Keep track of the first non-ascii word/char to avoid redoing the work.

src/heap.cc

Index: src/heap.cc
diff --git a/src/heap.cc b/src/heap.cc
index 4fd1a041eabfc73b4fa6ed112c0970447e163c06..afbc192b161c8384d7f14bb844c2da71bfcc99e1 100644
--- a/src/heap.cc
+++ b/src/heap.cc
@@ -4428,13 +4428,14 @@ MaybeObject* Heap::AllocateStringFromAscii(Vector<const char> string,
 
 
 MaybeObject* Heap::AllocateStringFromUtf8Slow(Vector<const char> string,
+                                              const char* non_ascii_start,

[Yang, 2012/10/18 14:46:23]

Maybe it's cleaner to pass non_ascii_start as string index and not as an address
pointing into the backing store? You need that index anyways.

[Toon Verwaest, 2012/10/18 15:06:56]

Done.

                                               PretenureFlag pretenure) {
-  // Count the number of characters in the UTF-8 string and check if
-  // it is an ASCII string.
+  // Continue counting the number of characters in the UTF-8 string, starting
+  // from the first non-ascii character or word.
+  int chars = non_ascii_start - string.start();
   Access<UnicodeCache::Utf8Decoder>
       decoder(isolate_->unicode_cache()->utf8_decoder());
-  decoder->Reset(string.start(), string.length());
-  int chars = 0;
+  decoder->Reset(non_ascii_start, string.length() - chars);
   while (decoder->has_more()) {
     uint32_t r = decoder->GetNext();
     if (r <= unibrow::Utf16::kMaxNonSurrogateCharCode) {