Re: large-scale polynomial gcd calculations



Raymond Toy writes:
> Vadim V. Zhytnikov wrote:
> 
>>Andrey G. Grozin writes:
>>
>>
>>>3. The platforms which will be used are: Pentium-4, AMD Opteron,
>>>Itanium2 (32-processor SGI Altix server), SUSE Linux everywhere. What
>>>lisps can run on all of them? What about 64-bit addressing - possible?
>>>desirable?
>>>
>>
>>Recently I have some experience with quite big Maxima
>>problem on i86 32-bit system with 1Gb RAM.
>>I've tried GCL, SBCL, CLISP and CMUCL.
>>None of tis Lisps can allocate more than 1Gb of RAM
> 
> 
> CMUCL can allocate up to about 1.6 GB using the -dynamic-space-size switch.
> 
> 
>>on 32-bit system.  Usually even less, just 500-800Mb.
>>I'm only not sure about CMUCL - maybe it can allocate
>>more than 1Gb.  In any case you probably
>>need 64-bit system with lots of RAM.  I've heard
>>people use GCL on 64-platform (IA64 ?) with really
>>huge heap size.  You have to build GCL with
> 
> 
> SBCL has a port to x86-64 (AMD).  I'm pretty sure it works.
> 

Right, I've tried sbcl 0.9.4 x86_64.  It passes various tests
and in general look quite stable.  Clisp also works on x86_64.
But I never tried all these 64-bit lisps on large problems.
To do so one need more than 1Gb of RAM.  To feel real advantage
over 32-bit system 4Gb or better 8Gb is required.

-- 
      Vadim V. Zhytnikov