An embedded and charset-unspecified text was scrubbed... Name: not available URL: <https://stat.ethz.ch/pipermail/r-sig-hpc/attachments/20100304/4cc13f16/attachment.pl>
care to send me some big working example R on multiprocessor systems
5 messages · Debabrata Midya, Melwin, Hao Yu
Hi Deb, PPSO in parallel mode should also run on a single processor, although you won't gain anything from that beyond just testing (in fact, I will be slower due to the overhead involved, unless your objective function takes long to evaluate without computational burden, e.g. due to extensive file i/o). Which Windows MPI-variant do you use (MPICHII or DeinoMPI)? I tested your code snippet successfully on Windows 2000/DeinoMPI/R-2.10.0. In any case, I recommend using optim_ppso_robust, which is more elaborated by now. Please let me know if you experience further problems. cheers, Melwin Am 04.03.2010, 00:50 Uhr, schrieb Debabrata Midya <Debabrata.Midya at services.nsw.gov.au>:
Melwin,
Thanks in advance.
I am using Windows XP and trying to use ppso especially the parallel
mode using Rmpi.
I have installed R-2.10.1, ppso_0.9-921.zip and Rmpi_0.5-9.zip.
I need your assistance to solve the problem below:
1. Currently I have only 1 processor. Is it possible to use optim_ppso
or optim_ppso_robust on 1 processor?
2. When I run:
result = optim_ppso(objective_function=rastrigin_function,
nslaves=1, max_number_function_calls=200, projectfile=NULL,
logfile=NULL) #standard version
The system hangs.
Once again thank you very much for the time you have given.
Regards,
Deb
It works for me as well. Deb: Can you check if Rmpi runs on its own (spawning slaves) before ppso? If Rmpi is not setting properly on Windows, it will hang. Melwin: Do you have more time consuming codes to share? Intel asks me to provide some R codes to test their MPI against others under R. But my codes are mostly related to simulation and overhead with MPI is minimum. Thanks. Hao
Melwin wrote:
Hi Deb, PPSO in parallel mode should also run on a single processor, although you won't gain anything from that beyond just testing (in fact, I will be slower due to the overhead involved, unless your objective function takes long to evaluate without computational burden, e.g. due to extensive file i/o). Which Windows MPI-variant do you use (MPICHII or DeinoMPI)? I tested your code snippet successfully on Windows 2000/DeinoMPI/R-2.10.0. In any case, I recommend using optim_ppso_robust, which is more elaborated by now. Please let me know if you experience further problems. cheers, Melwin Am 04.03.2010, 00:50 Uhr, schrieb Debabrata Midya <Debabrata.Midya at services.nsw.gov.au>:
Melwin,
Thanks in advance.
I am using Windows XP and trying to use ppso especially the parallel
mode using Rmpi.
I have installed R-2.10.1, ppso_0.9-921.zip and Rmpi_0.5-9.zip.
I need your assistance to solve the problem below:
1. Currently I have only 1 processor. Is it possible to use optim_ppso
or optim_ppso_robust on 1 processor?
2. When I run:
result = optim_ppso(objective_function=rastrigin_function,
nslaves=1, max_number_function_calls=200, projectfile=NULL,
logfile=NULL) #standard version
The system hangs.
Once again thank you very much for the time you have given.
Regards,
Deb
_______________________________________________ R-sig-hpc mailing list R-sig-hpc at r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-hpc
Department of Statistics & Actuarial Sciences Fax Phone#:(519)-661-3813 The University of Western Ontario Office Phone#:(519)-661-3622 London, Ontario N6A 5B7 http://www.stats.uwo.ca/faculty/yu
An embedded and charset-unspecified text was scrubbed... Name: not available URL: <https://stat.ethz.ch/pipermail/r-sig-hpc/attachments/20100305/46ab1033/attachment.pl>
Hi Deb, I was NOT trying to suggest that you mandatorily need DeinoMPI. However, afaik either MPICHII *OR* DeinoMPI is required for Rmpi under Windows. Please note that there are two versions of the Rmpi package, depending on if you use MPICHII or DeinoMPI. Although the aim is parallel computing, I doubt that it is a good idea to run these two in parallel ;-) Anyway, you should follow Hao's advice and first make sure Rmpi is working properly on its own. Please try some of the simple Rmpi examples to see if they work and let me know the result. Melwin PS: As for the configuration of DeinoMPI, I have to admit that I never got a true multi-machine Rmpi session working. Although I could include remote machines in DeinoMPI, in R Rmpi-slaves were exclusively spawned locally. I didn't try MPICHII at all, but resorted to Ubuntu-Linux, where everything went much smoother. Am 05.03.2010, 01:22 Uhr, schrieb Debabrata Midya <Debabrata.Midya at services.nsw.gov.au>:
Melwin, Thanks for your reply. I need you assistance to run ppso using DeinoMPI as well. As per your advice, I will start using optim_ppso_robust. 1. Which Windows MPI-variant do you use (MPICHII or DeinoMPI)? I have installed Windows MPICH2 and provided the path \MPICH2\bin\ in the path variable and it works well on R-2.10.1. 2. DeinoMPI I am unable to run the same code as in (1) using DeinoMPI. I have installed DeinoMPI.2.0.1.msi and I have tried to set the followings. I feel I have not done correctly. Do you have any advice on it? The systems hangs. a. Click "enable create store options" b. Click "Create Credential Store" c. In "Add Credential", add your account and password Once again thank you very much for the time you have given. Regards, Deb
Melwin <win at comets.de> 4/03/2010 7:35 pm >>>
Hi Deb, PPSO in parallel mode should also run on a single processor, although you won't gain anything from that beyond just testing (in fact, I will be slower due to the overhead involved, unless your objective function takes long to evaluate without computational burden, e.g. due to extensive file i/o). Which Windows MPI-variant do you use (MPICHII or DeinoMPI)? I tested your code snippet successfully on Windows 2000/DeinoMPI/R-2.10.0. In any case, I recommend using optim_ppso_robust, which is more elaborated by now. Please let me know if you experience further problems. cheers, Melwin Am 04.03.2010, 00:50 Uhr, schrieb Debabrata Midya <Debabrata.Midya at services.nsw.gov.au>:
--