Re: [ccp4bb]: Comments on G4 Ti

[Ethan Merritt <merritt@u.washington.edu>]

***  For details on how to be removed from this list visit the  ***
***          CCP4 home page http://www.ccp4.ac.uk         ***

On Friday 03 January 2003 11:31, wgscott@chemistry.ucsc.edu wrote:
> Does anyone know how much better the Absoft and other proprietary
> compilers do?  We've had lots of trouble getting vectorization and
> altivec to work with the Absoft compiler and gave up.

We've made only preliminary tests with the Intel compiler, and only
on Athlon.  Bottom line so far is that the Intel and Portland Group (PGI)
compilers give essentially identical run times, but we have at least one
piece of code [raster3d :-( ] where the Intel compiler produces incorrect
code.  So for now we're sticking with PGI.

The big surprise, however, was that the gnu compilers sped
up dramatically with release 3.2.  g77 used to produce code 50% to
100% slower than pgf77, but now it's only a few percent slower.
It still has other problems, though, like not supporting file-handling
modes common to programs developed on other platforms.

-- 
Ethan A Merritt       merritt@u.washington.edu
Biomolecular Structure Center Box 357742
University of Washington, Seattle, WA 98195
phone: (206)543-1421
FAX:   (206)685-7002