<div dir="ltr"><div>On Fri, Feb 8, 2013 at 1:28 PM, Simon Marlow <span dir="ltr">&lt;<a href="mailto:marlowsd@gmail.com" target="_blank">marlowsd@gmail.com</a>&gt;</span> wrote:<br></div><div class="gmail_extra"><div class="gmail_quote">

<blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;padding-left:1ex;border-left-color:rgb(204,204,204);border-left-width:1px;border-left-style:solid"><div class="im">On 08/02/13 10:10, Nicolas Frisby wrote:<br>


<blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;padding-left:1ex;border-left-color:rgb(204,204,204);border-left-width:1px;border-left-style:solid">
Hi Johan. I&#39;m seeing much different numbers when I test for regressions.<br>
Also 64-bit Linux.<br>
Could you explain your method? Origin of the GHC/libraries you&#39;re<br>
measuring and the options your running NoFib with? Thanks.<br>
I&#39;m using the 28 Jan ghc-7.6.2-x86_64-unknown-<u></u>linux.tar.bz2 and<br>
ec9377b1e51fd59e18 (Wed Feb 6 14:10:39 2013 +0000) for HEAD.<br>
Thanks.<br>
</blockquote>
<br></div>
The usual source of differences is -O vs. -O2, either for the benchmarks or for the libraries: you want GhcLibOpts=-O2 and NoFibHcOpts=-O2 in your mk/<a href="http://build.mk" target="_blank">build.mk</a> file.<br>
<br>
Cheers,<br>
        Simon<br>
<br></blockquote><div><div> </div><div>I figured Johan is doing O2 for libraries and programs. I have done libraries at O2, programs at O1, which likely does explain the difference.</div><div> </div><div>I&#39;m still in a holding pattern in case there&#39;s also some other setting he&#39;s using.</div>

<div><div> </div><div>Then I&#39;ll run both variations on the programs.</div><div> </div><div>Thanks much.</div></div></div><div> </div><div> </div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;padding-left:1ex;border-left-color:rgb(204,204,204);border-left-width:1px;border-left-style:solid">


<br>
<br>
<blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;padding-left:1ex;border-left-color:rgb(204,204,204);border-left-width:1px;border-left-style:solid"><div class="im">
<br>
On Thu, Feb 7, 2013 at 9:18 PM, Johan Tibell &lt;<a href="mailto:johan.tibell@gmail.com" target="_blank">johan.tibell@gmail.com</a><br></div><div><div class="h5">
&lt;mailto:<a href="mailto:johan.tibell@gmail.com" target="_blank">johan.tibell@gmail.com</a><u></u>&gt;&gt; wrote:<br>
<br>
    Hi all,<br>
<br>
    I just ran nofib on current HEAD and compared it to 7.6.2 on my<br>
    64-bit Linux machine. There are some regressions I think we should<br>
    look into before a release:<br>
<br>
    ------------------------------<u></u>------------------------------<u></u>--------------------<br>
             Program           Size    Allocs   Runtime   Elapsed  TotalMem<br>
    ------------------------------<u></u>------------------------------<u></u>--------------------<br>
                anna          -0.8%     +1.5%      0.08      0.08     +0.0%<br>
                ansi          -0.5%     +0.5%      0.00      0.00     +0.0%<br>
                atom          -0.9%     -0.0%     -3.1%     -0.8%     +0.0%<br>
              awards          -0.9%     +0.1%      0.00      0.00     +0.0%<br>
              banner          +0.1%     +2.2%      0.00      0.00     +0.0%<br>
          bernouilli          -0.6%     +2.7%      0.12      0.13     +0.0%<br>
               boyer          -0.7%     +0.0%      0.03      0.04     +0.0%<br>
              boyer2          -0.1%     +0.0%      0.01      0.01     +0.0%<br>
                bspt          -0.6%     -0.0%      0.01      0.02     +0.0%<br>
           cacheprof          -0.9%     +6.9%     +0.8%     +4.0%     +0.0%<br>
            calendar          -0.9%     +0.1%      0.00      0.00     +0.0%<br>
            cichelli          -0.1%     -0.0%      0.06      0.06     +9.4%<br>
             circsim          -0.7%     +0.1%     +1.6%     +7.0%    +13.6%<br>
