Result of today. I cherry-picked nd/read-directory-recursive-optim to
see how far I can get after pulling all the tricks. This is a slower
machine so time is longer. Anyway, read_directory time is reduced
about 70% in the end.
function before after
----------------------------------
treat_leading_path: 0.000 0.000
read_directory: 4.102 1.235
+treat_one_path: 2.843 0.531
++is_excluded: 2.632 0.102
+++prep_exclude: 0.225 0.040
+++matching: 2.054 0.035
++dir_exist: 0.035 0.035
++index_name_exists: 0.292 0.225
lazy_init_name_hash: 0.258 0.155
+simplify_away: 0.085 0.083
+dir_add_name: 0.446 0.000
I don't expect all these patches to go in. The meat is
nd/read-directory-recursive-optim (or 10/13) and 09/13. Some other
patches are safe even if they do not contribute much to the gain. The
last two are probably not worth the trouble.
Nguyễn Thái Ngọc Duy (13):
dir.c: add MEASURE_EXCLUDE code for tracking exclude performance
match_pathname: avoid calling strncmp if baselen is 0
dir.c: inline convenient *_icase helpers
match_basename: use strncmp instead of strcmp
match_{base,path}name: replace strncmp_icase with memequal_icase
dir: pass pathname length to last_exclude_matching
exclude: avoid calling prep_exclude on entries of the same directory
exclude: record baselen in the pattern
exclude: filter out patterns not applicable to the current directory
read_directory: avoid invoking exclude machinery on tracked files
Preallocate hash tables when the number of inserts are known in advance
name-hash: allow to lookup a name with precalculated base hash
read_directory: calculate name hashes incrementally
Makefile | 1 +
attr.c | 6 +-
cache.h | 2 -
diffcore-rename.c | 1 +
dir.c | 392 ++++++++++++++++++++++++++++++++++++++++++++----------
dir.h | 26 +++-
hash.h | 7 +
name-hash.c | 49 ++++---
name-hash.h (new) | 45 +++++++
read-cache.c | 1 +
unpack-trees.c | 1 +
11 files changed, 431 insertions(+), 100 deletions(-)
create mode 100644 name-hash.h
--
1.8.1.2.536.gf441e6d
--
To unsubscribe from this list: send the line "unsubscribe git" in
the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
