karolherbst wrote:
> > this PR breaks atan2:
> > `ERROR: atan2: -3294199.000000 ulp error at {-0x1.fffffep+127 (0xff7fffff),
> > -0x1.fffffep+127 (0xff7fffff)}: *-0x1.2d97c8p+1 vs. -0x1.921fb6p+1
> > (0xc0490fdb) at index: 198`
>
> On what, with what? Can you post the compiler output? If I interpret the IR
> with these values, I get the expected result
Tested on Intel with rusticl. Though atm we don't support denorms with fp32 and
I'm wondering if that's related? I found a couple of other regressions and
created reverts for those for now until I have time to dig into it: #189781
But could also be just some hardware inaccuracy or something else.
https://github.com/llvm/llvm-project/pull/188706
_______________________________________________
cfe-commits mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
