Announcement

Collapse
No announcement yet.

FLANG Fortran Compiler Preparing To Land In The LLVM Source Tree

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

  • FLANG Fortran Compiler Preparing To Land In The LLVM Source Tree

    Phoronix: FLANG Fortran Compiler Preparing To Land In The LLVM Source Tree

    It looks like the FLANG Fortran compiler (also previously referred to as f18) could soon be landing in the LLVM source tree and in time for the LLVM 10.0 branching...

    http://www.phoronix.com/scan.php?pag...-FLANG-Landing

  • #2
    Benchmarks ?

    Comment


    • #3
      Originally posted by pyler View Post
      Benchmarks ?
      Put it on my TODO list earlier today when seeing it was close to merging, might have out some initial data by next week.
      Michael Larabel
      http://www.michaellarabel.com/

      Comment


      • #4
        Would be good to see if MSVC can actually build the source tree. I was keeping an eye out for Flang as a means to compile multiple projects with Fortran components outside compat layers like MinGW or Cygwin. MSVC just simplifies bootstrapping the process, ultimately if it builds with Clang / libstdc++ that's enough.

        Comment


        • #5
          I thought FORTRAN support is only needed to support legacy apps but judging by the support it gets it seems to be thriving, so is it better than C/C++/Java/C#/whatever at something other than supporting legacy FORTRAN code?

          Comment


          • #6
            Originally posted by cl333r View Post
            I thought FORTRAN support is only needed to support legacy apps but judging by the support it gets it seems to be thriving, so is it better than C/C++/Java/C#/whatever at something other than supporting legacy FORTRAN code?
            New code is being written in FORTRAN, but less and less. However that's irrelevant. FORTRAN is used in simulations on literally all of the top super computers. Some of those codebase have been evolving for half a century and are not going away, ever. Many of those projects are funded by the military and/or UN/EU and/or universities backed with serious funding. So tl;dr;, they have unlimited (other people) money and making new compilers is worthwhile. *Someone* will pay for it just because it can be done and billed.

            edit: Legend has it that it's faster than C because of aliasing. However, that has not been true since the 80's.

            Comment


            • #7
              Originally posted by Elv13 View Post

              New code is being written in FORTRAN, but less and less. However that's irrelevant. FORTRAN is used in simulations on literally all of the top super computers. Some of those codebase have been evolving for half a century and are not going away, ever. Many of those projects are funded by the military and/or UN/EU and/or universities backed with serious funding. So tl;dr;, they have unlimited (other people) money and making new compilers is worthwhile. *Someone* will pay for it just because it can be done and billed.

              edit: Legend has it that it's faster than C because of aliasing. However, that has not been true since the 80's.
              Fortran has better precision for Numerical Analysis and more than C.

              Comment


              • #8
                Originally posted by Marc Driftmeyer View Post

                Fortran has better precision for Numerical Analysis and more than C.
                Care to expand on this? I don't see any way in which this can be true, precision is defined by the number of bits used for representation and the operations performed with them (and their order).

                Originally posted by Elv13 View Post
                edit: Legend has it that it's faster than C because of aliasing. However, that has not been true since the 80's.
                Aliasing can be prevented in C, but is definitely cumbersome. I have read that it is easier for the compiler to produce proper SIMD assembly in Fortran than in C. Inlining it is easier to decide to.

                Another advantage of Fortran (in lowercase for the last 30 years) is that math is legible and easy to write.

                It all comes down to what you are doing. If you are writing 95% math, 5% general purpose programming, you will be okay with Fortran (and perhaps some bindings to C/C++ libs for GP stuff).
                If you are writing 95% general purpose and 5% math, you will be better with C/C++ and eventually if you have a very strong motive you can bind to Fortran or inline assembly yourself.

                Comment


                • #9
                  This is good news as it fleshes out the LLVM project.

                  The use of FLang has me hoping the trend continues into the future. SLang would be good for Swift once it migrated back into mainstream LLVM. RLang would work for Rust. PLang for a pascal if somebody ever does that. The more that I think about it the more I like.

                  Comment


                  • #10
                    Originally posted by Marc Driftmeyer View Post

                    Fortran has better precision for Numerical Analysis and more than C.
                    Funny. As the Apple fanboy that you normally are, I thought you would drop a comment about great Swift is. Refreshing to see a non-Apple comment from you

                    Comment

                    Working...
                    X