Mercurial > hg-stable
changeset 2577:fa76c5d609c9
bdiff: improve worst case behavior by 100x.
on 5.8MB (244.000 lines) text file with similar lines, hash before
this change made diff against empty file take 75 seconds. this change
improves performance to 0.6 seconds. result is that clone of smallish
repo (137MB) with some files like this takes 1 minute instead of 10
minutes.
common case of diff is 10% slower now, probably because of worse cache
locality. but diff does not affect overall performance in common case
(less than 1% of runtime is in diff when it is working ok), so this
tradeoff looks good.
author | Vadim Gelfer <vadim.gelfer@gmail.com> |
---|---|
date | Fri, 07 Jul 2006 15:02:55 -0700 |
parents | 6a961a54f953 |
children | cf4f0322851d |
files | mercurial/bdiff.c |
diffstat | 1 files changed, 11 insertions(+), 2 deletions(-) [+] |
line wrap: on
line diff
--- a/mercurial/bdiff.c Fri Jul 07 11:23:53 2006 +0200 +++ b/mercurial/bdiff.c Fri Jul 07 15:02:55 2006 -0700 @@ -65,7 +65,7 @@ int splitlines(const char *a, int len, struct line **lr) { - int h, i; + int g, h, i; const char *p, *b = a; struct line *l; @@ -82,7 +82,16 @@ /* build the line array and calculate hashes */ h = 0; for (p = a; p < a + len; p++) { - h = *p + rol32(h, 7); /* a simple hash from GNU diff */ + /* + * a simple hash from GNU diff, with better collision + * resistance from hashpjw. this slows down common + * case by 10%, but speeds up worst case by 100x. + */ + h = *p + rol32(h, 7); + if ((g = h & 0xf0000000)) { + h ^= g >> 24; + h ^= g; + } if (*p == '\n' || p == a + len - 1) { l->len = p - b + 1; l->h = h * l->len;