Page 2 of 2

Re: Looking for help to identify a compiler bug in Clang uncovered by LuxMark/LuxCore

Posted: Wed Oct 19, 2022 9:44 pm
by Dade
In C99, fminf() of a number and NaN alaways returns the number: https://en.cppreference.com/w/c/numeric/math/fmin

If it is the same in OpenCL C, LuxMark v3.1 code is correct.

Re: Looking for help to identify a regression when using Clang-based OpenCL platform with LuxMark/LuxCore

Posted: Fri Oct 21, 2022 1:04 am
by illwieckz
That's also what Karol Herbst told me, that he expected that `min(1.0, infinty)` should return `1.0`… I believe it makes sense but I'm not a compiler wizard.

Here are some talks about it: https://github.com/llvm/llvm-project/issues/54947#issuecomment-1280119445