xdiff: give up scanning similar lines early
In a corner case of large files whose lines do not match uniquely, the loop to eliminate a line that matches multiple locations adjacent to a run of lines that do not uniquely match wasted too much cycles. Fix this by giving up early after scanning 100 lines in both direction. Signed-off-by: Junio C Hamano <gitster@pobox.com>
Davide Libenzi committed
Nov 7, 2008 at 21:24 UTC
733070bea9ce28d557f0b0950acc05b3fe6ff6c6
1 file changed
+13
-2
xdiff/xprepare.c
+13
-2
@@ -23,10 +23,9 @@
23
#include "xinclude.h"
24
25
26
-
26
#define XDL_KPDIS_RUN 4
27
#define XDL_MAX_EQLIMIT 1024
29
-
28
+#define XDL_SIMSCAN_WINDOW 100
29
30
31
typedef struct s_xdlclass {
@@ -312,6 +311,18 @@ void xdl_free_env(xdfenv_t *xe) {
311
static int xdl_clean_mmatch(char const *dis, long i, long s, long e) {
312
long r, rdis0, rpdis0, rdis1, rpdis1;
313
314
+ /*
315
+ * Limits the window the is examined during the similar-lines
316
+ * scan. The loops below stops when dis[i - r] == 1 (line that
317
+ * has no match), but there are corner cases where the loop
318
+ * proceed all the way to the extremities by causing huge
319
+ * performance penalties in case of big files.
320
+ */
321
+ if (i - s > XDL_SIMSCAN_WINDOW)
322
+ s = i - XDL_SIMSCAN_WINDOW;
323
+ if (e - i > XDL_SIMSCAN_WINDOW)
324
+ e = i + XDL_SIMSCAN_WINDOW;
325
+
326
/*
327
* Scans the lines before 'i' to find a run of lines that either
328
* have no match (dis[j] == 0) or have multiple matches (dis[j] > 1).