Re: [OMPI users] Dual core Intel CPU
On Wed, 16 Aug 2006, Allan Menezes wrote: Hi AnyOne, I have an 18 node cluster of heterogenous machines. I used fc5 smp kernel and ocsar 5.0 beta. I tried the following out on a machine with Open mpi 1.1 and 1.1.1b4 versions. The machine consists of a Dlink 1gigb/s DGE-530T etherent card 2.66GHz dual core Intel Cpu Pentium D 805 with Dual Cannel 1 gig DDR 3200 ram. I compiled the ATLAS libs (ver 3.7.13beta) for this machine and HPL (xhpl executable) and ran the following experiment twice: content of my "hosts" file1 for this machine for 1st experiment: a8.lightning.net slots=2 content of my "hosts" file2 for this machine for 2nd experiment: a8.lightning.net On the single node I ran for HPL.dat N =6840 and NB=120 : 1024 MB of Ram N = sqrt(0.75* ((1024-32 video overhead)/2 )*100*1/8)=approx 6840; 512MB Ram per CPU otherwise the OS uses the hard drive for virtaul memory. This way it resides totally in Ram. I ran this command twice for the two different hosts files above in two experiments: # mpirun --prefix /opt/openmpi114 --hostsfile hosts -mca btl tcp, self -np 1 ./xhpl In both cases the performance remains the same around 4.040 GFlops I would expect since I am running slots =2 as two CPU's I would get a performance increase from expt 2 by 100 -50% But I see no difference.Can anybody tell me why this is so? You are only launching 1 process in both cases. Try `mpirun -np 2 ...` to launch 2 processes, which will load each of your processors with an xhpl process. Please read the faq: http://www.open-mpi.org/faq/?category=running#simple-spmd-run It includes a lot of information about slots and how they should be set as well. Hugh I have not tried mpich 2. Thank you, Regards, Allan Menezes ___ users mailing list us...@open-mpi.org http://www.open-mpi.org/mailman/listinfo.cgi/users
[OMPI users] Open Mpi and Dual Core Intel CPU's
Hi AnyOne, Soory about the SPAM. But I tried Mpich2-1.0.4p1 and got ^ Gflops this way. I first tried mpich2 configured with --with-device=ch3 and -with-comm=ch3:shm ans then make, make install. I then tried the first expt. I ran # mpd --ncpus=2 & and #.mpiexec -np 1 -f hosts where hosts is the file defined below in my email. I got the same 4.00GFlops. HPL.dat was set at P=1 Q=1 for 1 node with 2 processors. I then modified HPL.dat to P=1 and Q=2 for the same hosts file of one machine of a18.lightning.net and got 6GigaFlops performance. I shall try the same with open mpi modifying the HPL.dat P's and Q's for open mpi and post my results for open mpi. Remember I am trying this out on a single dual core machine with SMP kernel. For Open Mpi below in both cases with slots=2 or slots=1 I tried the HPL.dat with P=1 and Q=1. I shall now try with P=1 and Q=2 with slots=2 in the hosts file. I have an 18 node cluster of heterogenous machines. I used fc5 smp kernel and ocsar 5.0 beta. I tried the following out on a machine with Open mpi 1.1 and 1.1.1b4 versions. The machine consists of a Dlink 1gigb/s DGE-530T etherent card 2.66GHz dual core Intel Cpu Pentium D 805 with Dual Cannel 1 gig DDR 3200 ram. I compiled the ATLAS libs (ver 3.7.13beta) for this machine and HPL (xhpl executable) and ran the following experiment twice: content of my "hosts" file1 for this machine for 1st experiment: a8.lightning.net slots=2 content of my "hosts" file2 for this machine for 2nd experiment: a8.lightning.net On the single node I ran for HPL.dat N =6840 and NB=120 : 1024 MB of Ram N = sqrt(0.75* ((1024-32 video overhead)/2 )*100*1/8)=approx 6840; 512MB Ram per CPU otherwise the OS uses the hard drive for virtaul memory. This way it resides totally in Ram. I ran this command twice for the two different hosts files above in two experiments: # mpirun --prefix /opt/openmpi114 --hostsfile hosts -mca btl tcp, self -np 1 ./xhpl In both cases the performance remains the same around 4.040 GFlops I would expect since I am running slots =2 as two CPU's I would get a performance increase from expt 2 by 100 -50% But I see no difference.Can anybody tell me why this is so? I have not tried mpich 2. Thank you, Regards, Allan Menezes
[OMPI users] OPEN MPI and Intel dual core cpu
Hi All, Regarding my previous posts good news I get 6 GFlops with open mpi ver 1.b4 with host containing the single line a18.lightning.net slots=2 . This time with the HPL.dat modified to be P=1 Q=2 for two nodes/processes on the same machine( Dual core Pentium D805) My Question: I have 18 heterogenous nodes all x86. Almost all single core(14) and two dual cores and two hyperthreading CPU's what should my P's and Q's be to benchmark the true performance? I am guessing P=4 and Q=6. Am i right? Thank you for your consideration Allan Menezes
