Announcement

Collapse
No announcement yet.

Benchmarking AMD's New AOCC Compiler For Ryzen

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

  • #21
    Originally posted by reavertm View Post
    Would be interesting to see ICC in here. (and funny to see and quite possibly outperforming AOCC even at AMD hardware).
    Well, to test icc, you need to test both baseline performance, and performance after removing the Intel specific runtime checks that reduce it to worst-case scenario on AMD. I am not even sure anyone still maintains uncrippling tools for icc. I haven't tried using it since they stopped providing it for free. And back then it was an arms race, we kept finding ways to make it not suck on AMD chips and they kept breaking them.

    Comment


    • #22
      Originally posted by carewolf View Post

      Well, to test icc, you need to test both baseline performance, and performance after removing the Intel specific runtime checks that reduce it to worst-case scenario on AMD. I am not even sure anyone still maintains uncrippling tools for icc. I haven't tried using it since they stopped providing it for free. And back then it was an arms race, we kept finding ways to make it not suck on AMD chips and they kept breaking them.
      Now, that's interesting. Got any sources?

      Comment


      • #23
        Originally posted by hugo8621 View Post

        sounds interesting with the -mtune=haswell. Is there some data that backs that up? Did phoronix run some benchmark comparisons on that? Might give that a try myself
        The scimark2 MonteCarlo benchmark with gcc-7.1 on 1800x jumps from 411 (znver1) to 1745 (haswell). The rest pretty much stay the same.
        Adding "-mprefer-avx128" raises LU benchmark score from 5000 to 6000.

        Comment


        • #24
          Originally posted by wdb974 View Post

          Now, that's interesting. Got any sources?
          http://www.agner.org/optimize/blog/read.php?i=49#49

          It is a bigger problem on Windows where icc is used by many games and benchmarks.

          Comment


          • #25
            I'm not agree with conclusions of test. This is the FIRST version of this tool and it's very good against compilers that are around from years. A good beginning

            Comment


            • #26
              Originally posted by tildearrow View Post
              They should open this project
              to open it they need to close it first

              Comment


              • #27
                Originally posted by chuckula View Post
                It's not much of a surprise, RyZen performs about as well as it's going to in Linux anyway and the lack of real number crunching hardware (only 1/2 speed support for AVX and no AVX-512 at all) can't just be magically optimized away in a compiler.
                to not much of a surprise, you posted irrelevant bullshit. this was compiler benchmark, not cpu benchmark. and if you need avx, you are better off with videocard
                Last edited by pal666; 20 May 2017, 03:03 PM.

                Comment


                • #28
                  Was openssl configured with "no-asm" option?
                  Was ffmpeg configured with "--disable-asm" option?

                  Comment


                  • #29
                    Originally posted by puleglot View Post
                    Was openssl configured with "no-asm" option?
                    Was ffmpeg configured with "--disable-asm" option?
                    All the build details are available via looking at the test profile scripts if running PTS locally or navigating to the test profile pages on Openbenchmarking.org.
                    Michael Larabel
                    http://www.michaellarabel.com/

                    Comment


                    • #30
                      Originally posted by Michael View Post

                      All the build details are available via looking at the test profile scripts if running PTS locally or navigating to the test profile pages on Openbenchmarking.org.
                      Indeed. Thanks!

                      Code:
                      cd ffmpeg-2.8.1/
                      ./configure --disable-zlib --disable-doc --prefix=$HOME/ffmpeg_/
                      Code:
                      cd openssl-1.0.1g/
                      ./config no-zlib

                      Comment

                      Working...
                      X