Re: Re: Illusory Multicore Support in 5.2?
- To: mathgroup at smc.vnet.net
- Subject: [mg66139] Re: [mg66101] Re: Illusory Multicore Support in 5.2?
- From: George Woodrow III <georgevw3 at mac.com>
- Date: Sun, 30 Apr 2006 04:22:09 -0400 (EDT)
- References: <200604260838.EAA02742@smc.vnet.net> <email@example.com> <200604270836.EAA01610@smc.vnet.net> <firstname.lastname@example.org> <200604290741.DAA23489@smc.vnet.net>
- Sender: owner-wri-mathgroup at wolfram.com
With some help from Pratik Desai, I was able to get multi-core to work on my Quad Core G5. On OS X, the functionality 'just works'. Your question should be why Windows is not set up with the right env variable out of the box. There may be unintended consequences that force the default values to be where they are. One might be suspicious of a program that silently changed an OS parameter this way. Beyond that, Wolfram could be a bit clearer about which subset of problems take advantage of multi-core computers. The 'What's New' documents all focus on certain linear algebra functions, so I suspect that it is the underlying library for these functions that use multiple processors. If you don't read these, the more general statements could be considered to be mild marketing hype. As soon as I can get the funding, I intend to upgrade to the Personal Grid version. It is clear that (for Macs, at least) multiple cores are the future. All of Apple's new portables are dual-core, and I think that this trend will continue. One open question is how long the personal grid version of Mathematica will exist as a separate product when the bulk of users will have at least dual core systems. This topic is probably beyond the scope of this mail-list. george On Apr 29, 2006, at 3:41 AM, Antti Penttilä wrote: > OK, the answer below from Pratik Desai did the trick - now SVD is > faster > and the processor load reaches 100% from time to time. > > I just can't understand why my Mathematica wasn't configured like this > by default. Both Intel and AMD are very much into multicore computers > and also Mathematica advertise its multicore capabilities. Soon the > major part of new computers sold will be multicore. Where is this > 'trick' of new environment variable documented? One could expect > that a > professional software like Mathematica would either offer some kind of > choice when installing or then would have some option inside the > program > where this could be changed. > > Antti > > > Pratik Desai <pratikd at wolfram.com> wrote: >> Antti Penttilä wrote: >>> Pratik Desai wrote: >>> >>>> Try using SingularValueDecomposition on a sufficiently large matrix >>>> >>>> Pratik >>>> Wolfram Research, Inc. >>>> Tech Support >>>> >>>> >>> I have a new dual-core Athlon 64 X2 and WinXP 64-bit edition with >>> Mathematica 5.2. SingularValueDecomposition with a random 2000x2000 >>> matrix takes about 41 seconds to complete but the processor load >>> stays >>> around 50% all the time. It seems that we'll have to start >>> getting some >>> refund of the price of Mathematica. >>> >>> Best regards, >>> Antti Penttilä >>> >> Hello Antti >> >> Mathematica 5.2 defaults to using only one thread for linear >> algebra on >> Windows. To use multiple threads and take advantage of multiple >> processors or cores, you will need to set OMP_NUM_THREADS to a >> value of >> % NUMBER_OF_PROCESSORS% in your environment. You can do this by >> going to >> My Computer and get to the properties template by right clicking and >> clicking Properties, click the Advanced Tab and then click on >> Environment Variables. Click on New right below the System Variables >> list. In the the variable name type in OMP_NUM_THREADS and variable >> value type in % NUMBER_OF_PROCESSORS% >> >> ---------------------------------------- >> Variable Name : OMP_NUM_THREADS >> Variable Value: % NUMBER_OF_PROCESSORS% >> ---------------------------------------- >> >> Now you can try your linear algebra operations. Dot on two >> matrices is >> the best indicator of your multiprocessing capabilities. If you have >> have further questions please let me know, or you can e-mail us >> directly >> at support at wolfram.com. I apologize for any confusion. >> >> >> Best regards, >> >> Pratik >> >