[ODE] Intel/Msvc Compiler

Peter Honeder peter.honeder at hlw.co.at
Tue Jan 6 17:01:27 MST 2004


Hi Tom!

Tom De Muer wrote:
> Hi Folks,
> 
> Just did a little test yesterday, I compiled ODE single precision (the msvc7
> contrib/default) with MSVC and got my car simulated at +/- 700 fps, the
> intel compiled code run at 870 fps.  The intel compiler also has automatic
> vectorisation that converts regular loops on floats into loops with SIMD but
> my Athlon only has limited simd support and it still is the _intel_ compiler
> ;)...  I thought it may be interesting in the people trying optimising stuff
> as I see topics about that reguraly popping up :o)  Has anybody else tried
> to fiddle with the intel/msvc compiler settings for optimal performance?
> 

The intel compiler does optimize really, really good and has proven to 
be most of the time 20-30% faster than msvc or g++. Especially maths 
code seems to get optimized really good. We had some test programs to 
test typical maths applications (Matrix libraries, calculating fractals, 
image libraries,...) and nearly all the time the intel compiler won, 
most of the time being 20-30% faster.
Notable was also, that the intel compiler can reach the speedup even on 
non-intel cpus.

Your test results seem reasonable and in the range of the normal 
possible speed optimization possible with the intel compiler.

An interesting link for further reading might be:
http://www.linuxjournal.com/article.php?sid=6766
which also includes some links to performance tests and optimization 
explanations.

Best regards,

Peter

-- 
Peter Honeder
http://engine.hlw.co.at




More information about the ODE mailing list