Announcement

Collapse
No announcement yet.

Zlib-ng 2.2 Speeds Up Compression By ~12% On x86_64 CPUs

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • Zlib-ng 2.2 Speeds Up Compression By ~12% On x86_64 CPUs

    Phoronix: Zlib-ng 2.2 Speeds Up Compression By ~12% On x86_64 CPUs

    The first release candidate of Zlib-ng 2.2 for this drop-in replacement to the Zlib data compression library is now available for testing. Zlib-ng continues to ship new performance optimizations and other tuning for providing faster Zlib performance on modern processors...

    Phoronix, Linux Hardware Reviews, Linux hardware benchmarks, Linux server benchmarks, Linux benchmarking, Desktop Linux, Linux performance, Open Source graphics, Linux How To, Ubuntu benchmarks, Ubuntu hardware, Phoronix Test Suite

  • #2
    Fedora ships with zlib-ng by default
    It's opt-in for Gentoo

    Comment


    • #3
      Also, it uses CMake instead of autotools. That's reasonable

      Last edited by Hazel; 19 June 2024, 11:09 AM.

      Comment


      • #4
        Originally posted by Hazel View Post
        Also, it uses CMake instead of autotools. That's reasonable

        https://github.com/zlib-ng/zlib-ng
        In fact, the original zlib also has a CMake build system.

        Comment


        • #5
          Originally posted by Hazel View Post
          Also, it uses CMake instead of autotools. That's reasonable

          https://github.com/zlib-ng/zlib-ng
          It would be hilarious to use autotools on anything-ng.

          Comment


          • #6
            Originally posted by Hazel View Post
            Also, it uses CMake instead of autotools. That's reasonable

            https://github.com/zlib-ng/zlib-ng
            But is it written in Rust?

            If it's not written in Rust then it's not kewl, right?

            Comment


            • #7
              Originally posted by NotMine999 View Post

              But is it written in Rust?

              If it's not written in Rust then it's not kewl, right?
              Rust shill here: The 5-10% speed edge C has is worth it for this

              Comment


              • #8
                C doesn't have a meaningful "speed edge". Sometimes Rust is faster, sometimes C is. In any case the reason for zlib-ng's speed is down to using CPU intrinsics and not anything to do with the language.

                The reason zlib-ng is written in C, is because zlib was written in C, and zlib-ng is a fork - not for any other reason.
                Last edited by dralley; 19 June 2024, 08:51 PM.

                Comment


                • #9
                  Originally posted by dralley View Post
                  C doesn't have a meaningful "speed edge". Sometimes Rust is faster, sometimes C is. In any case the reason for zlib-ng's speed is down to using CPU intrinsics and not anything to do with the language.

                  The reason zlib-ng is written in C, is because zlib was written in C, and zlib-ng is a fork - not for any other reason.
                  True. What I meant (and what I hope others mean) by "speed edge" is that C gives you much finer control to do complicated optimizations than Rust. Though I guess Rust does also have ASM macros which would give you the same control...
                  Naive (or the equivalent "pythonic" terms for each language) will definitely vary based on what optimizations the languages can do but as I understand it C takes less effort to do the exact assembly you want

                  Comment


                  • #10
                    What I meant (and what I hope others mean) by "speed edge" is that C gives you much finer control to do complicated optimizations than Rust.
                    It doesn't, though. Anything you could do with C can be done through unsafe Rust, including ASM and SIMD. C may make it very slightly easier to do those things as a result of not needing as much ceremony, but it's kinda counterbalanced by the 90% of pedestrian code which Rust makes far easier.

                    Comment

                    Working...
                    X