grok_bin_oct_hex: Shortcut leading zeros

I noticed that leading zeros are quite common for octal and hex
constants.  This code is structured for speed, with a partially unrolled
loop structured so that it is impossible to overflow the unrolled part.

If we get to the end of the unrolled portion, and the accumulated value
is still zero, it's because there have been only leading zeroes so far,
and instead of dropping into the loop, we can re-enter the unrolled part
without having to consider the possibility of overflowing.  This allows
the next chunk of digits to be processed without branching.

Author: khwilliamson

numeric.c

@@ -427,6 +427,7 @@ Perl_grok_bin_oct_hex(pTHX_ const char *start,
 
     /* Unroll the loop so that the first 8 digits are branchless except for the
      * switch.  A ninth hex one overflows a 32 bit word. */
+  redo_switch:
     switch (e - s) {
       default:
           if (UNLIKELY(! generic_isCC_(*s, class_bit)))  break;
@@ -473,6 +474,15 @@ Perl_grok_bin_oct_hex(pTHX_ const char *start,
               return value;
           }
 
+          /* If we get here, and the accumulated value is still 0, it is
+           * because there are more leading zeros than the cases of this
+           * switch(),  These are common enough with these kinds of
+           * binary-style numbers that it is worth this extra conditional to
+           * continue absorbing them via the switch. */
+          if (value == 0) {
+              goto redo_switch;
+          }
+
           break;
     }
 
@@ -549,7 +559,16 @@ Perl_grok_bin_oct_hex(pTHX_ const char *start,
                                         != PERL_SCAN_ALLOW_MEDIAL_UNDERSCORES)))
         {
             ++s;
-            goto redo;
+
+            /* To get here with the value so-far being 0 means we've only had
+             * leading zeros, then an underscore.  We can continue with the
+             * branchless switch() instead of this loop */
+            if (value == 0) {
+                goto redo_switch;
+            }
+            else {
+                goto redo;
+            }
         }
 
         if (*s) {