https://gcc.gnu.org/bugzilla/show_bug.cgi?id=62291
Steven Bosscher changed:
What|Removed |Added
Status|UNCONFIRMED |NEW
Last reconfirmed|
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=62291
--- Comment #1 from Richard Biener ---
Author: rguenth
Date: Fri Aug 29 12:39:50 2014
New Revision: 214727
URL: https://gcc.gnu.org/viewcvs?rev=214727&root=gcc&view=rev
Log:
2014-08-29 Richard Biener
PR tree-optimization/62291
* tree
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=62291
--- Comment #2 from Richard Biener ---
As said in the mail for the just applied patch:
"The second patch (once done) will refactor insertion phase
to do a dominator walk similar to what elimination does
computing AVAIL_OUT on-the-fly (and only k
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=62291
--- Comment #3 from Richard Biener ---
So what remains is insertion computing whether a value is available on some
predecessor (partial redundancy) or all (full redundancy) which needs AVAIL_OUT
for the predecessors and the immediate dominator.
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=62291
--- Comment #4 from Richard Biener ---
Btw, the testcase is absymally slow in compiling even with -O1 if you enable
glibc memory checking with both(!) MALLOC_CHECK_=3 and MALLOC_PERTURB_=69
I have no idea why... (constant overhead seems to be on
