Kévin Petit

Currently specified as follows >To create an OpenCL context from a specific device type, call clCreateContextFromType >Only devices that are returned by clGetDeviceIDs for device_type are used to create the...

Image conversion rules duplicated between environment spec and OpenCL C spec

1

See https://github.com/KhronosGroup/OpenCL-Docs/pull/1352 for an example of an extension that needs to add the same (lenghty) descriptions twice.

c11_atomics `{svm_,}atomic_fence` test issues

1

Several implementers have reported issues over the years with the `atomic_fence` and `svm_atomic_fence` tests, the latest being #2496. Analysis from a few years ago (e.g. https://www.khronos.org/members/login/bugzilla/show_bug.cgi?id=15814) suggests that the tests...

There are a number of places where we should make use of the SPIR-V validator: * to check the output of the offline compilation flow * to check that the...

See https://github.com/KhronosGroup/OpenCL-Docs/pull/217.

The CTS doesn't have consistency tests to check that when `CL_DEVICE_{COMPILER,LINKER}_AVAILABLE` return `CL_FALSE` `clBuildProgram`, `clCompileProgram` and `clLinkProgram` return `CL_{COMPILER,LINKER}_NOT_AVAILABLE` when appropriate.

These queries are loosely specified (see https://github.com/KhronosGroup/OpenCL-Docs/issues/763 for the spec discussion) and the value returned by implementations is not validated by the CTS (except in the testing for `cl_khr_dx9_media_sharing`).

- The limit for SPIR-V seems to be the universal limit for the size of a literal string: 65536 - I could not find a limit for OpenCL C after...