Announcement

Collapse
No announcement yet.

Pyston 0.3 Released For High-Performance Python

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

  • Pyston 0.3 Released For High-Performance Python

    Phoronix: Pyston 0.3 Released For High-Performance Python

    Version 0.3 of the high-performance Python programming implementation is now available...

    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
    1% faster than CPython
    not to sound rude, but wasn't Pyston announced about a year ago?
    That was the best they could do in a year?

    Comment


    • #3
      Originally posted by peppercats View Post
      not to sound rude, but wasn't Pyston announced about a year ago?
      That was the best they could do in a year?
      Did you read the relevant section of the blog/announcement?

      Comment


      • #4
        ll

        Originally posted by DanL View Post
        Did you read the relevant section of the blog/announcement?
        No because I have no interest in using software from a company that would hire people like condoleezza rice

        Comment


        • #5
          Originally posted by peppercats View Post
          not to sound rude, but wasn't Pyston announced about a year ago?
          That was the best they could do in a year?
          If you're interested in python performance, you can give a shot at pypy. It can be way faster than CPython on your code.
          Last edited by VinceLe; 22 April 2019, 02:13 PM. Reason: typo

          Comment


          • #6
            Originally posted by peppercats View Post
            not to sound rude, but wasn't Pyston announced about a year ago?
            That was the best they could do in a year?
            Someone who has Wordpress account please tell them to benchmark against PyPy. CPython is the slow reference implementation, it's not that impressive to be faster than it

            Comment


            • #7
              Does anyone remember Unladen Swallow? Makes me wonder if the Pyston guys have the endurance and resources to get LLVM jitter up to speed.

              Comment


              • #8
                Originally posted by log0 View Post
                Does anyone remember Unladen Swallow? Makes me wonder if the Pyston guys have the endurance and resources to get LLVM jitter up to speed.
                Personally I suspect their time would be better spent contributing to PyPy than developing another JIT'ing Python

                Comment


                • #9
                  Originally posted by nanonyme View Post
                  Personally I suspect their time would be better spent contributing to PyPy than developing another JIT'ing Python
                  This has always been the standard way the python ecosystem progresses. Several different groups try different approaches to the same problem. This allows people to see what works and what doesn't. After some time, either one project comes out as clearly superior and the other projects are abandoned, or a new project is created combining the best bits of the other projects, or the projects move apart so they either cover substantially different use-cases or different levels of a software stack. This has happened over and over in various parts of the python ecosystem (numerical computing, html parsing, web frameworks, plotting, installer creation), and is happening right now in other areas (faster interpreter, high-level plotting, next-gen array interface, unit testing).

                  Pypy is a great project, but it may or may not be the best approach to the problem. Some approaches may ultimately run into fundamental limitations that restrict their performance compared to other approaches, or it may be much harder to get comprable performance. That is why it is good to try a few different things to see what ultimately turns out to be superior.

                  Comment


                  • #10
                    This is somewhat pointless

                    Ok, let's assume they get pure python working.

                    Now what about the countless CPython extensions written in C? Numpy? Others? Will they be compatible? How long will it take to make them compatible?

                    And CPython is ridiculously slow. As in ~60x slower than C in some cases. They should benchmark against PyPy.

                    These resources should be spent working on PyPy IMO.

                    Or if they want to do something novel with python- contribute to Jython and make Python work well with the JVM and use InvokeDynamic and other modern JVM features. This is another way to sqeeze out performance on a mature JIT'ed VM. Current Jython is ridiculously slow, but they haven't spend much time on improving speed or using new JVM features for dynamic languages.

                    Comment

                    Working...
                    X