Re: PPC optimisations

[David Edelsohn <dje at watson dot ibm dot com>]

>>>>> Ralph Schmidt writes:

Ralph> Mostly peephole optimisations...so..architecture-specific.

	Peephole optimizations mean something specific in GCC which is
more restrictive than the general concept of peephole optimizations used
in the compiler community.  Again, you need to be careful with your words
and definition or you will send people down the wrong path.

Ralph> I know...i want to go into more detail when i have franz snapshot really
Ralph> working on my host so i don't report already fixed optimisation misses.

	Again, Franz's snapshots do not correspond to the GCC development
snapshots.  I do not know what Franz is including -- only gcc-2.95 release
branch, additional improvements which may or may not have been contributed
back to GCC (as opposed to Franz's code base), additional improvements
from the main development trunk not integrated into the gcc-2.95 release
branch, etc.  We need to get Franz involved to know what you are
comparing.

	In general, comments about the gcc-2.95 release branch are not
that useful because only bugfixes are applied to that source base, not new
optimizations.  If you want to know if a missed optimization already has
been addressed, you need to test the (unstable) public development
sources.

	I think that you need to coordinate more with Franz or else your
comments may have very limited use.  We want GCC on PowerPC to produce
good, optimized code, but comments based on a non-standard GCC code base
which itself is based on a version of GCC originally forked from the
development trunk over a year ago is quite distant from what is being
developed today.

David