comparison mercurial/bdiff.c @ 9534:8e202431d620

bdiff: gradually enable the popularity hack Patch from Jason Orendorff The lower the threshold, the stronger the popularity hack's influence. So at 3999 lines, the hack is disabled; and at 4000 lines, the hack is enabled at maximum strength (t=4). No source file in mercurial/crew is over 4000 lines. But there are, oh, a few such files in Mozilla. I can testify that this hack causes hg to generate some correct but eyebrow-raising patches. I think the hack should phase in gradually. The threshold should be high for small files where we don't need it so much. Like this: t = (bn < 31000) ? 1000000 / bn : bn / 1000; That would leave the popularity hack disabled for small files, then gradually phase it in: bn < 1000 -- t > bn (popularity hack is completely disabled) bn == 1000 -- t = 1000 (still effectively disabled) bn == 2000 -- t = 500 (only hits unusual files) bn == 10000 -- t = 100 (only hits especially common lines) bn == 31000 -- t = 31 (hack is at maximum power) bn == 32000 -- t = 32 (hack could backfire, ease off)
author Benoit Boissinot <benoit.boissinot@ens-lyon.org>
date Sat, 03 Oct 2009 23:36:08 +0200
parents 16f6c13706df
children 08a0f04b56bd
comparison
equal deleted inserted replaced
9490:67df9f46f907 9534:8e202431d620
149 h[j].pos = i; 149 h[j].pos = i;
150 h[j].len++; /* keep track of popularity */ 150 h[j].len++; /* keep track of popularity */
151 } 151 }
152 152
153 /* compute popularity threshold */ 153 /* compute popularity threshold */
154 t = (bn >= 4000) ? bn / 1000 : bn + 1; 154 t = (bn >= 31000) ? bn / 1000 : 1000000 / (bn + 1);
155 155
156 /* match items in a to their equivalence class in b */ 156 /* match items in a to their equivalence class in b */
157 for (i = 0; i < an; i++) { 157 for (i = 0; i < an; i++) {
158 /* find the equivalence class */ 158 /* find the equivalence class */
159 for (j = a[i].h & buckets; h[j].pos != INT_MAX; 159 for (j = a[i].h & buckets; h[j].pos != INT_MAX;