On Sun, 05 Feb 2017 13:58:59 -0600 Andreas Kloeckner <li...@informa.tiker.net> wrote:
> Nick Curtis <nicholas.cur...@uconn.edu> writes: > > I was more wondering if there was support for translating the pyopencl > > kernel call--specifically the wrapper generated by loopy that generates the > > kernel, adds arguments, etc.--into the C equivalent. > > No, no such support right now. But that code is due a generalization > anyway, for example to be able to also work with ispc and CUDA. Hi Nick and Andreas, To jump on Nick's question, I searched recently to dump the assembly code generated after compilation: As you may know, nvidia discontinued the support for OpenCL in their profiler (nvvp) and dumping the PTX, after minor modification, it should be able to profile it with NVVP. Apparently their PTX assembly is available with: clGetProgramInfo(program, CL_PROGRAM_BINARIES, ...); as described in: https://arrayfire.com/generating-ptx-files-from-opencl-code/ I believe this appliaction may have some usage :) Cheers, -- Jérôme Kieffer tel +33 476 882 445 _______________________________________________ PyOpenCL mailing list PyOpenCL@tiker.net https://lists.tiker.net/listinfo/pyopencl
