develooper Front page | perl.perl5.porters | Postings from June 2008

Re: More results from llvm-gcc

Thread Previous | Thread Next
From:
Steve Peters
Date:
June 4, 2008 10:27
Subject:
Re: More results from llvm-gcc
Message ID:
fd7a59d30806041027y2c4af02am28747ffb53177dc3@mail.gmail.com
On Wed, Jun 4, 2008 at 12:13 PM, Yuval Kogman <nothingmuch@woobling.org> wrote:
> On Wed, Jun 04, 2008 at 17:46:15 +0100, Nicholas Clark wrote:
>
>> But you're also comparing 4.2 with 4.0. What's llvm-gcc like against 4.2?
>
> Given that no GCC optimizations are performed by llvm-gcc it doesn't
> matter that much, but the results are:
>
>                      system gcc-4.2  llvmbc
>                      ------ -------  ------
> arith/mixed              100      82     119
> arith/trig               100      88     103
> array/copy               100      96     130
> array/foreach            100      83     134
> array/index              100      96     134
> array/pop                100      98     125
> array/shift              100      95     130
> array/sort-num           100     100     103
> array/sort               100      96     109
> call/0arg                100      81     148
> call/1arg                100      79     133
> call/2arg                100      86     133
> call/9arg                100      89     132
> call/empty               100      67     139
> call/fib                 100      87     130
> call/method              100      86     126
> call/wantarray           100      93     125
> hash/copy                100      87     112
> hash/each                100      97     127
> hash/foreach-sort        100      94     106
> hash/foreach             100      85     126
> hash/get                 100      76      83
> hash/set                 100      82     108
> loop/for-c               100      81      95
> loop/for-range-const     100     113     118
> loop/for-range           100     111     120
> loop/getline             100      94     106
> loop/while-my            100      82      96
> loop/while               100      85      85
> re/const                 100      98      99
> re/w                     100     106     117
> startup/fewmod           100     102     105
> startup/lotsofsub        100     100     103
> startup/noprog           100     108     104
> string/base64            100     105     115
> string/htmlparser        100      90     104
> string/index-const       100      99     109
> string/index-var         100      84     114
> string/ipol              100      86      83
> string/tr                100      97      96
>
> AVERAGE                  100      92     115
>

I would be curious to see how Intel C++ and Sun Studio on Linux are
doing in comparison.  Based on my previous smokes, they usually beat
the pants off of gcc.

Steve Peters
steve@fisharerojo.org

Thread Previous | Thread Next


nntp.perl.org: Perl Programming lists via nntp and http.
Comments to Ask Bjørn Hansen at ask@perl.org | Group listing | About