Conversation
|
@bashbaug I have added one more commit that fixes an issue with |
| /* "signed" and "unsigned" type names are valid (int is implied) */ | ||
| if (arg_type.find("signed") != std::string::npos) |
There was a problem hiding this comment.
Is this guaranteed to be valid? Neither signed nor unsigned (without the int) are listed in https://registry.khronos.org/OpenCL/specs/3.0-unified/html/OpenCL_C.html#built-in-scalar-data-types.
There was a problem hiding this comment.
OK, i didn't check the list of the builtin scalar data types; however, in the code of the test, there is "signed" and "unsigned" (both without int) in the std::vector returned from method generate_all_type_arguments(cl_device_id device) and then the code calls get_param_size on each element of that vector.
I don't remember exactly, but IIRC the problem was that the get_param_size returned 0 for both these "implied int" types, and as a result total_param_size was miscalculated by 8 bytes, which then triggered an error because the actual total size of arguments was over limit, when used with SPIR-V compilation mode. IIRC Clang doesn't really care about PoCL's argument-related limits, so it will accept larger actual arguments, but the SPIRV translation path isn't so forgiving. I could try to recreate the error & get more details if you'd like.
test_conformance/run_conformance.py
Outdated
| if result == 0 or result == 2 or result == -100: | ||
| log_file.write(" (" + get_time() + ") Test " + test_name + " passed in " + str(run_time) + "s\n") |
There was a problem hiding this comment.
Interested in feedback from others in the working group: Should we log test skips separately? Or is it OK to log these the same as a passing test?
There was a problem hiding this comment.
I realized it's currently inconsistent, so I updated it print skipping separately from passing in both places where it's printed.
If the queried type was "signed" or "unsigned" (with an implied int), the returned size was zero.
"consistency_il_programs" test was using bogus SPIRV input. However, clCreateProgramWithIL() must return CL_INVALID_VALUE for invalid SPIRV: ... CL_INVALID_VALUE if the length-byte block of memory pointed to by 'il' does not contain well-formed intermediate language input that can be consumed by the OpenCL runtime. therefore the test was triggering two different error conditions (with different error values) simultaneously. Accept both error values.
KhronosGroup#2454 added a REQUIRE_EXTENSION("cl_khr_spir"); which returns -100 (TEST_SKIPPED_ITSELF) when the device doesn't have the extension. The run_conformance.py considers all nonzero values to be errors. This fixes the runner script to accept SKIP values too.
2 participants
No description provided.