[libc-commits] [libc] [llvm] [libc][math] Implement single precision lgamma function (PR #205490)

[via libc-commits]

lntue wrote:

> the OpenGroup standard says that signgam is unspecified for NaN, -Inf or a negative integer. Thus for x=-0x1.eep+6 it should be set if you follow that standard.

Thanks Paul for the review!  About signgam for POSIX standard, most likely for regular functions, we will just stick to C standard and ignore the global signgam (by default, at least).  Instead, we will provide the POSIX lgamma_r functions that return the sign.

If there are real needs for the global signgam with the lgamma functions, we will provide a build option for that later.

https://github.com/llvm/llvm-project/pull/205490