John, Thanks. I've tested ex19 on our machine with the following configure options and I don't see any slowdown issues using 2 or more threads.
Configure options: --with-threadcomm=1 --with-pthreadclasses --with-openmp --with-debugging=0 --with-blas-lapack-dir=/soft/com/packages/intel/13/079/mkl Using C compiler: mpicc -Wall -Wwrite-strings -Wno-strict-aliasing -Wno-unknown-pragmas -O -fopenmp ${COPTFLAGS} ${CFLAGS} Using Fortran compiler: mpif90 -Wall -Wno-unused-variable -Wno-unused-dummy-argument -O -fopenmp ${FOPTFLAGS} ${FFLAGS} ----------------------------------------- Using include paths: -I/nfs2/abhyshr/petsc-dev/opt-mode-with-mkl/include -I/nfs2/abhyshr/petsc-dev/include -I/nfs2/abhyshr/petsc-dev/include -I/nfs2/abhyshr/petsc-dev/opt-mode-with-mkl/include -I/usr/include/mpich2 ----------------------------------------- Using C linker: mpicc Using Fortran linker: mpif90 Using libraries: -Wl,-rpath,/nfs2/abhyshr/petsc-dev/opt-mode-with-mkl/lib -L/nfs2/abhyshr/petsc-dev/opt-mode-with-mkl/lib -lpetsc -lX11 -lpthread -Wl,-rpath,/soft/com/packages/intel/13/079/mkl/lib/intel64 -L/soft/com/packages/intel/13/079/mkl/lib/intel64 -lmkl_intel_lp64 -lmkl_gnu_thread -lmkl_core -lgomp -lpthread -lm -L/usr/lib/gcc/x86_64-linux-gnu/4.4.7 -L/usr/lib/x86_64-linux-gnu -L/lib/x86_64-linux-gnu -lmpichf90 -lgfortran -lm -L/usr/lib/gcc/x86_64-linux-gnu/4.6 -lgfortran -lm -lquadmath -lm -ldl -lmpich -lopa -lmpl -lrt -lcr -lpthread -lgcc_s -ldl There are no external packages and MKL is using GNU compilers -lmkl_intel_lp64 -lmkl_gnu_thread -lmkl_core -lgomp -lpthread instead of Intel compiler that you have -lmkl_intel_lp64 -lmkl_intel_thread -lmkl_core -liomp5 -lpthread Shri On Oct 31, 2012, at 1:35 PM, John Fettig wrote: > On Wed, Oct 31, 2012 at 2:15 PM, Shri <abhyshr at mcs.anl.gov> wrote: >> Dave, >> I configured PETSc with MKL on our machine and tested ex2 using the >> options you sent (./ex2 -pc_type jacobi -m 1000 -n 1000 -threadcomm_type >> pthread -threadcomm_nthreads 1). However, I could not reproduce the problem >> you encountered. Using more threads did not reproduce it either. What >> configure options did you use? > > Shri, > > I had this problem as well, and you can read my configure options from > the pthreads.out log summary that I attached at the beginning of this > thread. > > Regards, > John