Re: [OMPI users] Problem compiling OMPI with Intel C compiler on Mac OS X
Today I ran into the same problem as Warner Yuen (see thread below), openmpi does not compile with icc and fails with an error where libtool ask for --tag. The error is macosx specific. It occurs in the compile for xgrid in openmpi-1.1/orte/mca/pls/xgrid the Makefile fails; xgrid uses some objective-C stuff that needs to be compiled with gcc [I guess] after adjusting the Makefile.in from xgrid>grep -n "\-\-tag=OBJC" Makefile.in 216:LTOBJCCOMPILE = $(LIBTOOL) --mode=compile $(OBJC) $(DEFS) \ 220:OBJCLINK = $(LIBTOOL) --mode=link $(OBJCLD) $(AM_OBJCFLAGS) \ to xgrid>grep -n "\-\-tag=OBJC" Makefile.in 216:LTOBJCCOMPILE = $(LIBTOOL) --tag=OBJC --mode=compile $(OBJC) $ (DEFS) \ 220:OBJCLINK = $(LIBTOOL) --tag=OBJC --mode=link $(OBJCLD) $ (AM_OBJCFLAGS) the change elicits a warning that OBJC is not a known tag, but it keeps going and compiles fine. I do not use the xgrid portion so I do not now whether this is clobbered or not. Standard runs using orterun work fine. Peter Brian Barrett wrote in July: On Jul 14, 2006, at 10:35 AM, Warner Yuen wrote: > I'm having trouble compiling Open MPI with Mac OS X v10.4.6 with > the Intel C compiler. Here are some details: > > 1) I upgraded to the latest versions of Xcode including GCC 4.0.1 > build 5341. > 2) I installed the latest Intel update (9.1.027) as well. > 3) Open MPI compiles fine with using GCC and IFORT. > 4) Open MPI fails with ICC and IFORT > 5) MPICH-2.1.0.3 compiles fine with ICC and IFORT (I just had to > find out if my compiler worked...sorry!) > 6) My Open MPI confguration was using: ./configure --with-rsh=/usr/ > bin/ssh --prefix=/usr/local/ompi11icc > 7) Should I have included my config.log? It looks like there are some problems with GNU libtool's support for the Intel compiler on OS X. I can't tell if it's a problem with the Intel compiler or libtool. A quick fix is to build Open MPI with static libraries rather than shared libraries. You can do this by adding: --disable-shared --enable-static to the configure line for Open MPI (if you're building in the same directory where you've already run configure, you want to run make clean before building again). I unfortunately don't have access to a Intel Mac machines with the Intel compilers installed, so I can't verify this issue. I believe one of the other developers does have such a configuration, so I'll ask him when he's available (might be a week or two -- I believe he's on vacation). This issue seems to be unique to your exact configuration -- it doesn't happen with GCC on the Intel Mac nor on Linux with the Intel compilers. Brian -- Brian Barrett Open MPI developer http://www.open-mpi.org/
[OMPI users] mpi.h - not conforming to C90 spec
Hi, Compiling an mpi program with gcc options -pedantic -Wall gives the following warning: mpi.h:147: warning: ISO C90 does not support 'long long' So it seems that the openmpi implementation doesn't conform to C90. Is this by design, or should it be reported as a bug? Thanks, Jonathan
Re: [OMPI users] mpi.h - not conforming to C90 spec
On Aug 17, 2006, at 4:43 PM, Jonathan Underwood wrote: Compiling an mpi program with gcc options -pedantic -Wall gives the following warning: mpi.h:147: warning: ISO C90 does not support 'long long' So it seems that the openmpi implementation doesn't conform to C90. Is this by design, or should it be reported as a bug? Well, MPI_LONG_LONG is a type we're supposed to support, and that means having 'long long' in the mpi.h file. I'm not really sure how to get around this, especially since there are a bunch of users out there that rely on MPI_LONG_LONG to send 64 bit integers around on 32 bit platforms. So I suppose that it's by design. Brian
Re: [OMPI users] mpi.h - not conforming to C90 spec
On 18/08/06, Brian Barrett wrote: On Aug 17, 2006, at 4:43 PM, Jonathan Underwood wrote: > Compiling an mpi program with gcc options -pedantic -Wall gives the > following warning: > > mpi.h:147: warning: ISO C90 does not support 'long long' > > So it seems that the openmpi implementation doesn't conform to C90. Is > this by design, or should it be reported as a bug? Well, MPI_LONG_LONG is a type we're supposed to support, and that means having 'long long' in the mpi.h file. I'm not really sure how to get around this, especially since there are a bunch of users out there that rely on MPI_LONG_LONG to send 64 bit integers around on 32 bit platforms. So I suppose that it's by design. OK, that seems reasonable. I wonder then if the non-C90 conforming parts should be surrounded with #ifndef __STRICT_ANSI__ - this is predefined when gcc is expecting C90 conforming code. I am not sure if this is portable to other compilers however. Probably not. Best wishes, Jonathan