Re: Performance on compilers

[Mark Johnson <mark_h_johnson@raytheon.com>]

Kjell Mesch wrote:

> Hi
>
> I'm looking for some kind of performance test on ada compilers.
> (Green Hill vs Gnat)
> Can anyone point me in some direction? :-)
>
> Anyone out that like/dislike GHS enviroment?
> (sun/ppc_vxworks configuration of AdaMulti2000)

Another solution could be ACES - it is a huge benchmark suite (over a
thousand tests) which includes compilation speed, as one of its measures
(along with memory & CPU usage for the generated application). Many
people have had problems building w/ it. I've gotten farther than many
have but I've never been able to get all the tests to work. If you are
interested in a several day investment, look at...

http://www.informatik.uni-stuttgart.de/ifi/ps/AdaBasis/pal_1195/ada/ajpo/testing/aces/

This seems to be the only site that hasn't discarded all the old data.
If you want to follow up on building the stuff under GNAT, I'll provide
that separately.
  --Mark

PS - with GNAT in particular, I've noticed having LOTS of memory makes a
huge difference when compiling big packages w/ optimization enabled.
I've seen memory usage grow to >170 Mbyte and a 10 minute compile time
(on an 800mhz, 512 Mbyte Pentium III) on a particularly large file. The
typical compile time is much better - I rebuild a few thousand packages
(for 22 main programs) in about two hours of CPU time - even with that
big one in the list....