Re: Omit frame pointer bug in m68k gcc 3.3.1

[Dan Kegel <dkegel at ixiacom dot com>]

David Brown wrote:
> Is it worth trying 3.4 rather than 3.3.1 ?  From my testing and brief
> benchtesting, it looks like the back-end of 3.3.1 is pretty much the same as
> for my old 2.95 version (a few things were slightly faster, and bitfield
> manipulation seemed to be a bit slower) on the cpu32.  The front-end
> optomisations (such as moving blocks of code around, and inline functions
> even if they were defined later in the code, with -O3 turned on) have
> definitely improved.  As far as I can tell, changes in 3.4 are going to be
> more on improving the speed of the compiler itself rather than changes at
> the back-end - is that correct?

http://gcc.gnu.org/gcc-3.4/changes.html seems to indicate only one
runtime speedup: arm-elf floating point.
It also indicates a bunch of new platforms.

So I'd say it's only worth trying 3.4 if you're interested
in future-proofing your project (3.4 is a lot stricter
about what it accepts), or if you want to try out
the compile-time speedups like precompiled headers.

It's easy enough to build it, at least for the platforms
I've tried.  I only had to add a couple changes to crosstool
to support it.

The last 3.4 snapshot I tried had a bunch of regression failures,
but not enough to ensure every project using it would fail :-)

- Dan


------
Want more information?  See the CrossGCC FAQ, http://www.objsw.com/CrossGCC/
Want to unsubscribe? Send a note to crossgcc-unsubscribe@sources.redhat.com