MathGroup Archive 2007

[Date Index] [Thread Index] [Author Index]

Search the Archive

Re: Mathematica, Quad Core, Linux

  • To: mathgroup at smc.vnet.net
  • Subject: [mg73251] Re: Mathematica, Quad Core, Linux
  • From: Antti Penttilä@smc.vnet.net
  • Date: Fri, 9 Feb 2007 02:12:44 -0500 (EST)
  • Organization: University of Helsinki
  • References: <eqc7tb$snu$1@smc.vnet.net> <eqeodl$530$1@smc.vnet.net>

Hi,

I don't have a Quad-core system running - only dual core Windows. But I 
think (based on my experience with 2-core) that the answer to your 2. 
question is: _If_ your calculation can be parallelized and distributed 
to 4 cores it will not redistribute itself in the middle of the 
calculation. It will use all the CPU-time it will get from the 4 cores. 
Moreover, with Mathematica you have to set the number of the available 
cores with an environment variable so its inconvenient to change this 
all the time (and might need restarting the Mathematica kernel?).

Regards,
Antti

>> 2. I got the following negative general comment on running multi-processor for parallel computing. "The parallelization is done at the beginning of the run. So,
>> if you tell the program (or the program finds) there are 4 cores available at the start of the program, it grabs them and assumes they will be available exclusively to the program. If a second user logs in and tries to run something, processing speed will be terrible." Is this also the case with the Mathematica Quad Core combination?



  • Prev by Date: Re: Would someone confirm they also see this problem?
  • Next by Date: Re: How to parse result from Reduce[ ] function
  • Previous by thread: Re: Re: Mathematica, Quad Core, Linux
  • Next by thread: Trouble accessing "Windows"-type Mathematica