AW: [Wien] Compiling Wien2k for beginners

[Gerhard Fecher]

Dear Peter,
I updated already a little, some errors are indeed not Wien2k specific, I hope this is now more clear.

I still have SIGSEGV errors for large cases if using full static linking.
This is independent on the -ulimit settings.
All Fortran (Intel) libraries can be linked statically, but if the GNU libraries are linked with -static, then the error occurs. However, I did not succeed to use the debugger, it seems that the SIGSEGV ist just a error message being not the "real" cause of the crash.
With KKR I realized that this type of error occurs if the dimensions of an array becomes larger then the compiler limit, that is still (2^31)-1 if you don't use an Itanium. (64bit extension is not 64bit).

I may need to comment a little more on the needed linker or compiler switches. Youre experience with -lsvml seems to be similar to my observation on -lpthread, but I did not figure out how ld was finding it. The need for -lsvml may also depend on the optimisation level.

The timings had some typos, that are corrected, you are right, lapw1 makes mainly use of mkl thats also the reason why the optimization doesn't bring much speed. I will check what happens if I use a regular blas/lapack instead (I did already for other programs but never with Wien). 

Ciao Gerhard

-----Ursprüngliche Nachricht-----
Von: wien-bounces at zeus.theochem.tuwien.ac.at im Auftrag von Peter Blaha
Gesendet: Sa 31.12.2005 18:18
An: A Mailing list for WIEN2k users
Betreff: Re: [Wien] Compiling Wien2k for beginners
 
Thank's for this effort. I'll for sure make a link from the WIEN2k-faq 
page to your site.

In particular the compiler/library instrallation should be very helpfull 
for the users.

The hints for segmentation faults are of quite general nature and apply 
in particular in a "program development" phase. It gives 
a little bit the impression that these problems are in the fortran source 
code of WIEN2k (which I hope they are not), while I think they are in the 
fortran library routines.
Therefore the hint to set all system variables to "unlimited" is vital in 
my experience, otherwise larger "cases" will not run in WIEN. 

The specific compiler options will still differ from system to system, in 
particular 32 or 64 bit operating systems of various Linux flavours may / 
may not work with your recommended options/libraries.

Just as example: on my 64 bit RedHat Xeon system I can use -Bstatic for 
-lguide, but NOT for -lguide_stats (and the latter library is necessary 
ONLY for static lguide).
I also do NOT need   -lsvml (when the compiler is installed properly, the 
compiler libraries (all libs in "fc90") should be found automatically, 
while the mkl-libs need to be given explicitly by name AND -L/...

I should also mention, that on my systems the timing differences between 
static and dynamic linking is (more or less) not visible.

Anyway, many thanks and a happy new year!


> NOTFor those who still have troubles to compile Wien2k
> 
> visit http://www.ghfecher.de/html/bandstructure.html
> 
> It is not finished but hopefully helps to avoid questions.
> 
> If someone is interested to help me with the introduction, let me know, I will send the tex file. Proof reading by a native English speaker would also be welcome.
> 
> Would be nice if someone can prepare a part about dual-core processors and/or other multi-processor set-up.
> 
> I will try to add something about working on an OPTERON Cluster, but may take some time.
> 
> 


                                      P.Blaha
--------------------------------------------------------------------------
Peter BLAHA, Inst.f. Materials Chemistry, TU Vienna, A-1060 Vienna
Phone: +43-1-58801-15671             FAX: +43-1-58801-15698
Email: blaha at theochem.tuwien.ac.at    WWW: http://info.tuwien.ac.at/theochem/
--------------------------------------------------------------------------
_______________________________________________
Wien mailing list
Wien at zeus.theochem.tuwien.ac.at
http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien

-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/ms-tnef
Size: 5330 bytes
Desc: not available
Url : http://zeus.theochem.tuwien.ac.at/pipermail/wien/attachments/20060102/ddc0f4cc/attachment.bin