On IBM Power 6, which subrontines need the lowest optimization level?

[Brane]

Hi All,

I tried to compile VASP on an IBM Power 6 system using the modifed makefile.rs6000 or makefile.sp2. The compilation has no problem. However, the calculated total energy is wrong when I use -O3 or -O2 opimization level to compile VASP. If the optimiztion level is downgraded to the lowest one (-O0), as expected, the calculated energy is correct. A similar problem was reported ealier: http://cms.mpi.univie.ac.at/vasp-forum/ ... php?2.4616

I think I have to use the lowest optimiztion level for certain subroutines. I tried to use -O0 for radial, nonl, nonlr, paw, pseudo, and wave without any success. I can confirm that there is no problem to use -O3 for vasp.lib. Does anybody have a clue on this?

Thanks.

Brane

[admin]

Dear colleague,
please find a makefile optimized for IBM-Power 6 (makefile.pwr6.64) in the latest vasp-release download on our ftp server

[ktao]

Hi,

I download the new makefile (makefile.pwr6.64), and tried to compile VASP on IBM-Power 6. However, I encountered some problems:

1, I deleted aedens.o from the makefile for I could not find it in the directory;

2, I still could not compile it because of the lack of the path to -lscalapack -lpblas -lblacs -ltools. I added it as: -L /afs/ipp/@sys/lib

3, However, I failed again and got fellowing error information:

ld: 0711-738 ERROR: Input file ../vasp.4.lib/linpack_double.o:
XCOFF32 object files are not allowed in 64-bit mode.
make: 1254-004 The error code from the last command is 8.

Any suggestion would be appreciated, thanks.

Kun Tao

PS: If I want to do some calculations with more than 100 atoms (including 5 elements), should I recompile the VASP again?

[ravi]

I have problem in using the makefile.ibm_hlrn . I am getting the following error when using it

###
make: 1254-002 Cannot find a rule to create target OFLAG from dependencies.
Stop.
###

Do you know how to resolve it?