6 Feb
2024
6 Feb
'24
1:48 p.m.
On Tue Feb 6 09:52:52 2024 +0000, Giovanni Mascellani wrote:
Why do you use two different approaches for sine (just setting the ULP tolerance) and cosine (quantizing and handling differently SM1-5 and SM6)? Not that I have a big problem with it, mostly curious. I'm not sure why the variation is so large for `asin()`in particular, but that's the reason.
-- https://gitlab.winehq.org/wine/vkd3d/-/merge_requests/602#note_60221