Re: Deciding new arm EABI port name
Riku Voipio wrote:
On Tue, Apr 04, 2006 at 12:09:11PM +0200, Pjotr Kourzanov wrote:
Thats unreleated to architecture strings. It's worth remembering
that compliler flags rarely turn O(n^2) solutions into O(1) solutions. The
linear max 5% increase will not solve the real big performance
problems.
*small* performance problems manifest themselves *big* way on *smaller*
systems. When you have to fit a distribution in 16 megs, even 5% counts...
If a clicking a button takes 5sec to react with default compiler options.
With best gcc tuning options you'll achieve 5% reduction in processing
time, and behold, the button reaction takes only 4.7 sec now!
Yes, but clicking on a button does not represent a performance
problem of any size. I refer to code that is (re)used by many apps
such as kernel, libc, libstdc++, libgcc etc.
Slow/Small systems do not really change this basic issue IMHO. However,
you have appeared to change the subject to *size* optimizations, which
I agree *is* important on small systems.
The subject is small-system optimization in general, not only
for speed *or* size.
Multimedia codecs and games are ofcourse another story, as well as some
other places, I'm not trying to say that the compiler optimization is
*useless*, just that it often overrated.
I am not proposing to change Debian into Gentoo;-) I just propose
to help people building more customized Debian distributions by
allowing special archs like "xscale" or "arm-softfloat". This will also help
in transition to from "arm" to "arm-eabi".
Reply to: