PHPC - CONJUGATE GRADIENT PROJECT

HOWTO COMPILE AND RUN

Requirements :

• a recent compiler (like gcc or intel)
• a BLAS library (like openblas or intel MKL)

compile on SCITAS clusters :

$ module load gcc openblas
$ git clone ssh://git@c4science.ch/source/CG-PHPC.git
$ make

run on a SCITAS cluster interactively (reservation name is indicative) :

$ salloc -N 1 -n 1 --reservation=phpc-reservation
$ srun ./conjugategradient lap2D_5pt_n100.mtx

You should see this output (timing is indicative) :

$ srun ./conjugategradient lap2D_5pt_n100.mtx 
size of matrix = 10000 x 10000
Call cgsolver() on matrix size (10000 x 10000)
	[STEP 488] residual = 1.103472E-10
Time for CG = 36.269389 [s]

The given example is a 5-points stencil for the 2D Laplace problem. The matrix is in sparse format.

The matrix format is Matrix Market format (.mtx extension). You can use other matrices there or create your own.