Issue |
140252
|
Summary |
__builtin_round() uses dynamic rounding mode on x86 (only)
|
Labels |
new issue
|
Assignees |
|
Reporter |
enh-google
|
my assumption was that `__builtin_round()` was meant to be like the `round()` libc function, and for arm and riscv, that's how it behaves: Android's libc can use `__builtin_round()` to implement `round()`.
for x86/x86-64 though, `__builtin_round()` seems to use the _dynamic_ rounding mode rather than hard-coding "away from zero" as specified by round().
seems like a bug?
_______________________________________________
llvm-bugs mailing list
llvm-bugs@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-bugs