Tree-SSA versus 3.4 branch garbage production

[Jan Hubicka <jh at suse dot cz>]

Hi,
I've recompiled 3.4 branch and tree-ssa with my memory statistics patch.
The short story is that Gerald's testcase nees 522MB on tree-SSA and 417
for 3.3 branch.  (just FYI for my current tree it is 190MB).
For combine.c it is 50MB versus 81MB. (and 19MB on my tree).

I am attaching the report for combine.i as Gerald's testcase may be
affected by fact that we do fewer inlining as gimplified bodies have
more MODIFY_EXPRs, but I can send them if there is interest.

HTH
Honza

Attachment: combine.mmr2
Description: Text document

Attachment: combine.mmr
Description: Text document