AMBER Archive (2002)

Subject: Re: Intel ifc Compiler

From: Joachim Reichelt (Reichelt_at_gbf.de)
Date: Mon Jun 03 2002 - 01:51:54 CDT

Dear David Case,

I think this is the same I think of it now!
Even the parts that seems to compile did not work at all.
AMBER 6 is FORTRAN 66 in many places, and that is what the ifc
does not like!
e.g. "bad" array sizes in subroutines as in:
      interger a(1)
will crash at least at runtime.
It does not make sence to me to put work in it any more, as there
is a running g77 option.

AND:

It is better to wait for the arrival of AMBER 7 here.
I ordered it befor starting playing around with AMBER 6 and ifc.

Thanks for your help.

David Case wrote:

>On Fri, May 31, 2002, Joachim Reichelt wrote:
>
>
>
>>*Compiler Internal Error* : Please report to INTEL
>>compilation aborted for _gibb_.f (code 1)
>>make: *** [gibb.o] Error 1
>>
>>shall I send it to Intel???
>>
>>
>
>It's up to you. I'll repeat what I said before. Amber 6 was written before
>there was an Intel compiler, and it looks like it is not compatible. Amber
>7 (the current release) works fine with ifc. So my best suggestion is to
>either get the current release of Amber, or to use one of the compilers
>that is supported with Amber 6.
>
>You could try splitting gibb.f into smaller pieces -- I think that will make
>ifc happy. But you would be putting in a lot of work for very little gain:
>g77-compiled code is not much slower than ifc-compiled code. If you are
>truly concerned about performance, you should upgrade to the new version of
>Amber anyway, since it is faster for most jobs than is Amber 6.
>
>(All of this assumes you are planning to use gibbs in the first place; if
>you aren't, the above problems may be irrelevant to you.)
>
>....hope this helps....dac
>
>
> 

-- 
Mit freundlichen Gruessen                         Best Regards
 
         Joachim Reichelt