Mozilla's Llamafile 0.8.2 Scores Big With New AVX2 Performance Optimizations

[qarium]

AV1 is literally the same thing...

i am pretty sure that AV1/AVIF is not the same thing or else firefox and google chrome would not add it.

[Quackdoc]

i am pretty sure that AV1/AVIF is not the same thing or else firefox and google chrome would not add it.

Then I would be a lot less sure if I were you lmao

https://aomedia.googlesource.com/aom/+/refs/heads/main/PATENTS

1.3. Defensive Termination. If any Licensee, its Affiliates, or its agents initiates patent litigation or files, maintains, or voluntarily participates in a lawsuit against another entity or any person asserting that any Implementation infringes Necessary Claims, any patent licenses granted under this License directly to the Licensee are immediately terminated as of the date of the initiation of action unless 1) that suit was in response to a corresponding suit regarding an Implementation first brought against an initiating entity, or 2) that suit was brought to enforce the terms of this License (including intervention in a third-party action by a Licensee).

the fluff is different but the core is the exact same thing "If you sue us or anyone for infringing your patents, the right to use any of our patents is revoked"

[qarium]

Then I would be a lot less sure if I were you lmao

https://aomedia.googlesource.com/aom/+/refs/heads/main/PATENTS

the fluff is different but the core is the exact same thing "If you sue us or anyone for infringing your patents, the right to use any of our patents is revoked"

then it boils down who has the power to revoke it...

in case of AV1/AVIF it looks like google is fine with that because they are the one who revoke it

it looks like the jXL people are a different untrusted entity...

[Quackdoc]

then it boils down who has the power to revoke it...

in case of AV1/AVIF it looks like google is fine with that because they are the one who revoke it

it looks like the jXL people are a different untrusted entity...

libjxl/jpeg-xl is a google project...

[qarium]

libjxl/jpeg-xl is a google project...

they maybe discovered that someone has a patent in this field.

1 single patent discovered in this field and jXL is death.

[Quackdoc]

they maybe discovered that someone has a patent in this field.

1 single patent discovered in this field and jXL is death.

no it's not, that's the point of defensive patents, the truth of the matter is that firefox just doesn't care about what the people want anymore. there is absolutely nothing prohibiting firefox from implementing JXL except for effort, effort which has largely been done by third party contributors.

[qarium]

no it's not, that's the point of defensive patents, the truth of the matter is that firefox just doesn't care about what the people want anymore. there is absolutely nothing prohibiting firefox from implementing JXL except for effort, effort which has largely been done by third party contributors.

you make it look like it is just firefox who does not want to support it
but google chrome removed it as well...

to me it looks like there is a reason why they don't want it.

[Quackdoc]

you make it look like it is just firefox who does not want to support it
but google chrome removed it as well...

to me it looks like there is a reason why they don't want it.

Chrome removed it because of an extremely bias test presented by the AVIF team despite many people asking for it. (it's been alleged that there is over lap between the AVIF/AV1 teams and the chrome team dunno how true it is) The test was extremely poorly done, and even their own data didn't show a massive improvement over JXL.

as for firefox, firefox has no excuse, all they have is "we don't have the resources to implement this" which is complete bogus considering the ammount of funding mozilla gets, they could easily hire more developers or even just review PRs for the feature in nightly.

[qarium]

Chrome removed it because of an extremely bias test presented by the AVIF team despite many people asking for it. (it's been alleged that there is over lap between the AVIF/AV1 teams and the chrome team dunno how true it is) The test was extremely poorly done, and even their own data didn't show a massive improvement over JXL.

i would be interested about this test... do you have any source for this ?
you talked about the progressive decoding feature of jXL and AVIF/AV1 does not support this but has a much more important technology to have big-sized-video file or big-sized-image file on the server and only deliver small-resolution picture or video to the clinds.

so i think google with youtube and their google pixel foto cloud service they are only interested in having a big sized vidoe file or a big sized image file and then only deliver small resolution video or pictures to the clinds.
this saves them a lot of traffic and by this money

but if this combined with a better compression ratio in tests is what they want then why should they support jXL ?

as for firefox, firefox has no excuse, all they have is "we don't have the resources to implement this" which is complete bogus considering the ammount of funding mozilla gets, they could easily hire more developers or even just review PRs for the feature in nightly.

i am pretty sure that firefox did become irrelevant in setting standards. so even if their argument is complete bogus it does not matter means they just do whatever google chrome does. (even if this is stupid)

[Quackdoc]

i would be interested about this test... do you have any source for this ?

This is the testing google used https://storage.googleapis.com/avif-...son/index.html

This was done on a then quite outdated JXL, the metric they choose to highlight is not one that is highly reputable like butteraugli or ssimulacra(2) they had an extremely small sample size (around 70 or so images, Personally I do testing on 20k images at a minimum) they show "encoding MP/s per effort" which is a outright useless metric. They show "quality at a similar effort" which again is useless with how they compared it.

It's just an extremely horribly done test.

you talked about the progressive decoding feature of jXL and AVIF/AV1 does not support this but has a much more important technology to have big-sized-video file or big-sized-image file on the server and only deliver small-resolution picture or video to the clinds.

so i think google with youtube and their google pixel foto cloud service they are only interested in having a big sized vidoe file or a big sized image file and then only deliver small resolution video or pictures to the clinds.
this saves them a lot of traffic and by this money

but if this combined with a better compression ratio in tests is what they want then why should they support jXL ?

AVIF cannot deliver a "smaller file from the same source file" this can actually be verified really easily because a pre-requisite of this is progressive encoding. It is possible to abuse animated AVIF images to get a somewhat kinda progressive file, but you still need to send the full file or you will get a corrupt image because AVIF is based on MP4 in a lot of cases. but in the end, no mater what, if you want a smaller resolution image, you need to encode a second image with AVIF.

JXL on the other hand can actually provide multiple resolutions in a single file, but this is still at the expense of total size. so not many people will actually do this.

i am pretty sure that firefox did become irrelevant in setting standards. so even if their argument is complete bogus it does not matter means they just do whatever google chrome does. (even if this is stupid)

firefox is not irrelevant, they still quite literally have over 170'000'000 users on a monthly basis. That is not irrelevant.

EDIT: fixed formatting