            clausify          -0.8%     +0.0%      0.03      0.03     +0.0%<br>
       comp_lab_zift          -0.7%     +0.0%      0.14      0.14    -12.5%<br>
            compress          -0.1%     +0.0%      0.12      0.13     +0.0%<br>
           compress2          -0.0%     +0.0%      0.12      0.14     +2.3%<br>
         constraints          -0.8%     +0.0%     -3.3%     -0.7%     +0.0%<br>
        cryptarithm1          -0.1%     +0.0%     +2.1%     +4.1%     +0.0%<br>
        cryptarithm2          -0.0%     -0.8%      0.01      0.01     +0.0%<br>
                 cse          -0.0%     -0.0%      0.00      0.00     +0.0%<br>
               eliza          +0.1%     +7.7%      0.00      0.00     +0.0%<br>
               event          -0.8%     +0.0%      0.09      0.10     -8.7%<br>
              exp3_8          -0.8%     +0.0%      0.15      0.15   +100.0%<br>
              expert          -0.2%    +18.4%      0.00      0.00     +0.0%<br>
                 fem          +1.5%     +0.5%      0.02      0.02     +0.0%<br>
                 fft          -0.9%     +0.0%      0.02      0.03     +0.0%<br>
                fft2          +3.7%    +31.9%      0.05      0.05    +20.0%<br>
            fibheaps          -1.0%     +0.3%      0.03      0.03     +0.0%<br>
                fish          -0.0%     -0.0%      0.01      0.01     +0.0%<br>
               fluid          -1.3%    +13.0%      0.01      0.01     +0.0%<br>
              fulsom          -0.2%     -0.0%      0.19      0.20     +8.3%<br>
              gamteb          -0.7%     -0.2%      0.03      0.04     +0.0%<br>
                 gcd          -0.8%     +0.0%      0.02      0.03     +0.0%<br>
         gen_regexps          +0.0%     +2.2%      0.00      0.00     +0.0%<br>
              genfft          -0.8%     -0.3%      0.03      0.03     +0.0%<br>
                  gg          -0.6%    +40.4%      0.01      0.02    +50.0%<br>
                grep          -1.8%     +2.3%      0.00      0.00     +0.0%<br>
              hidden          -0.7%     +3.5%     +4.1%     +7.6%     +0.0%<br>
                 hpg          -1.3%     -1.3%      0.05      0.10     +0.0%<br>
                 ida          -0.7%     -1.0%      0.07      0.08    +11.1%<br>
               infer          -0.9%     +0.5%      0.05      0.05    +30.0%<br>
             integer          -0.8%     +1.1%     +1.1%     +2.3%     +0.0%<br>
           integrate          -0.9%    +56.2%      0.20      0.23     +1.0%<br>
               kahan          -0.9%   +144.9%    +77.6%    +78.4%     +0.0%<br>
             knights          -0.1%     -0.4%      0.01      0.01     +0.0%<br>
                lcss          -0.7%     +0.7%    -24.3%    -18.4%     +1.9%<br>
                life          -0.2%     +0.0%      0.16      0.16     +0.0%<br>
                lift          -0.0%     +0.1%      0.00      0.00     +0.0%<br>
           listcompr          -0.1%     -0.0%      0.06      0.06     +0.0%<br>
            listcopy          -0.1%     -0.0%      0.06      0.06     +0.0%<br>
            maillist          +0.0%     +1.9%      0.02      0.04    +16.5%<br>
              mandel          -0.8%     +0.0%      0.05      0.05     +0.0%<br>
             mandel2          -0.1%     -4.0%      0.00      0.01     +0.0%<br>
             minimax          -0.2%     -0.0%      0.00      0.00     +0.0%<br>
             mkhprog          -0.1%     +1.1%      0.00      0.00     +0.0%<br>
          multiplier          -1.2%     +0.0%      0.07      0.08     +0.0%<br>
            nucleic2          -3.3%    +19.5%      0.05      0.05     +0.0%<br>
                para          -0.0%    +25.0%      0.22      0.23     +0.0%<br>
           paraffins          -0.8%     +0.0%      0.06      0.08     +7.5%<br>
              parser          -1.3%    +22.2%      0.03      0.03     +0.0%<br>
             parstof          -0.5%     +4.6%      0.01      0.00     +0.0%<br>
