Re: [MiNT] Why not O3

[François LE COAT <lecoat@ief.u-psud.fr>]

Hi,

Konrad M. Kokoszkiewicz wrote:
> I wrote :

> > In facts, Atari machines have very tiny cache memory, and there's
> > no L2 caches. So, optimisations for speed (-O3) that generate
> > supposed faster code, by the way of unrolling loops and others
> > expensive optimisations etc ... are not efficient on our
> > platforms. The compacity of executable's code is much more
> > efficient.
>

> Exactly that's why it was changed from -O3 to -O2.

Thanks to a previous E-mail of Thomas Binder on that
list, I discovered the "-Os" option. So far, I believed
that "-O1" was an executable size optimization. "-O1"
is rather the first level of speed optimizations. And
that option is the fastest one that I tried on my
Hades 060. This is due to the reasons I explained
in my previous message. "-Os" is pure size optimizations,
but it is not as efficient as "-O1". I'm pleased that
you agree with me. "-O2" is the default optimization
level of GCC, but "-O1" seems to me to be the best
compromise between size and speed. That's what I
experienced on Atari machines. "-O2" is too fat,
and "-Os" is too slow. Maybe you would have a similar
experience ...

Regards,

Dr. François LE COAT
Author of Eureka 2.12 (2D Graph Describer 3D Modeller)
http://eureka.atari.org
mailto:lecoat@atari.org