Announcement

Collapse
No announcement yet.

Clasp 0.4 -- Lisp Over LLVM -- Generates Code 200x Faster

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

  • Clasp 0.4 -- Lisp Over LLVM -- Generates Code 200x Faster

    Phoronix: Clasp 0.4 -- Lisp Over LLVM -- Generates Code 200x Faster

    Clasp is a Common Lisp compiler based on LLVM that also provies seamless interoperation with C++ libraries. Clasp 0.4 has been released with some big improvements...

    http://www.phoronix.com/scan.php?pag...p-0.4-Released

  • #2
    Clasp is one of the most refreshing things that happened to Common Lisp in 2015. The developer is quite responsive too.

    Comment


    • #3
      And 4 times slower than C++. Being able to use C++ libs is pretty neat though. Will be interesting to see how far they can get it performance-wise.

      Comment


      • #4
        AFAIK, Lisp is still the "fastest language" that is neither C/C++ nor Assembler.

        Comment


        • #5
          Originally posted by Cthulhux View Post
          AFAIK, Lisp is still the "fastest language" that is neither C/C++ nor Assembler.
          What about other compiled non-vm languages like Fortran, rust and go? The first two don't even use a garb collector?

          Comment


          • #6
            I don't know of a "Common Lisp vs. Go" benchmark. Mind to make one?

            Comment


            • #7
              Originally posted by Cthulhux View Post
              AFAIK, Lisp is still the "fastest language" that is neither C/C++ nor Assembler.
              I don't believe so. http://benchmarksgame.alioth.debian.org/ click on the languages and then click on the language name on the page (taking you to http://benchmarksgame.alioth.debian.....php?lang=____ where '____' can be rust, lisp, java, haskell, etc...)

              Their Lisp benchmarks use SBCL, one of the faster (fastest?) free software Lisp implementations. It still loses in some benchmarks to Ada, Rust, Haskell, Java, Fortran, Go, and probably a few others.

              In my experience, outside of seriously resource-constrained environments it makes sense to develop in whatever language lets you write high quality code the fastest. Then profile, and when you find slow spots replace poor algorithms with better ones or even foreign function calls to fast libraries in other languages. Python is routinely used in scientific computing and data analysis despite being notoriously slow - because Python developers are very willing to dip into C and C++ libraries when they need them.

              Comment


              • #8
                Actually, Clozure CL is faster than SBCL in quite a lot of cases.

                Racket - a Lisp - leads the way.

                Comment


                • #9
                  Originally posted by Cthulhux View Post
                  Actually, Clozure CL is faster than SBCL in quite a lot of cases.

                  Racket - a Lisp - leads the way.
                  Racket does well versus many other languages, but it trails SBCL: https://benchmarksgame.alioth.debian...4q/racket.html

                  Thanks for the tip on Clozure CL. I couldn't find any recent benchmarks. 4 years ago SBCL was pretty consistently quicker: http://www.cliki.net/performance%20benchmarks

                  Comment


                  • #10
                    It depends on your code, really.

                    Comment

                    Working...
                    X