Re: [PATCH] PRE TLC, improve fake exit edge placement

2017-08-01 Thread Richard Biener
On Tue, 1 Aug 2017, Richard Biener wrote: > > When working on PR81181 I ran into some things I wanted to clean up > several times. First a few PRE cleanups done for the fix. Second, > the fake exit edges we add for infinite loops happen to start from > loop headers rather than latches which is

[PATCH] PRE TLC, improve fake exit edge placement

2017-08-01 Thread Richard Biener
When working on PR81181 I ran into some things I wanted to clean up several times. First a few PRE cleanups done for the fix. Second, the fake exit edges we add for infinite loops happen to start from loop headers rather than latches which is somewhat confusing and making PRE dataflow order

[PATCH] PRE TLC

2017-05-05 Thread Richard Biener
Bootstrapped and tested on x86_64-unknown-linux-gnu, applied. Richard. 2017-05-05 Richard Biener * tree-ssa-pre.c (get_or_alloc_expr_for): Simplify. Index: gcc/tree-ssa-pre.c === ---

[PATCH] PRE TLC

2014-09-02 Thread Richard Biener
The following patch removes dead code (blocks are never defered because we iterate in a proper CFG order now) and avoids building up the el_avail vector one element at a time. Bootstrapped on x86_64-unknown-linux-gnu, testing in progress. Richard. 2014-09-02 Richard Biener rguent...@suse.de

[PATCH] PRE TLC

2012-11-29 Thread Richard Biener
When working on PR55124 I noticed we do some useless work in PRE. This cleans it up and makes it more obvious what we do in compute_avail. Bootstrapped and tested on x86_64-unknown-linux-gnu, applied to trunk. Richard. 2012-11-29 Richard Biener rguent...@suse.de * tree-ssa-pre.c

[PATCH] PRE TLC

2012-09-21 Thread Richard Guenther
This removes the no longer required dominating stmt argument from the insertion routines (since the eliminate () reorg) and also reflects that insertion now never can fail (again). Bootstrapped on x86_64-unknown-linux-gnu, testing in progress. Richard. 2012-09-21 Richard Guenther

[PATCH] PRE TLC

2012-05-03 Thread Richard Guenther
When looking at PR53168 I noticed that clean () can be optimized and simplified quite a bit. Likewise that we do not dump whether we find a partial or a partial partial redundancy, nor dump the insert iteration. Also dumping the PRE sets is tedious from inside gdb so this adds a debug function