[Beowulf] OS for 64 bit AMD

Joe Landman landman at scalableinformatics.com
Mon Apr 4 22:45:39 PDT 2005



Mike Davis wrote:
> Joe,
> Thanks. I am running PG, ABsoft, intel, and gnu compilers on my variety 
> of clusters and smp machines. The move to an f95 makes sense overall. 
> But, you might not believe just how much f77 and earlier code (ported to 
> f77) is still in use. It's not unusual to have someone come by the 

Hi Mike:

   Actually I would believe it. :)

> office once or twice a semester with some code that is f77 or Fortran 4 
> and want to know if it can run on one of the clusters. Then, of course, 
> we have the whole parallelization discussion. "No it won't just work as 
> a parallel program. Yes we might be able to get some speedup with 
> automated tools and loop unrolling. No the speedup won't be linear, 1+1 
> will not equal 2 for this case. The best speedup would come from a 
> rewrite for mpi, etc. etc. "

The funny thing is that for well designed and well written code, the 
parallelization efforts usually are not terrible (though they shouldn't 
expect multiple orders of magnitude  speedups without a ground up 
rewrite in most cases).

I remember when Craig Burley proposed g77, and I remember using f2c 
prior to that on some platforms without fortran compilers.  Toon 
indicates that distro vendors are free to install gcc3 next to gcc4, so 
you won't "lose" g77 per se ...

Joe

> 
> thanks again,
> 
> Mike
> 
> Joe Landman wrote:
> 
>>
>>
>> Mike Davis wrote:
>>
>>> What are the good reasons not to have g77 in gcc4? I admit ignorance 
>>> on the subject of gcc4, but g77 is useful to many in the scientific 
>>> and academic communities.
>>
>>
>> Hi Mike:
>>
>>   g77 does not work with the tree-ssa branch that gcc4 is emerging/has 
>> emerged from (please correct me if I am wrong about the details on 
>> this Greg/Toon/others).  gfortran is going to take its place (again 
>> please correct me if this statement is not correct).  c.f. 
>> http://gcc.gnu.org/develop.html and related and the mailing list 
>> archives at http://gcc.gnu.org/ml/fortran/ and 
>> http://gcc.gnu.org/fortran/ .  The idea is that gfortran is a f95 
>> compiler to supplant g77 (an f77 + some f90 extensions) due in large 
>> part to the issues in carrying forward a pure f77 compiler when many 
>> in the fortran world are at/beyond f95.  That is, fortran will be 
>> available, but the compiler appears to be an f95 compiler going forward.
>>
>>   That said, there are excellent commercial compilers available from 
>> Greg's company (PathScale), the Portland Group (been playing with 
>> v6.0), Intel for their processors, and others on the x86/AMD64/IA64 
>> stream. For those of the PPC persuasion, the IBM compilers are quite 
>> good.
>>
>> Joe
>>
> 

-- 
Joseph Landman, Ph.D
Founder and CEO
Scalable Informatics LLC,
email: landman at scalableinformatics.com
web  : http://www.scalableinformatics.com
phone: +1 734 786 8423
fax  : +1 734 786 8452
cell : +1 734 612 4615



More information about the Beowulf mailing list