[PATCH] D127964: [DCE] Eliminate no-op atan and atan2 calls

[Rainer Orth via Phabricator via llvm-commits]

ro added a comment.

In D127964#3732256 <https://reviews.llvm.org/D127964#3732256>, @spatel wrote:

> 



> That would be the quick fix to get things passing. But does that mean the mathlib on Solaris is setting errno on "atan2(-0.0, 0.0)" ?
>
> IEEE says "atan2(±0, +0) is ±0". 
> POSIX has that case as optional - https://pubs.opengroup.org/onlinepubs/9699919799/functions/atan2.html :
> "If y is ±0 and x is +0, ±0 shall be returned."
> And also says:
> "If both arguments are 0, a domain error shall not occur." 
> ...but that's in the same optional block.

I've just checked: the difference is between clang 11.3.0 and clang 15.0.0 (and older):

- With gcc, I get both the expected return values and `errno` is `0`.
- With clang, the values remain correct, but somehow `errno` is set to `EDOM`.

However, this only happens because gcc, unlike clang, evaluates the `atan2` calls at compile time.  When compiling with `-fno-builtin-atan2`, both clang and gcc binaries behave the same.

I've found a copy of the Solaris 9 `libm` sources and am looking at the `atan2` implementation and error handling right now.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D127964/new/

https://reviews.llvm.org/D127964