This is the mail archive of the
mailing list for the GCC project.
Re: [tree-ssa] Re: COND_EXPR lowering patch
- From: Andrew MacLeod <amacleod at redhat dot com>
- To: Jeff Law <law at redhat dot com>
- Cc: Zdenek Dvorak <rakdver at atrey dot karlin dot mff dot cuni dot cz>, gcc mailing list <gcc at gcc dot gnu dot org>
- Date: 25 Aug 2003 14:24:02 -0400
- Subject: Re: [tree-ssa] Re: COND_EXPR lowering patch
- References: <200308251550.h7PFoJqL027956@speedy.slc.redhat.com>
On Mon, 2003-08-25 at 11:50, firstname.lastname@example.org wrote:
> I think these diffs were set up to allow passing in a flag which
> enabled removal of conditionals/control structures and thus
> avoided the overhead of pdom calculation in that case.
> I think the code still uses parents and marks them as necessary --
> that was done to see if using that got us through the worklist
> faster. I'm pretty sure it was a wash and you don't need to
> muck around calling parent_stmt at all.
> It cleaned up some of the code which cached cond_checked and
> goto_checked as well. Basically we just need to know which
> blocks were visited since any "magic" stuff happens once the first
> time the block becomes executable.
> Note there may be some debugging statements left in the code -- look
> look for "Yow".
I'll combine this and mine...