Post #2797274
2026-04-16 08:19 UTC
@lisyarus@mastodon.gamedev.place
1 & 4. (I still think they're related) . My comment on 1. wasn't to contradict what you're saying, but to point out that when epsilon comparisons are made (or if you prefer: when they do make sense), they should at the very least be used correctly, and this also means choosing the epsilon based on considerations about where you are (or expect to be) on the FP line, how you got to the results, etc.
Or if you prefer:
1/n
Replies (1)
-
@oblomov@sociale.network 2026-04-16 08:23
@lisyarus@mastodon.gamedev.place the approxEqual implementation known is bad not because one often don't need to use it, but because when one does need to use it, it may fail spectacularly because e.g. it's being used with x, y of the “wrong” magnitude (either too small or too large) for that somewhat arbitrary choice of epsilon. And sure enough, a lot of poorly written numerical code *does* have that issue, but in a discussion about whether epsilon checks are needed or not, 2/n