bcalidas

We pass with the following configuration For clGetDeviceInfo 1) Report CL_DEVICE_SVM_COARSE_GRAIN_BUFFER under DEVICE_SVM_CAPABILITIES 2) Report CL_DEVICE_ATOMIC_SCOPE_ALL_DEVICES under CL_DEVICE_ATOMIC_MEMORY_CAPABILITIES 3) Report __opencl_c_atomic_scope_all_devices under CL_DEVICE_OPENCL_C_FEATURES With this combination, the meaning of __opencl_c_atomic_scope_all_devices...

This proposal looks reasonable.

As mentioned in an earlier comment, OpenCL does not importing an external handle into an already created semaphore. This avoids the need for introducing the concept of temporary handle import.

Option 3 looks to be best.

We would strongly prefer option 1. Under option 2, we are forced to change our implementation to calculate values at a higher precision. This leads to loss of performance which...

Do we have agreement that it is conformant to implement dot and cross products with fp16 precision.

We have concerns about the channel mappings for image reads. This does not align with what we have currently. We also have concerns about writing to YUV images. We only...

Sharing some initial feedback. 1) For cl_ext_image_tiling_control, is there really a need for the CL_DEVICE_IMAGE_TILING_DEVICE_ACCESS_EXT token. 2) For cl_ext_image_tiling_control does CL_DEVICE_IMAGE_TILING_HOST_ACCESS_EXT relate to whether or not the implementation will detile...

Some initial thoughts 1) It should be possible for (non-linear) tiled images to be created with CL_MEM_USE_HOST_PTR. I believe the spec already allows for this implicitly. 2) It would be...

I wanted to leave some high-level thoughts about how this extension could work. The major impact of linear vs optimal tiling modes would be when the image is created from...