On 2025/5/8 12:19, Eric Biggers wrote: ...
BTW, I also have to wonder why this patchset is proposing accelerating zlib instead of Zstandard. Zstandard is a much more modern algorithm.
I think simply because QAT doesn't support the Zstandard native offload. At least, for Intel Xeon Sapphire Rapids processors (it seems to have built-in QAT 4xxx), only LZ4 and deflate-family are natively supported. I've confirmed that SPR QAT deflate hardware decompresion already surpasses LZ4 software decompression on our cloud server setup, which is useful since it greatly improves decompression performance (even compared to software LZ4) and saves CPU overhead completely. For Zstandard, currently it seems it [1] just leverages part of the QAT engine: hardware sequence producer (matchfinder) to boost up Zstandard match finding And I think the current hardware QAT will have no positive effect on the Zstandard decompression. Without hardware improvements, it seems it can be hardly changed. I think it's true for both 5th Sapphire Rapids processors and 6th Granite Rapids processors. Correct if I'm wrong here but that is what I got so far. Thanks, Gao Xiang [1] https://github.com/intel/QAT-ZSTD-Plugin
- Eric
