first success story on NEC SX

Rob Ross rross at mcs.anl.gov
Wed Jul 30 10:29:06 CDT 2003


Hi Rene,

I'll try to integrate your suggested changes and release a 0.8.8 today, 
which will hopefully build for you without need for modifications :).

Thanks for your patience!

Rob

On Wed, 30 Jul 2003, rene.redler wrote:

> 
> Hello,
> 
>   after some Makefile editing tests ran successfully on a SX.
> 
> > Could you tell me more about the 'other odd
> > things'?  It may drive me crazy, but we're going to try to get the
> > Makefiles to work with as many make implementations as possible.
> 
> Starting again with version 0.8.7 it does not work when using the cross 
> compiler on a front end machine (in my case a Linux PC). When specifying 
> only MPIF77 and MPICC for using the cross compiler some configure test are 
> done by using the g++ on the Linux system, where in particular pointer 
> size do not match with those on the SX. When specifying CC and CXX with 
> the cross compiler the configure fails since the executable that are 
> generated for the SX cannot be executed on the Linux system.
> 
> On the SX the following environment variables have been defined
> 
> MPIF77=mpif90
> MPICC=mpic++
> CC=c++
> CXX=c++
> ARFLAGS=r
> 
> Makefile.in in directory /src/lib
> 
> #       v1hpg.c
> #       ncio.c \
> #       putget.c \
> 
> PACKING_LIST = \
> 
> causes the whole PACKING_LIST to be in the comment line as well
> 
> same for
> 
> #       onstack.h \
> 
> LIB_OBJS = $(LIB_CSRCS:.c=.o)
> 
> I needed to eliminate the backslashes after putget.c and onesteck.h.
> 
> make and make install worked fine with these modifications.
> 
> 
> Makefile.in in directory test:
> 
> pushd and popd are not supported on the SX. I replaced
> 
> do pushd . ; cd $$i && $(MAKE); popd; done
> 
> by
> 
> do cd $$i && $(MAKE); cd .. ; done
> 
> 
> In directories test_int, test_float, test_double_int and test_double I 
> replaced the Makefile.in by the one that is attached to this mail.
> 
> test_read in test_int, test_float and test_double worked fine.
> 
> Rene
> 
> 




More information about the parallel-netcdf mailing list