Closed haonanya closed 2 days ago
I think the current expressions are wrong indeed, thanks for reporting it! Ideally we'd align with unary_float.cpp
and use s[j]
and t[j]
.
Are you happy to open a PR to fix it?
I think the current expressions are wrong indeed, thanks for reporting it! Ideally we'd align with
unary_float.cpp
and uses[j]
andt[j]
.Are you happy to open a PR to fix it?
Sure! Thanks for quickly reply.
Looking through the other tests, I think test_conformance/math_brute_force/macro_unary_double.cpp
may have a similar issue (it's a multi-threaded test that directly indexes into gIn
without taking the thread ID into account).
All issues have been fixed by #1995 now.
https://github.com/KhronosGroup/OpenCL-CTS/blob/main/test_conformance/math_brute_force/unary_double.cpp#L292,L293
input is cl_double p = (cl_double )gIn + thread_id buffer_elements; instead of (cl_double )gIn)[j], output ref is cl_double r = (cl_double )gOut_Ref + thread_id * buffer_elements; instead of gOut_Ref. Can you please double check? Thanks very much.
+@svenvh, @bashbaug