Luke I
Luke I
Hi @thornhale Thanks for opening the issue. We are aware of that. There are two causes of host to device memory transfers. The first occurs when graph node locality is...
If you are interested in re-implementing functionality into plain OpenCL 1.2 I would suggest to look at the Eigen code first. TensorFlow is closely integrated with it and most of...
I am working on creation of project thing in here.. it will contain a small task that need to be fixed ( failing tests ) so people interested in contributing...
@thornhale just small update: We are still in progress of performance improvement for AMD and mobile targets with focus on models like Resnet50 and VGG. Every commit in the upstream...
Hi @enihcam Thanks for the report. I will do my best to help. To clarify is that this device: https://www.techpowerup.com/gpudb/2197/radeon-hd-8280e ? As well, do you know if the device you...
@enihcam / @DuncanMcBain after Neo driver was released we have Skylake series SoC available for tests and benchmarks - there is nothing ad-hoc about this ;) @enihcam is there any...
Hi @kcauchy In principle.. there is ComputeCpp for Windows. Essentially, we did not try windows in Eigen yet - that would be a first step to do. Please let us...
@welnaseth the first step would be to test if Eigen is working on windows ( https://bitbucket.org/mehdi_goli/opencl/commits/branch/Eigen-Optimised-Tensor-Vector-Contraction ) Could you give it a spin?
@welnaseth thanks for checking this out! I created a separate issue so we can track that particular problem separately.