This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: Local optimization on i386 ?
- To: hubicka at horac dot ta dot jcu dot cz
- Subject: Re: Local optimization on i386 ?
- From: Marc Lehmann <pcg at opengroup dot org>
- Date: Thu, 2 Dec 1999 01:06:39 +0100
- Cc: Marc Lehmann <pcg at opengroup dot org>, egcs at egcs dot cygnus dot com
- References: <26996.943919594@upchuck> <199911301338.OAA04341@quatramaran.ens.fr> <19991130232208.O351@cerebro.laendle> <19991201143623.B20663@horac.ta.jcu.cz>
On Wed, Dec 01, 1999 at 02:36:23PM +0100, hubicka@horac.ta.jcu.cz wrote:
> would be smart enought to join two modifications of stack even when they
> are far away and possibly the stack references are between (so it ought to be
> able to change this references)
>
> This is far beyhond the scope of peephole pass.
The principle problem is that I don't think this is feasible in
*general*. I Have not looked into this specific problem, but it might
be the case that one of these stack adjusts is emitted very late in the
compiler. This means that you would have to add a pass that could simplify
these, and this is very likely to introduce other suboptimal sequences
again.
That is, something *slightly* more intelligent than the current peepholer
would be nice, but a very intelligent pass might not be doable at this
late stage.
--
-----==- |
----==-- _ |
---==---(_)__ __ ____ __ Marc Lehmann +--
--==---/ / _ \/ // /\ \/ / pcg@opengroup.org |e|
-=====/_/_//_/\_,_/ /_/\_\ XX11-RIPE --+
The choice of a GNU generation |
|