Ben Ashbaugh

missing test coverage for cl_khr_expect_assume

3

see: https://github.com/KhronosGroup/OpenCL-Docs/issues/723 We need to add CTS coverage for the `cl_khr_expect_assume` extension.

See comment on https://github.com/KhronosGroup/OpenCL-Docs/pull/738. We have several OpenCL C features that require other OpenCL C features, for example: > OpenCL C compilers that define the feature macro `__opencl_c_device_enqueue` must also...

add test for object comparability

3

See https://github.com/KhronosGroup/OpenCL-Docs/issues/797. We should add basic testing for object comparability, to ensure that object handles compare equal when they are queried from child objects.

The CTS tests currently define `isnan`, `isfinite`, and `isinf` macros for MSVC: https://github.com/KhronosGroup/OpenCL-CTS/blob/40aaa2bcf2460a083b89427cffaacdbd5f707ffc/test_common/harness/compat.h#L141-L151 Unfortunately, this means we cannot use the functions `std::isnan`, `std::isfinite`, and `std::isinf` functions from C++11 without generating...

Targeted Testing for SPIR-V Versions Greater Than 1.0

1

As far as I can tell all of the "spirv_new" tests are currently testing only SPIR-V 1.0 because the assemble_spirv.py script currently passes `--target-env spv1.0` to `spirv-as` (this is probably...

api consistency requirements for fp64 and fp16

1

We don't currently have any API consistency tests for fp64 (and fp16) but there are enough moving parts here so we probably should... assuming it's clear what the API consistency...

See: https://github.com/KhronosGroup/OpenCL-Docs/issues/790 We need to add CTS coverage for the `cl_hkr_uniform_work_group_arithmetic` extension. See also existing gaps for other work-group scan and reduction functions: https://github.com/KhronosGroup/OpenCL-CTS/issues/1364

See: https://github.com/KhronosGroup/OpenCL-ICD-Loader/pull/231 We should consider adding CI configurations to test different default calling conventions.

WIP: add an OpenCL ICD loader extension to shut down and free memory

8

This is a proposed solution to address #157. It adds an ICD loader extension that can be explicitly called by the host application (or possibly a layer?) to shut down...

do strings in the Windows resource file need a prefix

1

Found during a code review: https://github.com/KhronosGroup/OpenCL-ICD-Loader/pull/206#issuecomment-1432068544 > When looking at: https://learn.microsoft.com/en-us/windows/win32/menurc/stringfileinfo-block It seems that "040904E4" means that the values should be in U.S. English and in Multilingual character set (0x04E4...