Re: inline considered harmful

A while back I ran a large number of tests on optimisations across an
athlon t-bird 1.4g, pentium4M (dell laptop) and a cyrix 200 (for fun!). 
The applications were scientific, some home scripts and gimp
manipulations.

I was mainly interested in routines that run for many hours/days as this
was what was causing me the most grief at the time.  Basicly, ~10% gains
can be be seen between -Os (s must be for for slowest!) and -O2 (-O3 is
almost always slower than -O2)

Of course, results were highly variable, and application dependent.

I came to the conclusion that for me, compiling the system with
"-march=athlon-xp -O2 -pipe -fomit-frame-pointer -falign-functions=4"
gave the best general gain.  My next step was to be to compile all the
gui stuff with -Os to speed up startup - but I never got around to it.

Per package optimisations would be nice, but I can imagine it getting
into a far greater mess than USE flags are at the moment - getting
unusable.

BillK

On Thu, 2004-07-22 at 19:06, Jerry McBride wrote:
> On Thursday 22 July 2004 02:35 pm, Bart Alewijnse wrote:
> > I'm curious as to how many of you have considered and tried using -Os



--
[email protected] mailing list