This thread is locked.Only browsing is available.
Top Page > Browsing
Tips on Compilation of OpenMX in POWER7 machines
Date: 2012/07/05 15:50
Name: T.Ohwaki

Dear users,

Here are indicated a few ways of compiling the OpenMX 3.6 code
for the IBM's POWER7-CPU machines.
Hopefully the following information may be helpful for
OpenMX users using the IBM's POWER-platforms.

There are two methods as follows;

(A) Case of using "mpcc" and "mpif90":

CC = mpcc -q64 -O2 -qnosmp -qarch=pwr7 -qtune=pwr7 -qstrict \
-qlanglvl=extended -qmaxmem=-1 -Dnoomp -Df77 -Dnosse \

LD = mpif90 -64 -noparallel

LIB = -L/***** -llapack \
-L/***** -lblas \
-L/***** -lfftw3 \
-L/***** -lxlopt

openmx: $(OBJS)
$(LD) $(OBJS) $(STACK) $(CS) $(LIB) -lm -lc -o openmx

(B) Case of usign only "mpif90":

CC = mpif90 -O2 -Dnoomp -Df77 -Dnosse \

LIB = -L/***** -llapack \
-L/***** -lblas \
-L/***** -lfftw3 \
-L/***** -lxlopt

openmx: $(OBJS)
$(CC) $(OBJS) $(STACK) $(CS) $(LIB) -lm -o openmx

We strongly recommend users to refer to the first method (A)
because this method is the most optimum one.

Note that the above methods were confirmed and valid
only in the HITACHI SR16000 M1 (OS: AIX 7.1)
at Hokkaido University.

So if you would like to compile the OpenMX code
at other POWER-processor machines,
you need to try to alter several parts of
these compilation sentences appropriately
according to the machine systems.

* * * * * *

We would like to thank Prof. M.Omiya
(Information Initiative Center, Hokkaido Univ.)
for suggesting the most optimum method
of compiling the OpenMX code
in the SR16000 M1 at Hokkaido Univ..

Page: [1]