AAA – Analytical Anti-Aliasing

(blog.frost.kiwi)

513 points todsacerdoti | 5 comments | 20 Nov 24 08:03 UTC | HN request time: 0.851s | source

Show context

FrostKiwi ◴[20 Nov 24 11:20 UTC] No.42192825[source]▶

>>42191709 (OP) #

Thanks for sharing! Author here, happy to answer any questions.

replies(14): >>42193691 #>>42193836 #>>42193884 #>>42194114 #>>42195074 #>>42195398 #>>42196248 #>>42196427 #>>42196783 #>>42197142 #>>42197506 #>>42198224 #>>42198810 #>>42202062 #

1. ferbivore ◴[20 Nov 24 13:45 UTC] No.42193884[source]▶

>>42192825 #

> Mobile chips support exactly MSAAx4 [...] the driver will force 4x anyways

On what GPUs and through what APIs did you see this? This seems fairly weird. I especially wouldn't expect Apple to have problems.

replies(2): >>42197952 #>>42200340 #

2. mastax ◴[20 Nov 24 20:48 UTC] No.42197952[source]▶

>>42193884 (TP) #

I’m definitely seeing similar artifacts when at 2x on an iPhone 15 Pro.

replies(1): >>42198937 #

3. ferbivore ◴[20 Nov 24 22:48 UTC] No.42198937[source]▶

>>42197952 #

Through Metal or in a web browser? Apple's docs say it's supposed to use the standard 0.75/0.25 sample positions with 2x MSAA: https://developer.apple.com/documentation/metal/mtldevice/28...

replies(1): >>42199998 #

4. mastax ◴[21 Nov 24 01:13 UTC] No.42199998{3}[source]▶

>>42198937 #

Viewing the article in Safari.

5. FrostKiwi ◴[21 Nov 24 02:06 UTC] No.42200340[source]▶

>>42193884 (TP) #

Yes, that also surprised me and I tested it on multiple mobile Apple devices. It's not really a mistake per se, the implementation is free to do what it wants. Selecting MSAAx2 on these types of mobile GPUs simply has no upside and isn't really supported over MSAAx4 and I guess apple still wanted to make the choice somehow possible, as opposed to Android, where there is only an illusion of choice.

It's just so happens to produce visible artifacts in this case. I suppose for 3D scenes it's mostly fine.

↑