:
@gus
we are not able to make hpl sucessfully.
i think it has to do something with blas
i cannot find blas tar file on the net, i found rpm but de
installation steps is with tar file.
First of all, this mail list is for Open MPI issues. On this list are
people who are
Ankush Kaul wrote:
@Eugene
they are ok but we wanted something better, which would more clearly
show de diff in using a single pc and the cluster.
Another option is the NAS Parallel Benchmarks. They are older, but
well known, self-verifying, report performance, and relatively small
and acce
jody wrote:
Hi
I wanted to profile my application using gprof, and proceeded like
when profiling a normal application:
- compile everything with option -pg
- run application
- call gprof
This returns a normal-looking output, but i don't know
whether this is the data for node 0 only or accumulate
Ankush Kaul wrote:
let me describe what i want to do.
i had taken linux clustering as my final year engineering project as i
m really iintrested in 0networking.
to tell de truth our college does not have any professor with
knowledge of clustering.
the aim of our project was just to make a
Shaun Jackman wrote:
Eugene Loh wrote:
What's the rest? I said the shared-memory area is much smaller, but
I was confused about which OMPI release I was using. So, the
shared-memory area was 128 Mbyte and it was getting mapped in once
for each process, and so it was counted doubly.
Eugene Loh wrote:
Shaun Jackman wrote:
What's the purpose of the 400 MB that MPI_Init has allocated?
It's for... um, I don't know. Let's see...
About a third of it appears to be
vt_open() -> VTThrd_open() -> VTGen_open
which I'm guessing is due to the Vam
Shaun Jackman wrote:
Wow. Thanks, Eugene. I definitely have to look into the Sun HPC
ClusterTools. It looks as though it could be very informative.
Great. And, I didn't mean to slight TotalView. I'm just not familiar
with it.
What's the purpose of the 400 MB that MPI_Init has allocated?
Shaun Jackman wrote:
Eugene Loh wrote:
On the other hand, I assume the memory imbalance we're talking
about is rather severe. Much more than 2500 bytes to be
noticeable, I would think. Is that really the situation you're
imagining?
The memory imbalance is drastic. I'm ex
On Apr 14, 2009, at 12:02 PM, Shaun Jackman wrote:
Assuming the problem is congestion and that messages are backing up, ...
I'd check this assumption first before going too far down that path.
You might be able to instrument your code to spit out sends and
receives. VampirTrace (and PERU
Shaun Jackman wrote:
Eugene Loh wrote:
At 2500 bytes, all messages will presumably be sent "eagerly" --
without waiting for the receiver to indicate that it's ready to
receive that particular message. This would suggest congestion, if
any, is on the receiver side. Some kin
Ankush Kaul wrote:
Finally, after mentioning the hostfiles the cluster is working fine.
We downloaded few benchmarking softwares but i would like to know if
there is any GUI based benchmarking software so that its easier to
demonstrate the working of our cluster while displaying our cluster.
Shaun Jackman wrote:
Eugene Loh wrote:
I'm no expert, but I think it's something like this:
1) If the messages are short, they're sent over to the receiver. If
the receiver does not expect them (no MPI_Irecv posted), it buffers
them up.
2) If the messages are long, only
Shaun Jackman wrote:
When running my Open MPI application, I'm seeing three processors that
are using five times as much memory as the others when they should all
use the same amount of memory. To start the debugging process, I would
like to know if it's my application or the Open MPI library
Steve Kargl wrote:
I can rebuild 1.2.9 and 1.3.1. Is there any particular configure
options that I should enable/disable?
I hope someone else will chime in here, because I'm somewhat out of
ideas. All I'm saying is that 10-usec latencies on sm with 1.3.0 or
1.3.1 are out of line with what o
Iain Bason wrote:
But maybe Steve should try 1.3.2 instead? Does that have your
improvements in it?
1.3.2 has the single-queue implementation and automatic sizing of the sm
mmap file, both intended to fix problems at large np. At np=2, you
shouldn't expect to see much difference.
And th
Iain Bason wrote:
There are a bunch changes in the shared memory module between 1.2.9
and 1.3.1. One significant change is the introduction of the "sendi"
internal interface. I believe George Bosilca did the initial
implementation. This is just a wild guess, but maybe there is
somethin
Steve Kargl wrote:
I recently upgraded OpenMPI from 1.2.9 to 1.3 and then 1.3.1.
One of my colleagues reported a dramatic drop in performance
with one of his applications. My investigation shows a factor
of 10 drop in communication over the memory bus. I've placed
a figure that iilustrates the
Ah. George, you should have thought about that. I understand your
eagerness to share this exciting news, but perhaps an April-1st
announcement detracted from the seriousness of this grand development.
