The ABINIT Users Mailing List ( CLOSED )

[Masayoshi Mikami <mmikami@rc.m-kagaku.co.jp>]

Dear all,

After running tests_fast,v1-v5 with the binary compiled
with -O3 on Itanium2/ifort(8.1), I had runtime errors with
v1/t55, t56, t57, t58, t71, t72, t73 ; v4/t46, v5/t09. t10.
Numerically, v1/t97, v2/t26, v5/t21 were strange...

Well, I will have to spend some time to play with
~/.abinit/build/*.ac by tuning "fcflags_opt_*" line, such as
fcflags_opt_21drive="-O0"
(This is it, Yann ?)

BTW, from Gilles,
> Thanks for the lesson you taught us....
> One more, if you have some time: do you know exactly the routine 
(s) whose time increase (nonlop is made of many subroutines like  
opernl etc..) when you change the optimization option?

Well, I would like to find them when I have time.
(it will take more time ... first, I would like to eliminate
 runtime errors/numerical errors. Then I will see the optimization  
effect
 as you suggested, Gilles.)

Everyone here, if you have similar experience as described
in these e-mails and already find some clues, could you kindly
share your experience with us ?
("I'll be back," <- pronounce like "Terminator" ;-)
 when I have something to be shared about this issue)

Regards,
Masayoshi

On 2007/05/16, at 14:14, Masayoshi Mikami wrote:

> Dear all,
>
> The story is still going on ...
>
> On 2007/05/14, at 15:44, Masayoshi Mikami wrote:
> > My preoccupation came from the experience with v.4.6.x,
> > where I used attached makefile_macros with "-O2" ...
>         (snip)
> > Thus, I would like to suggest "-O3 -tpp2" as default in configuration
> > for Itanium2/ifort... Yann, how do you think ?
>
> Then, I remember I had to check by running "make tests_dev".
> When I run the tests_dev with v.5.3.4, I am having some runtime  
> errors,
> e.g. v1/t55 ...  on display, I found the following error message:
>
> File abfiles.run created, starting abinit
> forrtl: severe (174): SIGSEGV, segmentation fault occurred
> Image              PC                Routine            Line         
> Source
> abinis             4000000000A8F241  Unknown               Unknown   
> Unknown
> abinis             40000000005BE9B0  Unknown               Unknown   
> Unknown
>
> In passing, from t55_STATUS
> Status file, with repetition rate  49, status number   638
>   Level abinit         : call driver
>   Level driver         : call gstate
>   Level gstate         : call scfcv
>   Level scfcv          : call vtorho
>   istep   =    8
>   Level vtorho         : call vtowfk
>   isppol  =    1
>   ikpt    =    1
>   Level vtowfk         : before loop
>   Level cgwf           : loop iband
>   iline   =  901
>
> I did not have such errors with the binary compiled with "-O2".
> Thus, I will need to reduce optimization level as -O2 for
> some subroutines in ~ABINIT/src. Thus let me ask something :
>
> In v.4.x.x, we had the following trick in makefile_macros:
> FFLAGS=-FR -O3 -w -tpp2  # default
> FFLAGS_Src_2psp =-FR -O2 -w -tpp2       # -O3 => -O2
> FFLAGS_Src_6response =-FR -O2 -w -tpp2   # -O3 => -O2
> FFLAGS_LIBS= -O3 -w  -tpp2
> to have different optimization levels for the different source  
> directories.
>
> In v.5.3.x, how to set the FCFLAGS for the directories individually
> when I run "./configure" ?
>
> We might want to tune the optimization levels
> for each of the directories in "src": namely,
> 00basis,  01manage_mpi, 11contract, 11util, 12ffts,
> 12geometry, 12nlstrain, 12parser,12poisson, 12spacepar
> 13io_mpi, 13ionetcdf, 13iovars, 13nonlocal, 13paw, 13psp
> 13recipspace, 13xc, 13xml, 14bader, 14iowfdenpot,
> 14occeig,14wfs, 14wvl_wfs, 15common, 15gw, 15rsprc
> 16geomoptim.16response, 17ddb, 17lwf, 17suscep.
> 18seqpar, 19cut3d, 21drive, 21paral_md, defs,
> lib01cg, lib01fftnew. ib01hidempi, main
>
> Regards,
> Masayoshi