Re: [Freeon-users] Ktrax-preprocessed.f with --enable-parallel-clones

[Jose R. Valverde]

Thanks for all the details, guys.

        I am too busy with other topics, but keep on trying. Now, for
MPI if it is going to be a total rewrite, I can leave that out, and 
forget about it for the time being, leaving the clone code only.

        Now, regarding the clone code, I did run in a sort of glitch
as the validation test "fails" on 12-BPA due to precision differences.
I am compiling on an x86_64 and I guess the check is for a 32 bit 
architecture or that I may be using something that makes the precision
slightly different for the numbers are the virtually same with errors
in the range of 10e-7 to 10e-8.

        But the code runs OK manually, so I'll likely modify my copy
of the validation makefile just to be able to complete all the tests
and be sure so I can let you  know.

        BTW, this is a 48 CPU, 128 GB RAM, Linux  2.6.38-11-generic
#50-Ubuntu SMP Mon Sep 12 21:17:25 UTC 2011 x86_64 x86_64 x86_64 
GNU/Linux machine, with GCC version 4.5.2 (Ubuntu/Linaro 4.5.2-8ubuntu4)
and Gfortran  version 4.5.2 (Ubuntu/Linaro 4.5.2-8ubuntu4).

                                j

On Wed, 19 Oct 2011 20:25:55 -0500
Jeff Hammond <address@hidden> wrote:
> It is not surprising at all that those flags cause the compiler to
> explode.  Compiling KTrax_7_6 with "-msse -msse2 -mfpmath=sse
> -ffast-math -ftree-vectorize" is cruel and unusual punishment.  That
> subroutine has a single loop that is more than 10000 lines long and is
> completely vectorizable.
> 
I did a few checks, and what makes it fail is the option -ftree-vectorize,
you can compile Ktrax with

gfortran -g -O2 -march=native -msse -msse2 -mfpmath=sse -ffast-math
-pipe -I. -I../Modules -I. -I.. -I../OneE
-I../Modules/MMA/LookUpTables_800_6x   -c -o KTrax.o
KTrax-preprocessed.f 

with no problem at all, so at least the SSE optimizations are taken into
account.

-- 
                        EMBnet/CNB
                Scientific Computing Service
        Solving all your computer needs for Scientific
                        Research.

                http://bioportal.cnb.csic.es
                  http://www.es.embnet.org