On Mon, 26 Feb 2024 06:43:43 GMT, Joe Darcy <da...@openjdk.org> wrote:
> To improve the "fingerprinting" coverage of the StrictMath tests, I've added > test cases where the worst-case of a non-FDLIBM library is larger than the > FDLIBM worst-case. Assuming reasonable methodology of the paper, the output > of the other math library must differ from FDLIBM at such a point. In a subsequent push, I'll fill those in for the functions not already so updated (cbrt, expm1, log10, log1p, etc.) ------------- PR Comment: https://git.openjdk.org/jdk/pull/15879#issuecomment-1963425696
