search for: shinpei0208

.../bugs.freedesktop.org/attachment.cgi?id=51217) Shinpei's Kato talk Before being able to run GPGPU kernels on the GPU, we need to have an API to handle the upload, allocate memory and read/write from/to the GPU. We think that our runtime should follow the CAL specifications (https://github.com/shinpei0208/cal-spec) and so should every other driver. If you would like to do this task, you'll have to communicate implement the interface for nvidia but also try to make it portable so as AMD could also use this library to send commands to their GPU. Please have a look at Shinpei's presentation....

...hat we should do. This can basically be summed up as using the LLVM IR for compute. This means we need to implement LLVM backends for our GPUs. See the Mesa_compilers.pdf attachment. Also, the runtime bits should be written according the Compute Abstraction Layer specifications (https://github.com/shinpei0208/cal-spec). This bug report is actually a meta one that is tracking the development of OpenCL on Nouveau. -- Configure bugmail: https://bugs.freedesktop.org/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are the assignee for the bug.