Hi Amid, On Wed, 05 Jun 2019 12:25:06 -0000 "Amitoz AZAD" <rteja...@gmail.com> wrote:
> I have been using pyopencl to do to processing on big numpy matrices for > quite some time. > > I need to process large sparse matrices this time. It is just not possible > to convert them numpy and then do the processing ( as this will consume all > the memory on the device). I am familiar with using with the scipy sparse > matrixes and was wondering if any work has been done to deal with them using > pyopencl. In my project I am using intensively sparse matrix x dense vector multiplication using the sparse representation CSR. The Python code is here: https://github.com/silx-kit/pyFAI/blob/master/pyFAI/opencl/azim_csr.py The corresponding OpenCL code is here: https://github.com/silx-kit/pyFAI/blob/master/pyFAI/resources/openCL/ocl_azim_CSR.cl It operates in single precision but performs the accumulation of the results using compensated arithmetics to reduce the numerical noise. There is also a single-threaded implementation to have it running on macintosh computers. If needed, I can provide some help. Cheers, -- Jérôme Kieffer _______________________________________________ PyOpenCL mailing list -- pyopencl@tiker.net To unsubscribe send an email to pyopencl-le...@tiker.net