                 pic          +0.5%     +0.0%      0.00      0.00     +0.0%<br>
               power          -1.0%     +0.0%     -0.5%     +1.0%     +0.0%<br>
              pretty          -0.2%     +0.0%      0.00      0.00     +0.0%<br>
              primes          -0.8%     -0.0%      0.04      0.05     +0.0%<br>
           primetest          -0.7%     +0.0%      0.07      0.07     +0.0%<br>
              prolog          -0.2%    +16.0%      0.00      0.00     +0.0%<br>
              puzzle          -0.1%     -2.1%      0.09      0.10     +0.0%<br>
              queens          -0.8%     +0.0%      0.02      0.02     +0.0%<br>
             reptile          -0.8%     +0.6%      0.01      0.02     +0.0%<br>
             rewrite          -0.8%     +0.7%      0.02      0.02     +0.0%<br>
                rfib          -1.0%     +0.4%      0.02      0.02     +0.0%<br>
                 rsa          -0.7%     +2.5%      0.02      0.02     +0.0%<br>
                 scc          -0.1%     +0.0%      0.00      0.00     +0.0%<br>
               sched          -0.8%     +0.0%      0.01      0.02     +0.0%<br>
                 scs          -1.9%     +0.8%     -5.3%     -2.7%     +0.0%<br>
              simple          -0.3%     -0.0%      0.15      0.16     +6.9%<br>
               solid          -0.7%     +0.0%      0.09      0.09     +0.0%<br>
             sorting          +0.0%    +55.9%      0.00      0.00     +0.0%<br>
              sphere          -0.9%     -1.8%      0.04      0.04     +0.0%<br>
              symalg          -0.8%     +0.3%      0.01      0.01     +0.0%<br>
                 tak          -0.8%     +0.9%      0.01      0.01     +0.0%<br>
           transform          -0.8%     +0.0%     -3.3%     -5.7%     +0.0%<br>
            treejoin          +0.1%   +109.3%      0.15      0.17     -7.4%<br>
           typecheck          -0.8%     +0.0%      0.14      0.15     +0.0%<br>
             veritas          -0.8%     +0.0%      0.00      0.00     +0.0%<br>
                wang          -0.9%     +0.0%      0.07      0.08     +0.0%<br>
           wave4main          -0.8%     +1.8%      0.18      0.19     -7.1%<br>
        wheel-sieve1          -0.7%     +0.0%     +0.0%     +1.5%    -12.5%<br>
        wheel-sieve2          -0.7%     +0.0%      0.11      0.12     +2.1%<br>
                x2n1          +7.4%    +43.3%      0.01      0.01   +200.0%<br>
    ------------------------------<u></u>------------------------------<u></u>--------------------<br>
                 Min          -3.3%     -4.0%    -24.3%    -18.4%    -12.5%<br>
                 Max          +7.4%   +144.9%    +77.6%    +78.4%   +200.0%<br>
      Geometric Mean          -0.5%     +5.5%     +1.8%     +4.3%     +3.2%<br>
<br>
    I haven&#39;t had time to look through the regressions yet, so if<br>
    someone has time, please grab a benchmark that looks bad and have a<br>
    look at the Core to see what&#39;s going on. I suggest starting with<br>
    &quot;kahan&quot;, which is a relatively simple benchmark.<br>
<br>
    -- Johan<br>
<br>
<br>
    ______________________________<u></u>_________________<br>
    ghc-devs mailing list<br></div></div>
    <a href="mailto:ghc-devs@haskell.org" target="_blank">ghc-devs@haskell.org</a> &lt;mailto:<a href="mailto:ghc-devs@haskell.org" target="_blank">ghc-devs@haskell.org</a>&gt;<br>
    <a href="http://www.haskell.org/mailman/listinfo/ghc-devs" target="_blank">http://www.haskell.org/<u></u>mailman/listinfo/ghc-devs</a><div class="im"><br>
<br>
<br>
<br>
<br>
______________________________<u></u>_________________<br>
ghc-devs mailing list<br>
<a href="mailto:ghc-devs@haskell.org" target="_blank">ghc-devs@haskell.org</a><br>
<a href="http://www.haskell.org/mailman/listinfo/ghc-devs" target="_blank">http://www.haskell.org/<u></u>mailman/listinfo/ghc-devs</a><br>
<br>
</div></blockquote>
<br>
</blockquote></div><br></div></div>