Here's another desirable MPI feature. People talk about "error
detection/correction". We
Jeff Squyres wrote:
On Mar 28, 2009, at 5:31 AM, François PELLEGRINI wrote:
MPI_Comm_size (MPI_COMM_WORLD, &procglbnbr);
MPI_Comm_rank (MPI_COMM_WORLD, &proclocnum);
dataloctab = malloc (2 * (procglbnbr + 1) * sizeof (int));
dataglbtab = dataloctab + 2;
Why +2? Shouldn't it be +pr
Shaun Jackman wrote:
On Tue, 2009-03-24 at 07:03 -0800, Eugene Loh wrote:
I'm not sure I understand this suggestion, so I'll say it the way I
understand it. Would it be possible for each process to send an
"all done" message to each of its neighbors? Conversely, each
Simon Köstlin wrote:
I'm new to MPI and I've got a question about blocking routines like
the Send-, Wait-Function and so on. I wrote a parallel program that
uses the blocking Send and the Nonblocking Isend function. Now my
question: If I'm sending something with the blocking Send function it
Ashley Pittman wrote:
On 23 Mar 2009, at 23:36, Shaun Jackman wrote:
loop {
MPI_Ibsend (for every edge of every leaf node)
MPI_barrier
MPI_Iprobe/MPI_Recv (until no messages pending)
MPI_Allreduce (number of nodes removed)
} until (no nodes removed by any node)
Previously, I attempted to use
Shaun Jackman wrote:
I've just read in the Open MPI documentation [1]
That's the MPI spec, actually.
that collective operations, such as MPI_Allreduce, may synchronize,
but do not necessarily synchronize. My algorithm requires a collective
operation and synchronization; is there a better (m
Åke Sandgren wrote:
Hi!
Valgrind seems to think that there is an use of uninitialized value in
mca_mpool_sm_alloc, i.e. the if(mpool_sm->mem_node >= 0) {
Backtracking that i found that mem_node is not set during initializing
in mca_mpool_sm_init.
The resources parameter is never used and the mp
Jeff Squyres wrote:
If you're exchanging data at the end of an iteration, then you
effectively have a synchronization anyway -- no need for an extra
barrier synchronization.
Ralph Castain wrote:
Ummmnot to put gasoline on the fire, but...if the data exchange
is blocking, why do you n
Mark Allan wrote:
Dear all,
With this simple code I find I am getting a memory leak when I run on 2
processors. Can anyone advise why?
I'm fairly new to MPI (have only done very simple things in the past).
I'm trying to do a non-blocking send
I think the question is about passing NULL as a buffer pointer. E.g.,
MPI_Send(NULL, 0, mytype,dst, tag,comm);
vs
MPI_Send(&dummy,0,mytype,dst,tag,comm);
George Bosilca wrote:
The only restriction the MPI standard impose on the sendcounts and
recvcounts arrays is that the values be non-neg
Prentice Bisbal wrote:
Jeff Squyres wrote:
On Feb 2, 2009, at 4:48 PM, Prentice Bisbal wrote
No. I was running just a simple "Hello, world" program to test v1.3 when
these errors occured. And as soon as I reverted to 1.2.8, the errors
disappeared.
FW
Jeff Squyres wrote:
As you have notes, MPI_Barrier is the *only* collective operation that
MPI guarantees to have any synchronization properties (and it's a
fairly weak guarantee at that; no process will exit the barrier until
every process has entered the barrier -- but there's no guarantee
Jack Bryan wrote:
I need to transfer data from multiple sources to one
destination.
The requirement is:
(1) The sources and destination nodes may work asynchronously.
(2) Each source node generates data package in their own paces.
And, there may be many packages to send. Whenever,
Win Than Aung wrote:
thanks for your reply jeff
so i tried following
#include
#include
int main(int argc, char **argv) {
int np, me, sbuf = -1, rbuf = -2,mbuf=1000;
int data[2];
MPI_Init(&argc,&argv);
MPI_Comm_size(MPI_COMM_WORLD,&np);
MPI_Comm_rank(MPI_COMM
Win Than Aung wrote:
MPI_Recv() << is it possible to receive the message sent from
other sources? I tried MPI_ANY_SOURCE in place of source but it
doesn't work out
Yes of course. Can you send a short example of what doesn't work? The
example should presumably be less than about 20 line
Jeff Squyres wrote:
On Dec 12, 2008, at 11:46 AM, Eugene Loh wrote:
FWIW, I've run into the need for this a few times due to HPCC tests
on large (>100 MPI procs) nodes or multicore systems. HPCC (among
other things) looks at the performance of a single process while all
ot
Jeff Squyres wrote:
FWIW, Open MPI does have on its long-term roadmap to have "blocking"
progress -- meaning that it'll (probably) spin aggressively for a
while and if nothing "interesting" is happening, it'll go into a
blocking mode and let the process block in some kind of OS call.
