xref: /freebsd/contrib/arm-optimized-routines/math/aarch64/experimental/README.contributors (revision f3087bef11543b42e0d69b708f367097a4118d24)
1*f3087befSAndrew TurnerCode in this sub-directory should follow the GNU Coding Standard, but it is
2*f3087befSAndrew Turnernot expected to be upstreamed into glibc without modification, so
3*f3087befSAndrew Turnerglibc-specific conventions need not be followed.
4*f3087befSAndrew Turner
5*f3087befSAndrew TurnerThe requirements for portable code apply to non-portable code with the
6*f3087befSAndrew Turnerfollowing differences:
7*f3087befSAndrew Turner
8*f3087befSAndrew Turner1. Worst-case ULP error should be encoded in filenames (e.g. sin_u35.c). There
9*f3087befSAndrew Turner   are no specific restrictions on acceptable ULP error, but if functions
10*f3087befSAndrew Turner   provide significantly less accuracy than portable equivalents then a clear
11*f3087befSAndrew Turner   justification for inclusion should be stated in comments at the top of the
12*f3087befSAndrew Turner   source file. Error bounds of the approximation should be clearly documented
13*f3087befSAndrew Turner   in comments.
14*f3087befSAndrew Turner
15*f3087befSAndrew Turner2. Functions are assumed to support round-to-nearest mode by default, unless
16*f3087befSAndrew Turner   stated; other rounding modes are not required to be provided.
17