I have spotted the biggest bottleneck in "bdiff.c". Actually it was
pretty easy to find after I recompiled the python interpreter and
mercurial for profiling.
In "bdiff.c" function "equatelines" allocates the minimum hash table
size, which can lead to tons of collisions. I introduced an
"overcommit" factor of 16, this is, I allocate 16 times more memory
than the minimum value. Overcommiting 128 times does not improve the
performance over the 16-times case.
#!/bin/sh
echo "[extensions]" >> $HGRCPATH
echo "mq=" >> $HGRCPATH
cat >> readlink.py <<EOF
import errno, os, sys
for f in sys.argv[1:]:
try:
print f, '->', os.readlink(f)
except OSError, err:
if err.errno != errno.EINVAL: raise
print f, 'not a symlink'
EOF
hg init
hg qinit
hg qnew base.patch
echo a > a
echo b > b
hg add a b
hg qrefresh
python readlink.py a
hg qnew symlink.patch
rm a
ln -s b a
hg qrefresh --git
python readlink.py a
hg qpop
hg qpush
python readlink.py a