AMD Releases Brotli-G For GPU-Accelerated Brotli Compression

[phoronix]

Phoronix: AMD Releases Brotli-G For GPU-Accelerated Brotli Compression

After open-sourcing its Radeon Raytracing Analyzer code last week, this week AMD's GPUOpen team has a new open-source project announcement: Brotli-G...

https://www.phoronix.com/news/AMD-GPUOpen-Brotli-G

[tildearrow]

Typo:

AMD is releasing te Brotli-G SDK under an MIT license.

[Danny3]

Wonderful, more open source stuff is always better to have!
Good job and thank you very much AMD!

[Michael]

Typo:

Thanks

[alpha_one_x86]

To see how it scale, any benchmark to difference of performance?

[FireBurn]

I wonder if popping it into a HLSL to SPIRV converter works

[jochendemuth]

just not AMD Radeon graphics products

Shouldn't it be "not just"? In my mind this has the completely opposite meaning. I got really confused reading this sentence.

[jochendemuth]

To see how it scale, any benchmark to difference of performance?

Yes, weird to announce a new algo, hinting at improved performance, but not releasing any numbers.

I had to dig through the articles, but it linked to https://quixdb.github.io/squash-benc...#results-table.
This does not contain Brotli-G numbers, but is a very nice and comprehensive comparison of compression algos.

Better than nothing, I guess.

[Meteorhead]

I wonder if popping it into a HLSL to SPIRV converter works

That converter is called Clang. Just fetch the DX compiler and compile it. Or wait until upstreaming of the HLSL front-end is finished (it's pending) and upstream Clang will do the trick, won't need MS's fork anymore.

[alpha_one_x86]

Yes, weird to announce a new algo, hinting at improved performance, but not releasing any numbers.

I had to dig through the articles, but it linked to https://quixdb.github.io/squash-benc...#results-table.
This does not contain Brotli-G numbers, but is a very nice and comprehensive comparison of compression algos.

Better than nothing, I guess.

The problem is some algorithm don't scale well, then use GPU to just have 30% more performance is useless