S3TC Support Will Land In Mesa Now That The Patent Has Expired

[rene]

another great example how patents foster innovation, not!

[bug77]

another great example how patents foster innovation, not!

How did the S3TC patent prevent other from coming up with better ideas?
Software patents aren't a big help in their current form, but in this case the only thing they stopped was implementation in an open source library because there was no one to foot the bill.

[oiaohm]

Huh? S3TC is 16 RGBA texels in 64-bit (DXT1) or 128-bit (DXT3/5). I also think there's a big "*" on the alpha in the 64-bit case in terms of ability to represent various values.

ASTC can go between 16 (ASTC_4x4) to 144 (ASTC_12x12) RGBA texels in 128-bit of data. Of course the quality stinks with 12x12, you get what you paid for at that compression level. Both are lossy.

Should have been a little more exact. https://www.youtube.com/watch?v=7bJ-D1xXEeg

Crunch implementation of S3TC takes you to 4x4 to 16x16 in variable. Please note Crunch need the S3TC patents most likely will have to be implemented in glsl/SPIR-V to use GPU the most as most hardware will not have decayed decoders for it all.. So S3TC based codecs can exceed the current ASTC codecs. We have been stuck in time with S3TC due to the patent and ASTC does not implement particular features to avoid the S3TC patents.

Basically I left out a word. Crunch S3TC beats ASTC. So the effects of this S3TC patents expiring are going to have many years of effects as research done on S3TC will be able to come back into mainline developments. Basically DXT1/3/5 is only part of the picture of S3TC effected codecs.

So the next 12 months could be a little bumpy for everyone in the graphical world and even more arguments over what texture encoding should be supported by hardware now that patents are out the way on S3TC based codecs.

[oiaohm]

How did the S3TC patent prevent other from coming up with better ideas?
Software patents aren't a big help in their current form, but in this case the only thing they stopped was implementation in an open source library because there was no one to foot the bill.

There are many improved versions to S3TC that were done in research that have never made it into active deployment due to the fact you need a patent license and the terms of the patent license forbid modified forms. This includes not needing hardware support for S3TC because using shading language to decode S3TC and once you start doing this you are no longer locked to what is in hardware of GPU cards. Please note it did not stop open source library existing but it prevented the open source library from being commercial used and deployed it also prevented commercial games and the like having software fall backs. Licensing a game under S3 license set-up would have be totally horrible..

This is the old unable to do "Standing on the shoulders of giants" problem. When you licensed S3TC you agreed to implement only as S3 defined so any modification to S3TC that would improve things like what crunch did was forbidden to hardware makers. Yes we see ASTC implemented bits that could make S3TC better and S3TC had features that could have made ASTC better so neither codec is the best possible and that was partly caused by the patent license terms.

Really limitation clauses in patent licenses cause a lot of poorer quality options to exist and in a lot of cases the best option not to exist at all because the patent licenses will not allow it to exist.

[andre30correia]

since steam need s3tc since ever, everybody have this install, at least who use steam to play, but finally the patent is over, maybe this way it will improve

[Nille_kungen]

Does this affect all GPUs in mesa or just Intel? Either way, I'm excited to see test results of this. Makes me wonder if this will eliminate the performance gap in some tests vs Windows (note I said some).

It's all GPUs.
This isn't performance but rather fixing broken textures you might have had if you didn't install libtxc_dxtn.
It's S3 Texture Compression so it's all about textures the black things you might have seen in some games where it should have been something else.

[bug77]

There are many improved versions to S3TC that were done in research that have never made it into active deployment due to the fact you need a patent license and the terms of the patent license forbid modified forms. This includes not needing hardware support for S3TC because using shading language to decode S3TC and once you start doing this you are no longer locked to what is in hardware of GPU cards. Please note it did not stop open source library existing but it prevented the open source library from being commercial used and deployed it also prevented commercial games and the like having software fall backs. Licensing a game under S3 license set-up would have be totally horrible..

