GNU bug report logs -
#79035
Remove micro-optimization for Funintern for performance
Previous Next
Full log
Message #5 received at submit <at> debbugs.gnu.org (full text, mbox):
[Message part 1 (text/plain, inline)]
Based on my measurement of my reentrant reader submission, I tried
just eliminating the static variable used to "optimize" out a
redundant lookup by stashing the index in a static variable used only
by Funintern. See attached - I just repeated the lookup code in
Funintern, removed the assignment from oblookup, and eliminated the
static variable. At least for -O0, storing this information in the
frequently called oblookup to save a lookup in the very infrequently
called Funintern is a bit of a pessimization. Based on summing the
profiles of all the byte-compile jobs, and filtering for
obarray-related functions in read.c, I get the following for commit
a8b65860a5e28ee0867e8506a17d74d4a9b7783a (the cumulative is all
functions in lread.c, most of which I am eliding):
Each sample counts as 0.01 seconds.
% cumulative self self total
time seconds seconds calls s/call s/call name
13.44 31.82 8.68 79508607 0.00 0.00 oblookup
0.85 58.97 0.55 260 2.12 2.21 grow_obarray
0.84 59.51 0.54 83352123 0.00 0.00 obarray_index
0.81 60.03 0.52 78691233 0.00 0.00
oblookup_considering_shorthand
0.37 62.19 0.24 6858863 0.00 0.00 intern_sym
0.29 63.03 0.19 4635384 0.00 0.00 read_internal_start
0.08 64.05 0.05 20570557 0.00 0.00 mapatoms_1
0.06 64.23 0.04 1060 0.04 0.04 map_obarray
0.05 64.30 0.03 6858863 0.00 0.00 intern_driver
0.05 64.33 0.03 1143939 0.00 0.00 Fintern
0.03 64.48 0.02 Funintern
0.02 64.57 0.01 66751 0.00 0.00 Fintern_soft
0.00 64.59 0.00 22626 0.00 0.00
Flread__substitute_object_in_subtree
0.00 64.59 0.00 22626 0.00 0.00 substitute_object_recurse
0.00 64.59 0.00 13223 0.00 0.00 Fobarray_make
0.00 64.59 0.00 13223 0.00 0.00 allocate_obarray
0.00 64.59 0.00 13223 0.00 0.00 make_obarray
0.00 64.59 0.00 10916 0.00 0.00 Fobarrayp
0.00 64.59 0.00 5032 0.00 0.00 Flocate_file_internal
0.00 64.59 0.00 2522 0.00 0.00 intern_1
0.00 64.59 0.00 1043 0.00 0.00 Fmapatoms
The same commit with the attached patch:
13.22 31.13 8.23 79563723 0.00 0.00 oblookup
0.67 57.22 0.42 264 1.59 1.66 grow_obarray
0.61 57.60 0.38 83407339 0.00 0.00 obarray_index
0.61 57.98 0.38 78745509 0.00 0.00
oblookup_considering_shorthand
0.25 60.64 0.16 6858324 0.00 0.00 intern_sym
0.19 61.32 0.12 4635494 0.00 0.00 read_internal_start
0.11 61.59 0.07 20669351 0.00 0.00 mapatoms_1
0.11 61.66 0.07 1067 0.07 0.07 map_obarray
0.10 61.84 0.06 Funintern
0.05 62.10 0.03 66913 0.00 0.00 Fintern_soft
0.01 62.27 0.01 6858324 0.00 0.00 intern_driver
0.00 62.28 0.00 1146143 0.00 0.00 Fintern
0.00 62.28 0.00 22460 0.00 0.00
Flread__substitute_object_in_subtree
0.00 62.28 0.00 22460 0.00 0.00 substitute_object_recurse
0.00 62.28 0.00 13219 0.00 0.00 Fobarray_make
0.00 62.28 0.00 13219 0.00 0.00 allocate_obarray
0.00 62.28 0.00 13219 0.00 0.00 make_obarray
0.00 62.28 0.00 10918 0.00 0.00 Fobarrayp
0.00 62.28 0.00 5031 0.00 0.00 Flocate_file_internal
0.00 62.28 0.00 2522 0.00 0.00 intern_1
0.00 62.28 0.00 1050 0.00 0.00 Fmapatoms
The total reported cpu time for
a8b65860a5e28ee0867e8506a17d74d4a9b7783a is 3479.51s, and with the
patch it is 3460.69.
You can see Funintern is more expensive, but also that in all those
jobs no actual calls were recorded in either case. I see similar
differences for these functions when I compare results for commit
2bdcf0250acecdb0719203ae711aedf5baad1783.
As a reminder, I submitted my assignment paperwork a couple of years
ago, although this seems like a pretty trivial change.
Lynn
[0001-Remove-micro-optimization-for-Funintern.patch (text/x-patch, attachment)]
This bug report was last modified 12 days ago.
Previous Next
GNU bug tracking system
Copyright (C) 1999 Darren O. Benham,
1997,2003 nCipher Corporation Ltd,
1994-97 Ian Jackson.