Difference between revisions of "S3D"
Line 10: | Line 10: | ||
used: | used: | ||
− | * | + | * make.SP2 - (AIX), this is copied from my old S3D installation |
<pre> | <pre> | ||
COMPILE_F90 = mpxlf95_r -g -O3 -qstrict -qarch=auto -qtune=auto -qsuffix=cpp=f90 -q64 | COMPILE_F90 = mpxlf95_r -g -O3 -qstrict -qarch=auto -qtune=auto -qsuffix=cpp=f90 -q64 | ||
Line 19: | Line 19: | ||
-qrealsize=8 -WF,-Dflush=FLUSH_ -WF,-D$(MACH) -lesslp2 | -qrealsize=8 -WF,-Dflush=FLUSH_ -WF,-D$(MACH) -lesslp2 | ||
</pre> | </pre> | ||
− | * | + | * make.LINUX (Using OpenMPI 1.0.2 built with icc/ifort) |
<pre> | <pre> | ||
F90=mpif90 -fpp1 | F90=mpif90 -fpp1 | ||
Line 27: | Line 27: | ||
LINK = $(F90) | LINK = $(F90) | ||
</pre> | </pre> | ||
− | * | + | * make.BGL |
<pre> | <pre> | ||
Latest revision as of 22:06, 7 March 2007
I've been able to build and run S3D on AIX 5.3, Linux (x86_64), and BGL.
Though without output confirmation, I can't really say that it's working
100%.
Here's the method I used, I didn't use the harness scripts.
1) Create a make.XXX for your machine in "Source/build". Here's what I've used:
- make.SP2 - (AIX), this is copied from my old S3D installation
COMPILE_F90 = mpxlf95_r -g -O3 -qstrict -qarch=auto -qtune=auto -qsuffix=cpp=f90 -q64 -qrealsize=8 -WF,-Dflush=FLUSH_ -WF,-D$(MACH) -c -o $@ COMPILE_F77 = mpxlf95_r -g -O3 -qstrict -qarch=auto -qtune=auto -qfixed -q64 -qrealsize=8 -WF,-Dflush=FLUSH_ -WF,-D$(MACH) -c -o $@ LINK = mpxlf95_r -g -O3 -qstrict -qarch=auto -qtune=auto -qsuffix=cpp=f90 -q64 -qrealsize=8 -WF,-Dflush=FLUSH_ -WF,-D$(MACH) -lesslp2
- make.LINUX (Using OpenMPI 1.0.2 built with icc/ifort)
F90=mpif90 -fpp1 COMPILE_F90 = $(F90) -c -o $@ COMPILE_F77 = $(F90) -c -o $@ COMPILE_C = icc -c -o $@ LINK = $(F90)
- make.BGL
COMPILE_F90 = mpixlf90 -g -O3 -qstrict -qarch=auto -qtune=auto -qsuffix=cpp=f90 -q32 -qrealsize=8 -WF,-Dflush=FLUSH_ -WF,-D$(MACH) -c -o $@ COMPILE_F77 = mpixlf90 -g -O3 -qstrict -qarch=auto -qtune=auto -qfixed -q32 -qrealsize=8 -WF,-Dflush=FLUSH_ -WF,-D$(MACH) -c -o $@ LINK = mpixlf90 -g -O3 -qstrict -qarch=auto -qtune=auto -qsuffix=cpp=f90 -q32 -qrealsize=8 -WF,-Dflush=FLUSH_ -WF,-D$(MACH)
2) Build the code with "make MACH=XXX"
3) Create the "Source/data" directory
4) Copy the "Source/input/s3d.in.XXX" file you want to "s3d.in"
5) Run "s3d_f90.x" from the "Source/run" directory with the number of processors given in s3d.in (npx * npy * npz)
All of the platform appeared to run properly. When we get some information on how to validate output, I can report about that.
Thanks, Alan