[Wien] Re: [WIEN]: bug in parallel script

[Torsten Andersen]

Dear Stefaan,

for some of us this is the desired behaviour, even if we use a mix of 
Intel Pentium 4 and Athlons, since the optimizations are different for 
the two architectures. Therefore the parallel lapw1 should use different 
binaries on P4, Athlon, Sun US2, Sun US3, Alpha ev6, Alpha ev67, etc. 
This should of course be selected in the .cshrc or whatever login 
initialization there is.

Depending on the *nix you are using (Linux, HP/UX, SunOS, Tru64, AIX, 
etc.) there are different environment variables indicating the 
architecture. Of course, as a last resort you can base the selection on 
the hostname command.

Best regards,
Torsten.

Stefaan Cottenier wrote:

>>====
>>Peter Blaha <pblaha at theochem.tuwien.ac.at>
>>submitted the following contribution:
>>====
>>
>>It's not a bug, it's a feature, although one can discuss what is more
>>common.
>>
>>Suppose you want to combine a P-IV with an Athlon (or even with a
>>DEC-Alpha!!!), than you MUST have different WIENROOTS for the two
>>computersi, since of course you want to use the P-IV executable on one
>>machine, but the DEC-alpha executable on the other. The present version
>>works for this case.
>>    
>>
>
>Here is something I cannot follow. If an Athlon and P-IV are in the same
>NFS, then also the .cshrc they use is the same, and so is the WIENROOT !? Or
>do you mean that in such cases the .cshrc has to assign the value of
>WIENROOT depending on the detected machine name ?
>
>  
>
>>Of course, your case is different. You have 2 WIEN installations, one is
>>in $WIENROOT, but you want to use the other one. Most other WIEN
>>scripts use the $bin/executables and not necessarily the one of
>>$WIENROOT", so you simply call the first WIEN-command with the full path
>>and from then on use those executables.
>>
>>However, we think having 2 WIEN versions on one computer is not for the
>>"regular" user, and finaqlly we decided to use the present version.
>>    
>>
>
>OK, I see the problem and understand your choice. However, it leads to an
>undesirable dissimilar behaviour between run_lapw and run_lapw -p : if a
>full path (different from $WIENROOT) is specified, then run_lapw will use
>for every subprogram the version of that path, while run_lapw -p will use
>the version in $WIENROOT. And therefore run_lapw -p is not just "run_lapw
>executed in parallel"...
>
>Stefaan
>
>_______________________________________________
>Wien mailing list
>Wien at zeus.theochem.tuwien.ac.at
>http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
>
>
>  
>

-- 
Dr. Torsten Andersen
Department of Physics, Condensed Matter Theory Group, Uppsala University
UU-WWW: http://www.fysik4.fysik.uu.se/    TA-WWW: http://deep.at/myspace
New: Summer School 2003: http://www.fysik4.fysik.uu.se/~thor/school.html