Thread 105737399 - /g/ [Archived: 749 hours ago]

Anonymous
6/28/2025, 10:44:06 PM No.105737399
20250628_164250
20250628_164250
md5: 34d6e5fb7b38016eb63b50de7dd23497🔍
Is this really true? Why can't we just use both? What's the point of war mongering for one over the other?
Replies: >>105737427 >>105737496 >>105737520
Anonymous
6/28/2025, 10:46:36 PM No.105737427
>>105737399 (OP)
Yes, AVIF is hardware accelerated because it's based on the video codec AV1. Video codecs actually need hardware accel, image codecs don't.
>Why can't we just use both?
Blame Google. We easily could, if not for them.
Replies: >>105737690
Anonymous
6/28/2025, 10:54:43 PM No.105737496
>>105737399 (OP)
Hardware acceleration for something that previously worked well with zero complaints is absolute retardation since old devices would get fucked; besides, fuck having even more dark silicone that's basically never used.
Replies: >>105737690 >>105739290
Anonymous
6/28/2025, 10:57:12 PM No.105737520
>>105737399 (OP)
We don't need anything other than PNG/APNG.
Anonymous
6/28/2025, 10:59:19 PM No.105737529
>needing hardware acceleration for a fucking static image
Replies: >>105737652 >>105737690 >>105740523
Anonymous
6/28/2025, 11:31:45 PM No.105737652
>>105737529
this
wtf is this autism
Anonymous
6/28/2025, 11:45:05 PM No.105737690
>>105737427
>>105737496
>>105737529
Isn't it supposed to give you better battery life? Old junk not being able to decode would create a lot of e-waste I guess.
Replies: >>105739406
Anonymous
6/29/2025, 3:18:49 AM No.105739210
476616720_17851834317389386_8064249241512471811_n
476616720_17851834317389386_8064249241512471811_n
md5: 33feb0447fbde3495e1d8a181cdd8d41🔍
Has anyone tried the new iq tune when encoding AVIF images? The only big difference between AVIF and JXL seems to be when trying to compress photographic images instead of pictures of big fat anime tiddies, JXL usually does like 20% better on average but the IQ tune seems to have chipped away at this advantage.

"- Compression Efficiency Improvements
* Tuning mode AOM_TUNE_IQ improves image compression efficiency on the CLIC dataset by up to 12% for the same SSIMULACRA 2 score, up to 14% for the same DSSIM score, and up to 17% for the same Butteraugli score.

https://www.phoronix.com/news/AOM-AV1-3.12
Anonymous
6/29/2025, 3:30:27 AM No.105739290
1678318160693615
1678318160693615
md5: 2a2b6cf294c72e8b79c12885d290c607🔍
>>105737496
>besides, fuck having even more dark silicone that's basically never used.
Good luck with that. ARM already has instruction specifically built to speed up *Javascript*.
Replies: >>105740499 >>105740524
Anonymous
6/29/2025, 3:46:37 AM No.105739406
>>105737690
No "old junk" that can connect to a 2g network or better has any difficulty with image decoding.
Replies: >>105740093
Anonymous
6/29/2025, 5:21:15 AM No.105740093
owl-surprised
owl-surprised
md5: 329633a5be963f2759419e9596ab4f23🔍
>>105739406
With JXL it shouldn't but AVIF seems to be significantly more demanding than webp. Ironically that probably means webp will practically speaking be adopted next despite it having no hardware acceleration. I say this because I still see brand new laptops in stores being sold with fucking celerons from 2016. This shouldn't even be legal desu.
Replies: >>105741712
Anonymous
6/29/2025, 6:31:37 AM No.105740499
>>105739290
the point of RISC is to make the most frequent operations into fast instructions and allow infrequent operations to be done with multiple simpler instructions. arm devices run JS all the time, it's very RISC-y to make a single instruction to speed it up
s0ychan
6/29/2025, 6:37:34 AM No.105740523
>>105737529
it may matter if you're a web server encoding many images, like generating 4chan thumbnails
Anonymous
6/29/2025, 6:37:43 AM No.105740524
>>105739290
FJCVTZS is really just "float to int with x86 semantics", which is what the JS spec encodes. It’s also useful for x86 emulators.
Anonymous
6/29/2025, 10:28:53 AM No.105741712
>>105740093
at least webp despite it's age has better lossless compression than AVIF