tests for cl_khr_spirv_queries

[bashbaug]

See: KhronosGroup/OpenCL-Docs#1385

[kpet]

If the SPIRV-Headers become a hard dependency, maybe we should just generate this file at build time?

[karolherbst]

looks good, tested my implementation of cl_khr_spirv_queries against this PR and everything passes.

Though I think if cl_khr_spirv_extended_debug_info is supported you could also check if OpenCL.DebugInfo.100 is.

[bashbaug]

Though I think if cl_khr_spirv_extended_debug_info is supported you could also check if OpenCL.DebugInfo.100 is.

Good catch. I added this, although I don't have a device to test it with, and according to opencl.gpuinfo.org there aren't any devices in the database that support it, either, hrm.

As an aside, the tests are currently (informationally) printing any non-required SPIR-V capabilities that are reported. Should I do something similar for SPIR-V extended instruction sets and SPIR-V extensions also? This might help to find other missing checks like this one... or it could just be noise, once usage of the SPIR-V queries becomes more widespread.

[karolherbst]

Though I think if cl_khr_spirv_extended_debug_info is supported you could also check if OpenCL.DebugInfo.100 is.

Good catch. I added this, although I don't have a device to test it with, and according to opencl.gpuinfo.org there aren't any devices in the database that support it, either, hrm.

don't have tests here for that either.

As an aside, the tests are currently (informationally) printing any non-required SPIR-V capabilities that are reported. Should I do something similar for SPIR-V extended instruction sets and SPIR-V extensions also? This might help to find other missing checks like this one... or it could just be noise, once usage of the SPIR-V queries becomes more widespread.

I think with this extension in mind it's fine to report more capabilities or extensions than necessary. They can also be part of vendor/ext extensions the tests here isn't aware off, so I'm pretty sure it's gonna cause some noise sooner or later.

[karolherbst]

need to add "cl_khr_spirv_queries" inside test_conformance/compiler/test_compiler_defines_for_extensions.cpp

otherwise test_compiler compiler_defines_for_extensions will fail: "FAIL: Extension cl_khr_spirv_queries is not in the list of approved Khronos extensions!"

This change could go on its own, but I think it's fine adding it here if this PR gets merged soonish.

[bashbaug]

need to add "cl_khr_spirv_queries" inside test_conformance/compiler/test_compiler_defines_for_extensions.cpp

Fixed.

I really want to get that test fixed; it's kind of silly how we need to modify it every time a new KHR extension gets published.

[karolherbst]

need to add "cl_khr_spirv_queries" inside test_conformance/compiler/test_compiler_defines_for_extensions.cpp

Fixed.

I really want to get that test fixed; it's kind of silly how we need to modify it every time a new KHR extension gets published.

yeah.. this list really should be generated