Announcement

Collapse
No announcement yet.

Yeppp: A High-Performance Math Library

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

  • Yeppp: A High-Performance Math Library

    Phoronix: Yeppp: A High-Performance Math Library

    Yeppp! is a very fast vector math library that works with a variety of programming languages, designed to work on a vast array of hardware across multiple architectures, and with that is designed to be super fast...

    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
    Can anyone comment from experience how it compares to eigen (http://eigen.tuxfamily.org/index.php?title=Main_Page) in C++? Speed, ease of use, etc.

    Comment


    • #3
      If benchmarks are evidence Eigen can't touch this library.

      Comment


      • #4
        It's open under the BSD license. So expect all faster math routines to be copied into glibc within the week.

        Comment


        • #5
          Is that supposed to be pronounced as https://www.youtube.com/watch?v=K0QHw7iy1Rg ?

          Comment


          • #6
            They don't compare - Eigen is a linear algebra library, Yepp is limited to function like log, sin, cos, exp

            Comment


            • #7
              Originally posted by budric View Post
              Can anyone comment from experience how it compares to eigen (http://eigen.tuxfamily.org/index.php?title=Main_Page) in C++? Speed, ease of use, etc.
              It doesn't, Eigen is for vector and matrix operations, yeppp is for elementary mathematical functions. They complement eachother. After getting my hopes up, I got disappointed. Yeppp lacks the functions I am most interested in, namely pow and sqrt.

              Comment


              • #8
                ARM assembly is completly missing for all but the trivial operations, code seems pretty like what a C compiler would spit out with intrisincs + some cleanups.
                The speed comes mainly from excessive loop unrolling, so unless you do ONE of the supported operation on a HUGE array you are better of with a slower but single pass.

                I dont still dont know the a good generic enough solution, intrisincs are bearable but rather specific to architectures, asm is still the best for performance but even more HW-Specific. Libraries like this have a very narrow usecase

                Comment


                • #9
                  Originally posted by curaga View Post
                  It's open under the BSD license. So expect all faster math routines to be copied into glibc within the week.
                  If you're crying so much maybe you should permit developers to release their code under the bsd license? If it will be included into glibc, yeppp developers should be very happy. Unless they were drank and they weren't aware what bsd license exist for.

                  Comment


                  • #10
                    There are vector functions, but no matrix functions as of now.

                    Comment

                    Working...
                    X