MathGroup Archive 2006

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

Search the Archive

Re: Re: why does not the Mathematica kernel seem to 'multi-task' between computations in different windows?

  • To: mathgroup at smc.vnet.net
  • Subject: [mg69892] Re: [mg69866] Re: why does not the Mathematica kernel seem to 'multi-task' between computations in different windows?
  • From: Benedetto Bongiorno <bbongiorno at attglobal.net>
  • Date: Wed, 27 Sep 2006 06:03:43 -0400 (EDT)

Mathgroup,

I have been talking to Wolfram on the same issue.
I am in the process of purchasing a computer configuration that will 
allow us to run
four notebooks simultaneously on a Windows XP 64 bit OS, using two 
Opteron dual cores.
Wolfram has advised that one license of 5.2 will allow you to run two 
notebooks simultaneously,
using one dual core processor. If we buy another license we can run two 
additional notebooks simultaneously,
using second dual core processor.
Wolfram has made it clear all variables for each notebook's kernel  is 
confined to a core of the dual processor.
Wolfram's PersonalGrid appears to do the same thing with more functionality.

Does anyone else have any experience in this area?

Thank you,
Ben

ben wrote:
> Hi,
>
> I suggest to check the manual about the difference between
> mathematica frontent (windows) and mathematica kernel (computation
> client):
>
> The frontend only displays your worksheet nicely, handles editing and
> sends
> commands to the kernel; its the kernel that does all the computation
> work.
>
> Now the point is: You can open several frontends which
> by default will all be linked to the same instance of the kernel.
>
> This has two surprising effects,
>
> 1. A kernel which is busy with a command from frontend no.1 is simply
> busy, it can not do computations for frontend no.2 in the same time
>
> 2. All variables are stored by the kernel, that is a variable defined
> in frontend no.1
> is visible also in frontend no.2
>
> Effect 2 makes clear why you cannot have multi-tasking: results would
> be not defined uniquely, since results from frontend no.2 can effect
> the state of the variables and thus computations in frontend no.1 and
> vice versa, and if you do not know which of two calculations running
> parallel finishes first you might end up in a whole mess.
>
> What's now the solution to your problem?
>
> Kernel->Kernel Configure Options allows you to add additional kernels
> (make them local for simplicity)
> Then by
> Kernel->Notebook's Kernel
> you can assign a unique kernel to each notebook.
> This will use up extra memory and maybe licences too.
>
> Hope that helps
> Ben
>
> Nasser Abbasi schrieb:
>
>   
>> Hi
>>
>> Sometime I start some computation in one window which can take few minutes
>> to half an hour or more.
>>
>> It seems anything I type meanwhile in a second window does not get processed
>> until the first computation is completed.
>>
>> The computation does have 'Print[]' statements in it for debugging, so it is
>> not completely CPU bound, and I would think that one should be able to do
>> someone work in the second window meanwhile even if it might be slower to
>> respond.
>>
>> It seems that the Mathematica kernel can't be preempted at all once it
>> starts a computation, even if I/O calls are being made.
>>
>> Right now, once I start something, I can't use Mathematica at all until the
>> computation is over because the kernel is busy. (Except it seems I am able
>> to do File->Save ok while it is busy)
>>
>> Sometimes I start a second copy of Mathematica, but the limit is 2 instances
>> and I do not have lots of RAM.  I also wonder if this limit should be
>> increased to say 5.
>>
>> I guess it is time for me to go buy a PC with those Intel new chips the
>> core2 with lots of RAM as I spend too much time now waiting for something to
>> complete.
>>
>> Nasser
>>     
>
>
>
>   


  • Prev by Date: Re: Program to calculate rational function with imbedded continued fraction
  • Next by Date: Re: linear second order homogeneous differential equation recursions
  • Previous by thread: Re: why does not the Mathematica kernel seem to 'multi-task' between computations in different windows?
  • Next by thread: Re: Re: Re: why does not the Mathematica kernel seem to 'multi-task' between computations in different windows?