Alth
douglas.gupt...@dal.ca wrote:
Proceeding from that, it seems that "mpi_recv" is implemented as
"poll forever until the message comes"
and NOT as
"sleep until the message comes"
I had assumed, until now, that mpi_recv would be implemented as the
second.
It isn't a binary situatio
Jens wrote:
Hi Terry,
I would like to run a paraview-server all time on our cluster (even
though it is not in use 24h) - but this would simply result in some kind
of "heating-thread".
Even though it has (in theory) no impact on the node performace (which
is part of a grid-engine), it would sim
Douglas Guptill wrote:
Hi:
I am using openmpi-1.2.8 to run a 2 processor job on an Intel
Quad-core cpu. Opsys is Debian etch. I am reaonably sure that, most
of the time, one process is waiting for results from the other. The
code is fortran 90, and uses mpi_send and mpi_recv. Yet
"gnome-sys
vladimir marjanovic wrote:
In
order to overlap communication and computation I don't want to use
MPI_Wait.
Right. One thing to keep in mind is that there are two ways of
overlapping communication and computation. One is you start a send
(MPI_Isend), you do a bunch of computation
vladimir marjanovic wrote:
I am new user of Open MPI, I've used MPICH before.
There is performance bug with the following scenario:
proc_B: MPI_Isend(...,proc_A,..,&request)
do{
sleep(1);
MPI_Test(..,&flag,&request);
Benjamin Lamptey wrote:
Hello,
I am new at using open-mpi and will like to know something basic.
What is the equivalent of the "mpif.h" in open-mpi which is normally
"included" at
the beginning of mpi codes (fortran in this case).
I shall appreciate that for cpp as well.
For Fortran:
I think the KINDs are compiler dependent. For Sun Studio Fortran,
REAL*16 and REAL(16) are the same thing. For Intel, maybe it's
different. I don't know. Try running this program:
double precision xDP
real(16) x16
real*16 xSTAR16
write(6,*) kind(xDP), kind(x16), kind(xSTAR16), kind(1.0_16)
Sangamesh B wrote:
I reinstalled all softwares with -O3 optimization. Following are the
performance numbers for a 4 process job on a single node:
MPICH2: 26 m 54 s
OpenMPI: 24 m 39 s
I'm not sure I'm following. OMPI is faster here, but is that a result
of MPICH2 slowing down? The or
Brian Dobbins wrote:
On Thu, Oct 9, 2008 at 10:13 AM, Jeff
Squyres
wrote:
On Oct 9, 2008, at 8:06 AM, Sangamesh B wrote:
OpenMPI
: 120m 6s
MPICH2 : 67m 44s
That seems to indicate that something else is going on -- with -np 1,
there
Eugene Loh wrote:
Sangamesh B wrote:
The job is run on 2 nodes - 8 cores.
OpenMPI - 25 m 39 s.
MPICH2 - 15 m 53 s.
I don't understand MPICH very well, but it seemed as though some of
the flags used in building MPICH are supposed to be added in
automatically to the mpicc/etc com
Sangamesh B wrote:
I wanted to switch from mpich2/mvapich2 to OpenMPI, as OpenMPI
supports both ethernet and infiniband. Before doing that I tested an
application 'GROMACS' to compare the performance of MPICH2 & OpenMPI.
Both have been compiled with GNU compilers.
After this benchmark, I cam
I guess this must depend on what BTL you're using. If I run all
processes on the same node, I get the behavior you expect. So, are you
running processes on the same node, or different nodes and, if
different, via TCP or IB?
Gregory D Abram wrote:
I have a little program which initializes,
Aurélien Bouteiller wrote:
You can't assume that MPI_Send does buffering.
Yes, but I think this is what Eric meant by misinterpreting Enrico's
problem. The communication pattern is to send a message, which is
received remotely. There is remote computation, and then data is sent
back. No
Dirk Eddelbuettel wrote:
On 6 September 2008 at 22:13, Davi Vercillo C. Garcia () wrote:
| I'm trying to execute some programs in my notebook (Ubuntu 8.04) using
| OpenMPI, and I always get a warning message like:
|
| libibverbs: Fatal: couldn't read uverbs ABI version.
| -
Jeff Squyres wrote:
OMPI currently polls for message passing progress. While you're in
MPI_BCAST, it's quite possible/ likely that OMPI will poll hard until
the BCAST is done. It is possible that a future version of OMPI will
use a hybrid polling+non- polling approach for progress, such th
201 - 300 of 303 matches
Mail list logo