This is the old unable to do "Standing on the shoulders of giants" problem. When you licensed S3TC you agreed to implement only as S3 defined so any modification to S3TC that would improve things like what crunch did was forbidden to hardware makers. Yes we see ASTC implemented bits that could make S3TC better and S3TC had features that could have made ASTC better so neither codec is the best possible and that was partly caused by the patent license terms.

Really limitation clauses in patent licenses cause a lot of poorer quality options to exist and in a lot of cases the best option not to exist at all because the patent licenses will not allow it to exist.

I'm confused. All the better compression algorithms out there are based on S3TC? Because if they aren't, they could be implemented regardless of S3TC being patented. And if they aren't, then maybe S3TC wasn't so trivial after all.

And no, I'm siding with patent holders, I still think 20 years protection for a bouncing animation or even a rectangle(ish) shape (a non-software patent) is idiotic.

[oiaohm]

I'm confused. All the better compression algorithms out there are based on S3TC? Because if they aren't, they could be implemented regardless of S3TC being patented. And if they aren't, then maybe S3TC wasn't so trivial after all.

The patented parts of S3TC are regarding colours. How you pick your colours and store your colours effect how your bias your lossy compression. So yes you might have better compression like ASTC has but then you cannot optimise it due to not being able to go anywhere near the S3TC patents. Basically Crunch S3TC takes ideas from ASTC and S3TC and uses them with each other.

S3TC was kind of trivial but trivial differences can make a difference between an image that looks good to the human eye or looks complete crap to the human eye.

Also knowing if texture compression is good bad or indifferent you want to be comparing the methods using the last ideas. ASTC look good compared to the original S3TC but not as good against crunch what is more modern. Of course if s3 had patented S3TC more openly we may have see more work in this field and another method better than S3TC may have tuned up.

You need competition between the different ideas to work in the direction of the best solution. The S3TC patents was prevent people working with the S3TC ideas to make them as good as they could be. So now when ASTC and others have been comparing it been as long as it better than the old S3TC it was good enough. Now this means ASTC developers never got pushed to be the best possible either.

So even compression solutions not based off S3TC due to S3TC patented status and no one able put improved S3TC versions out means those solutions are not as good as they could be because they never had proper competition.

You see the same thing with font hinting patents around freetype. Yes freetype was able to make kind of approx equals but they were not 100 percent right because they could not touch the patented items. Now that a lot of patents are gone from font hinting comparing the different design choices is more possible so freetype project is able to do more work in this field and have more competition between engines. This has seen engines in freetype that never depended on the patents even be improved.

Patents can restrict competition between solutions and in fact harm development in those areas. Patents that you can license and use however you want are a lot less harmful than patents that you license that say you must implement X way or no patent license. S3TC deployed version has almost been 100 percent suspended in time for 20 years so not providing effective competition to other solutions. Now that the patent is expired we hopefully will see competition in texture compression heat up.

[leonmaxx]

I used Crunch library for DXT (S3TC) compression for about a year and then switched to codecs from AMD Compress (old Compressonator) when it was open sourced.
AMD's codecs worked better for our tasks they gave compressed textures with good quality in much less time than Crunch.

Crunch S3TC

There is no such format. May be you meant CRN format?
Anyway it has NO SUPPORT in hardware and have to be transcoded to DXT1/DXT5 formats to be usable by hardware.

Crunch implementation of S3TC takes you to 4x4 to 16x16 in variable.

Not correct. From documentation page on CRN and clustered DDS (https://github.com/BinomialLLC/crunch):

any combination of 4x4, 8x4, 4x8, and 8x8 pixel blocks

Crunch is discontinued now. Author now working on another commercial texture compression library/tool called Basis.

[Guest]

It's all GPUs.
This isn't performance but rather fixing broken textures you might have had if you didn't install libtxc_dxtn.
It's S3 Texture Compression so it's all about textures the black things you might have seen in some games where it should have been something else.

To add onto this; I remember a while back I tried playing WoW with Wine; without libtxc_dxtn, certain textures would look washed-out and in a lower color bit.