for one I mistyped the option - it should be: --with-cudac='nvcc -ccbin icc' But then - should it not be c++? [as cuda internally uses g++ - not gcc]. So --with-cudac='nvcc -ccbin icpc'. But this gives me all kinds of wierd errors in configure.log.. Maybe it won't work [with intel compilers].. Satish On Fri, 24 Feb 2012, John Fettig wrote:

Unfortunately it doesn't work. I'll send the logs to petsc-maint, but here's the error. If you remove these lines:

#ifndef PETSC_HAVE___INT64 #define PETSC_HAVE___INT64 1 #endif

from $PETSC_ARCH/include/petscconf.h, it builds ok (and make test passes). Here's the error:

nvcc -O -arch=sm_20 -c --compiler-options=-fPIC -I/home/jfe/local/petsc-dev/include -I/home/jfe/local/petsc-dev/intel-opt-cuda-precise/include -I/usr/local/cuda/include -I/home/jfe/local/petsc-dev/intel-opt-cuda-precise/include/txpetscgpu/include -I/usr/local/encap/platform_mpi-8.01/include -D__INSDIR__=src/vec/vec/impls/seq/seqcusp/ veccusp.cu /home/jfe/local/petsc-dev/include/petscsys.h(219): error: identifier "__int64" is undefined

John

On Fri, Feb 24, 2012 at 10:53 AM, Satish Balay <[email protected]> wrote:

...

...

...

...

...

...

...

...

...

...

...

--compiler-bindir <directory> (-ccbin) Specify the directory in which the compiler executable (Microsoft Visual Studion cl, or a gcc derivative) resides. By default, this executable is expected in the current executable search path. <<<<<<<<

It says it will look for the default compiler in the specified path - and doesn't mention that it can switch the compiler. Ah - the manpage is misleading.. Looks like '--ccbin icc' works..

balay@bb30:~/junk>nvcc -v -ccbin icc sizeof.c #$ _SPACE_= #$ _CUDART_=cudart #$ _HERE_=/usr/local/cuda/bin #$ _THERE_=/usr/local/cuda/bin #$ _TARGET_SIZE_=64 #$ TOP=/usr/local/cuda/bin/.. #$ LD_LIBRARY_PATH=/usr/local/cuda/bin/../lib:/usr/local/cuda/bin/../extools/lib:/soft/intel/11.0.081/lib/intel64:/soft/intel/11.0.081/lib/intel64:/usr/local/cuda/lib64:/usr/local/cuda/lib64 #$ PATH=/usr/local/cuda/bin/../open64/bin:/usr/local/cuda/bin:/soft/intel/11.0.081/bin/intel64:/soft/intel/11.0.081/bin/intel64:/usr/mpi/gcc/mvapich2-1.5/bin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/usr/games:/usr/local/cuda/bin:/usr/local/cuda/bin #$ INCLUDES="-I/usr/local/cuda/bin/../include" "-I/usr/local/cuda/bin/../include/cudart" #$ LIBRARIES= "-L/usr/local/cuda/bin/../lib64" -lcudart #$ CUDAFE_FLAGS= #$ OPENCC_FLAGS= #$ PTXAS_FLAGS= #$ icc -c -x c "-I/usr/local/cuda/bin/../include" "-I/usr/local/cuda/bin/../include/cudart" -m64 -o "/tmp/tmpxft_00000199_00000000-1_sizeof.o" "sizeof.c" #$ icpc -m64 -o "a.out" -Wl,--start-group "/tmp/tmpxft_00000199_00000000-1_sizeof.o" "-L/usr/local/cuda/bin/../lib64" -lcudart -Wl,--end-group balay@bb30:~/junk>

So '--with-cc=icc' --with-nvcc='nvcc --ccbin icc' should work.. Will check..

Satish

On Fri, 24 Feb 2012, John Fettig wrote:

...

Some time ago I asked about using icc with nvcc. You can use icc with nvcc by passing it as an argument:

nvcc -ccbin /path/to/icc

Has anybody tried building PETSc with CUDA this way?

John