This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: Post reload copy propagation?
- To: Michael Hayes <m dot hayes at elec dot canterbury dot ac dot nz>
- Subject: Re: Post reload copy propagation?
- From: Jeffrey A Law <law at cygnus dot com>
- Date: Sun, 12 Sep 1999 22:50:56 -0600
- cc: gcc at gcc dot gnu dot org
- Reply-To: law at cygnus dot com
In message <14299.34598.526694.920052@andromeda.elec.canterbury.ac.nz>you wri
te:
>
> Could anyone familiar with the GCSE pass tell me how feasible would it
> be to run it on hard registers after reload? This could help mop up
> some unecessary copies introduced during reload and propagation of
> registers used as function arguments.
You do have some technical issues to solve if you want to use it after reload.
1. cse.c basically ignores hard registers.
2. gcse does not update death notes in any way shape or form.
3. redundant expression motion/removal after reload is effectively assignment
motion and you can't compute safe motions in parallel since one motion
is likely to effect another motion because we work strictly on hard regs.
Jeff