[libc-commits] [libc] [libc][math] Implement double precision sin correctly rounded to all rounding modes. (PR #95736)

[via libc-commits]

lntue wrote:

Thanks @zimmermann6 for letting me know about the recent results at: https://inria.hal.science/hal-04480440.

With the insight from that paper, I was able to make non-FMA targets also correctly rounded to all rounding modes.

So I rearrange the range reduction headers, with `range_reduction_double_fma.h` and `range_reduction_double_nofma.h` contain their own specialized look-up tables, and fast small range reduction, while `range_reduction_double_common.h` contains double-double large range reduction and 128-bit precision small and large range reductions.

https://github.com/llvm/llvm-project/pull/95736