* string.c (count_utf8_lead_bytes_with_word): Use __builtin_popcount

  only if it can use SSE 4.2 POPCNT whose latency is 3 cycle.

* internal.h (rb_popcount64): use __builtin_popcountll because now
  it is in fast path.

git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@54894 b2dd03c8-39d4-4d8f-98ff-823fe69b080e

4 files changed, 35 insertions, 4 deletions

diff --git a/ChangeLog b/ChangeLog
index 267b3433c0..9312462438 100644
--- a/ChangeLog
+++ b/ChangeLog
@@ -1,3 +1,11 @@
+Tue May  3 22:10:09 2016  NARUSE, Yui  <naruse@ruby-lang.org>
+
+	* string.c (count_utf8_lead_bytes_with_word): Use __builtin_popcount
+	  only if it can use SSE 4.2 POPCNT whose latency is 3 cycle.
+
+	* internal.h (rb_popcount64): use __builtin_popcountll because now
+	  it is in fast path.
+
 Tue May  3 14:19:18 2016  Nobuyoshi Nakada  <nobu@ruby-lang.org>
 
 	* parse.y (new_if_gen): set newline flag to NODE_IF to trace all
diff --git a/configure.in b/configure.in
index 14711a6837..ad8cc988c4 100644
--- a/configure.in
+++ b/configure.in
@@ -2429,6 +2429,8 @@ fi])
 RUBY_CHECK_BUILTIN_FUNC(__builtin_bswap16, [__builtin_bswap16(0)])
 RUBY_CHECK_BUILTIN_FUNC(__builtin_bswap32, [__builtin_bswap32(0)])
 RUBY_CHECK_BUILTIN_FUNC(__builtin_bswap64, [__builtin_bswap64(0)])
+RUBY_CHECK_BUILTIN_FUNC(__builtin_popcount, [__builtin_popcount(0)])
+RUBY_CHECK_BUILTIN_FUNC(__builtin_popcountll, [__builtin_popcountll(0)])
 RUBY_CHECK_BUILTIN_FUNC(__builtin_clz, [__builtin_clz(0)])
 RUBY_CHECK_BUILTIN_FUNC(__builtin_clzl, [__builtin_clzl(0)])
 RUBY_CHECK_BUILTIN_FUNC(__builtin_clzll, [__builtin_clzll(0)])
diff --git a/internal.h b/internal.h
index 0bac79e012..765758a962 100644
--- a/internal.h
+++ b/internal.h
@@ -271,21 +271,38 @@ nlz_intptr(uintptr_t x) {
 
 static inline int
 rb_popcount32(uint32_t x) {
+#ifdef HAVE_BUILTIN___BUILTIN_POPCOUNT
+    return __builtin_popcount(x);
+#else
     x = (x & 0x55555555) + (x >> 1 & 0x55555555);
     x = (x & 0x33333333) + (x >> 2 & 0x33333333);
     x = (x & 0x0f0f0f0f) + (x >> 4 & 0x0f0f0f0f);
     x = (x & 0x001f001f) + (x >> 8 & 0x001f001f);
     return (x & 0x0000003f) + (x >>16 & 0x0000003f);
+#endif
 }
 
 static inline int
 rb_popcount64(uint64_t x) {
+#ifdef HAVE_BUILTIN___BUILTIN_POPCOUNT
+    return __builtin_popcountll(x);
+#else
     x = (x & 0x5555555555555555) + (x >> 1 & 0x5555555555555555);
     x = (x & 0x3333333333333333) + (x >> 2 & 0x3333333333333333);
     x = (x & 0x0707070707070707) + (x >> 4 & 0x0707070707070707);
     x = (x & 0x001f001f001f001f) + (x >> 8 & 0x001f001f001f001f);
     x = (x & 0x0000003f0000003f) + (x >>16 & 0x0000003f0000003f);
     return (x & 0x7f) + (x >>32 & 0x7f);
+#endif
+}
+
+static inline int
+rb_popcount_intptr(uintptr_t x) {
+#if SIZEOF_VOIDP == 8
+    return rb_popcount64(x);
+#elif SIZEOF_VOIDP == 4
+    return rb_popcount32(x);
+#endif
 }
 
 static inline int
diff --git a/string.c b/string.c
index 35e53f7306..9ccf0933e2 100644
--- a/string.c
+++ b/string.c
@@ -1476,17 +1476,21 @@ count_utf8_lead_bytes_with_word(const uintptr_t *s)
     uintptr_t d = *s;
 
     /* Transform so that bit0 indicates whether we have a UTF-8 leading byte or not. */
-    d |= ~(d>>1);
-    d >>= 6;
+    d = (d>>6) | (~d>>7);
     d &= NONASCII_MASK >> 7;
 
     /* Gather all bytes. */
+#if defined(HAVE_BUILTIN___BUILTIN_POPCOUNT) && defined(__POPCNT__)
+    /* use only if it can use POPCNT */
+    return rb_popcount_intptr(d);
+#else
     d += (d>>8);
     d += (d>>16);
-#if SIZEOF_VOIDP == 8
+# if SIZEOF_VOIDP == 8
     d += (d>>32);
-#endif
+# endif
     return (d&0xF);
+#endif
 }
 #endif