Post #2797276
2026-04-16 08:29 UTC
@lisyarus@mastodon.gamedev.place I would expect it to go into more details about how the epsilon itself may be defined “sensibly” (or at least: more than the bullets in the user input case study ;-)).
2bis. Well, then I consider myself lucky to never having seen it used in that context 8-D
3. Which claim, sorry? That GJ is a poor algorithm in general or that the loss of precision due to only checking for 0 rather than epsilon can become catastrophic? Because
3/n
Replies (1)
-
@oblomov@sociale.network 2026-04-16 09:01
@lisyarus@mastodon.gamedev.place for the latter it's mainly personal experience with heavy-duty numerical codes, but that might just be my application field (science); videogames is likely to be much more tolerant towards such numerical anomalies, leading to funny videos about bouncing tanks or speedrunning exploits. 4/4