This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [PATCH] fwprop, updated patch and SPEC results
- From: Paolo Bonzini <paolo dot bonzini at lu dot unisi dot ch>
- To: Jan Hubicka <hubicka at ucw dot cz>
- Cc: GCC Patches <gcc-patches at gcc dot gnu dot org>, Roger Sayle <roger at eyesopen dot com>, Steven Bosscher <stevenb dot gcc at gmail dot com>, Kenneth Zadeck <zadeck at naturalbridge dot com>
- Date: Sat, 04 Nov 2006 18:30:17 +0100
- Subject: Re: [PATCH] fwprop, updated patch and SPEC results
- References: <454B91E2.9030300@lu.unisi.ch> <20061104154453.GA11911@atrey.karlin.mff.cuni.cz>
As reported on IRC, this is caused by the bitmaps for reaching
definitions. The increased peak consumption is because df allocates
many bitmaps at the same time, more so than out-of-ssa or flow.c.
fwprop is the first pass to use df for loopless testcases such as this one.
Bitmap Overall Allocated
Peak Leak
---------------------------------------------------------------------------------
df-problems.c:986 (df_rd_alloc) 46040 259809052 259801720 0
df-problems.c:987 (df_rd_alloc) 46040 259827812 259816280 0
It may be worth taking a look at how the testcases behaves in dataflow
branch. (Just to recall, this is bug.c attached to PR28071, compiled
with -O3 -fno-tree-pre -fno-tree-fre). I'm using the bitmap statistics
patch by Jan (available at
http://gcc.gnu.org/ml/gcc-patches/2006-07/msg01244.html with
s/bitmap_hash/bitmap_names_hash/g prior to applying it) to get the above
data.
Paolo