Change S_ckwarn_common() to looping over the packed warning value.

The core never uses WARN3() or WARN4(), and rarely uses WARN2(), so the
previous code, effectively an unwrapped loop, wasn't a speed up. Functionally
equivalent smaller code fits better into CPU caches.

1 files changed, 18 insertions, 7 deletions

diff --git a/util.c b/util.c
index 13b56a0802..b72f263dc5 100644
--- a/util.c
+++ b/util.c
@@ -1627,15 +1627,26 @@ S_ckwarn_common(pTHX_ U32 w)
     if (PL_curcop->cop_warnings == pWARN_NONE)
 	return FALSE;
 
+    /* Check the assumption that at least the first slot is non-zero.  */
+    assert(unpackWARN1(w));
+
+    /* Check the assumption that it is valid to stop as soon as a zero slot is
+       seen.  */
+    if (!unpackWARN2(w)) {
+	assert(!unpackWARN3(w));
+	assert(!unpackWARN4(w));
+    } else if (!unpackWARN3(w)) {
+	assert(!unpackWARN4(w));
+    }
+	
     /* Right, dealt with all the special cases, which are implemented as non-
        pointers, so there is a pointer to a real warnings mask.  */
-    return isWARN_on(PL_curcop->cop_warnings, unpackWARN1(w))
-	|| (unpackWARN2(w) &&
-	    isWARN_on(PL_curcop->cop_warnings, unpackWARN2(w)))
-	|| (unpackWARN3(w) &&
-	    isWARN_on(PL_curcop->cop_warnings, unpackWARN3(w)))
-	|| (unpackWARN4(w) &&
-	    isWARN_on(PL_curcop->cop_warnings, unpackWARN4(w)));
+    do {
+	if (isWARN_on(PL_curcop->cop_warnings, unpackWARN1(w)))
+	    return TRUE;
+    } while (w >>= WARNshift);
+
+    return FALSE;
 }
 
 /* Set buffer=NULL to get a new one.  */