[PATCH] Cache most recent regcomp() call in _FcStrRegexCmp().

[Nick Alcock <nick.alcock@xxxxxxxxxx>]

regcomp() is not a notably fast function call.  As the name suggests it is
basically a compiler, and even for calls with literal strings it is a fairly
intricate one.  In particular it calls malloc() more than once, and malloc()
itself is not a particularly fast function call, especially not when the arena
is full.

Under the circumstances, it is unfortunate that _FcStrRegexCmp() is called so
very often by fontconfig.  On my system (with ~5000 fonts, and a stripped-down
fonts.conf), mapping a new Emacs frame makes twelve calls to FcFontMatch()...
and those twelve calls proceed to call _FcStrRegexCmp() 175171 times, with a
call to regcomp() every time!  This then proceeds to invoke malloc() on the
order of three million times.  It is fairly easy to turn this into a
pathological slowdown at runtime.  My Emacsen are not small (arena sizes of a
gigabyte-plus are common, with a highly fragmented heap), and in that situation
those three million malloc() and free() calls can easily consume in excess of
fifty seconds (though on a fresh startup it takes 'only' a quarter of a second
or so).

Since each call to FcFontMatch() in this pathological case (and probably in most
other cases) calls _FcStrRegexCmp() with a constant argument, a trivial cache
consisting of a single static variable suffices to reduce the number of calls to
regcomp() to... eleven.  (The precise method used may need a little rethinking
for thread-safety, but the general principle, of reducing regcomp() calls via
some variety of caching, appears to be essential.)

---
 src/fcstr.c | 31 +++++++++++++++++++++++--------
 1 file changed, 23 insertions(+), 8 deletions(-)

[Resent for the second time: if this doesn't get through I will consider
 the fontconfig list broken...]

diff --git a/src/fcstr.c b/src/fcstr.c
index 339a346..fa4560d 100644
--- a/src/fcstr.c
+++ b/src/fcstr.c
@@ -247,19 +247,32 @@ static FcBool
 _FcStrRegexCmp (const FcChar8 *s, const FcChar8 *regex, int cflags, int eflags)
 {
     int ret = -1;
-    regex_t reg;
+    static regex_t reg;
+    static FcChar8 *last;
 
-    if ((ret = regcomp (&reg, (const char *)regex, cflags)) != 0)
+    if (!regex || !last || FcStrCmp (regex, last) != 0)
     {
-	if (FcDebug () & FC_DBG_MATCHV)
+	if (last != NULL)
 	{
-	    char buf[512];
+	    regfree (&reg);
+	    last = NULL;
+	}
 
-	    regerror (ret, &reg, buf, 512);
-	    printf("Regexp compile error: %s\n", buf);
+	if ((ret = regcomp (&reg, (const char *)regex, cflags)) != 0)
+	{
+	    if (FcDebug () & FC_DBG_MATCHV)
+	    {
+		char buf[512];
+
+		regerror (ret, &reg, buf, 512);
+		printf("Regexp compile error: %s\n", buf);
+	    }
+	    return FcFalse;
 	}
-	return FcFalse;
+	FcFree (last);
+	last = FcStrdup (regex);
     }
+
     ret = regexec (&reg, (const char *)s, 0, NULL, eflags);
     if (ret != 0)
     {
@@ -271,7 +284,9 @@ _FcStrRegexCmp (const FcChar8 *s, const FcChar8 *regex, int cflags, int eflags)
 	    printf("Regexp exec error: %s\n", buf);
 	}
     }
-    regfree (&reg);
+
+    if (last == NULL)  /* only on OOM */
+	regfree (&reg);
 
     return ret == 0 ? FcTrue : FcFalse;
 }
-- 
1.8.2.1.162.ge277c5f
_______________________________________________
Fontconfig mailing list
Fontconfig@xxxxxxxxxxxxxxxxxxxxx
http://lists.freedesktop.org/mailman/listinfo/fontconfig