Announcement

Collapse
No announcement yet.

NumPy 2.0 Brings Faster Performance Thanks To Intel's x86-simd-sort & Google's Highway

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

  • NumPy 2.0 Brings Faster Performance Thanks To Intel's x86-simd-sort & Google's Highway

    Phoronix: NumPy 2.0 Brings Faster Performance Thanks To Intel's x86-simd-sort & Google's Highway

    NumPy 2.0 was released on Sunday that's been in the making for the past year and their first major release since 2006. While it comes with API/ABI breakage, NumPy 2.0 delivers new features and performance improvements...

    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
    I started with NumPy 0.9 on Python 1.6 some twenty-something years ago.

    And now I think this Python craze went too far. There are much better, safer and faster languages out there.
    Python is dangerous, slow and overrated.

    We'd all be better off if people actually learned proper programming rather than gluing bits and pieces of Python code from Stack Overflow (or Copilot).

    Comment


    • #3
      Originally posted by pkese View Post
      There are much better, safer and faster languages out there.
      Python is dangerous, slow and overrated.
      We'd all be better off if people actually learned proper programming rather than gluing bits and pieces of Python code from Stack Overflow (or Copilot).
      Is there any language/tool where this don't apply? It's absolute generic (and wrong) statement.

      Comment


      • #4
        Originally posted by pkese View Post
        And now I think this Python craze went too far. There are much better, safer and faster languages out there.
        Python is dangerous, slow and overrated.
        Using Python for something like deep learning isn't necessarily slow. Basically, it gives you a scriptable shell atop a bunch of C/C++ libraries that do the real work. Matlab does this a lot, with toolboxes like Computer Vision being fairly direct wrappers of OpenCV (from what I recall).

        Not saying I love everything about it, but I think Python's popularity has something to do with being both approachable enough for education and capable enough for real work.


        BTW, thanks to the article for drawing my attention to Google Highway. I'll have to take a good look at it, next time I need something like it.
        Last edited by coder; 17 June 2024, 08:23 AM.

        Comment


        • #5
          Originally posted by Ilfirin View Post
          Is there any language/tool where this don't apply? It's absolute generic (and wrong) statement.
          In these forums, the answer is always Rust.


          Originally posted by pkese View Post
          And now I think this Python craze went too far. There are much better, safer and faster languages out there.
          Python is dangerous, slow and overrated.

          We'd all be better off if people actually learned proper programming rather than gluing bits and pieces of Python code from Stack Overflow (or Copilot).
          Seems like everyone who says stuff like this doesn't actually understand the point of Python.
          Everyone fluent in Python knows it has issues. If those issues were relevant to it's usefulness, they obviously they wouldn't use it.

          Comment


          • #6
            Originally posted by schmidtbag View Post
            In these forums, the answer is always Rust.
            I am Rust "fan" and it applies too. Every month there are new syntax addition to existing syntax, that even official documentation states "this could be written by this, this, this, or this". It's becoming (sadly) pretty bloated. Time for new modern language with clear design, again.

            Comment


            • #7
              Originally posted by Ilfirin View Post
              I am Rust "fan" and it applies too. Every month there are new syntax addition to existing syntax, that even official documentation states "this could be written by this, this, this, or this". It's becoming (sadly) pretty bloated. Time for new modern language with clear design, again.
              Or a new core specification that's easier to remember. Just mark some of these methods the proper way, like the way Python has "pythonic." Just to point out, I have never used Rust but I here there are so many concepts you have to learn and more are introduced all the time.

              Comment


              • #8
                Seems like everyone who says stuff like this doesn't actually understand the point of Python.
                I agree. We use Python 3 a lot when it comes to automating the boring stuff. It works really well in that role and even small GUI applications. It is NOT used where Rust or C/C++ needs to be used. Use the language appropriate for what you are doing. Python turns out to work for 80%+ of the work we do.... and does is very WELL.

                Nice to see improvements in numpy.

                Comment


                • #9
                  Originally posted by pkese View Post
                  We'd all be better off if people actually learned proper programming rather than gluing bits and pieces of Python code from Stack Overflow (or Copilot).
                  We would be better off, but there would be fewer of us.

                  NumPy is immensely popular in the scientific and research community exactly because it's based on Python, which has a very gentle cost of entry for people who are not primarily programmers. It's either that or use R or non-programming tools. Python, with its ecosystem + NumPy has enabled a revolution in the quality of programs made in the science community. And for some folk it's even a stepping stone towards using more advanced languages.

                  Offloading the computation work to optimized lower-level code, including machine code, is a reasonable balance between Python's usability and best performance.

                  Comment


                  • #10
                    Python can be slow enough to cause problems feeding ML frameworks with data. Among other things because Python has no threads that would use CPU time from multiple cores.

                    This is mentioned in the intro to the C interface to Tensorflow.

                    Comment

                    Working...
                    X