From 7f9ebf8e86e90ea1eb68388f1be12503a3f31e56 Mon Sep 17 00:00:00 2001
From: Jean-Francois Romang <jeanfrancois.romang@gmail.com>
Date: Thu, 11 Oct 2012 22:55:25 +0800
Subject: [PATCH] ARM lsb/msb assembly

Implement lsb/msb using armv7 assembly instructions.
msb is the easiest one, using a gcc intrinsic that generates
code using the ARM's clz instruction. lsb is also using this
clz instruction, but with the help of ARM's 'rbit' (bit
reversing) instruction. This leads to a >2% speed gain.

I also renamed 'arm-32' to the more meaningfull 'armv7' in the Makefile

No functional change.
---
 src/Makefile   | 14 +++++++-------
 src/bitboard.h | 15 +++++++++++++++
 2 files changed, 22 insertions(+), 7 deletions(-)

diff --git a/src/Makefile b/src/Makefile
index c73b94f3..17e61773 100644
--- a/src/Makefile
+++ b/src/Makefile
@@ -126,12 +126,12 @@ ifeq ($(ARCH),x86-32-old)
 endif
 
 #arm section
-ifeq ($(ARCH),arm-32)
-	arch = arm-32
+ifeq ($(ARCH),armv7)
+	arch = armv7
 	os = any
 	bits = 32
 	prefetch = yes
-	bsfq = no
+	bsfq = yes
 	popcnt = no
 endif
 
@@ -276,7 +276,7 @@ ifeq ($(optimize),yes)
 			endif
 		endif
 
-		ifeq ($(arch),arm-32)
+		ifeq ($(arch),armv7)
 			CXXFLAGS += -fno-gcse
 		endif
 	endif
@@ -315,7 +315,7 @@ endif
 
 ### 3.7 prefetch
 ifeq ($(prefetch),yes)
-	ifneq ($(arch),arm-32)
+	ifneq ($(arch),armv7)
 		CXXFLAGS += -msse
 		DEPENDFLAGS += -msse
 	endif
@@ -376,7 +376,7 @@ help:
 	@echo "osx-ppc-32           > PPC-Mac OS X 32 bit"
 	@echo "osx-x86-64           > x86-Mac OS X 64 bit"
 	@echo "osx-x86-32           > x86-Mac OS X 32 bit"
-	@echo "arm-32               > ARM 32 bit"
+	@echo "armv7                > ARMv7 32 bit"
 	@echo "general-64           > unspecified 64-bit"
 	@echo "general-32           > unspecified 32-bit"
 	@echo ""
@@ -466,7 +466,7 @@ config-sanity:
 	@test "$(debug)" = "yes" || test "$(debug)" = "no"
 	@test "$(optimize)" = "yes" || test "$(optimize)" = "no"
 	@test "$(arch)" = "any" || test "$(arch)" = "x86_64" || test "$(arch)" = "i386" || \
-	 test "$(arch)" = "ppc64" || test "$(arch)" = "ppc" || test "$(arch)" = "arm-32"
+	 test "$(arch)" = "ppc64" || test "$(arch)" = "ppc" || test "$(arch)" = "armv7"
 	@test "$(os)" = "any" || test "$(os)" = "osx"
 	@test "$(bits)" = "32" || test "$(bits)" = "64"
 	@test "$(prefetch)" = "yes" || test "$(prefetch)" = "no"
diff --git a/src/bitboard.h b/src/bitboard.h
index 4c7c948a..6aa35874 100644
--- a/src/bitboard.h
+++ b/src/bitboard.h
@@ -247,6 +247,21 @@ FORCE_INLINE Square msb(Bitboard b) {
   return (Square) index;
 }
 
+#  elif defined(__arm__)
+
+FORCE_INLINE int lsb32(uint32_t v) {
+  __asm__("rbit %0, %1" : "=r"(v) : "r"(v));
+  return __builtin_clz(v);
+}
+
+FORCE_INLINE Square msb(Bitboard b) {
+  return (Square) (63 - __builtin_clzll(b));
+}
+
+FORCE_INLINE Square lsb(Bitboard b) {
+  return (Square) (uint32_t(b) ? lsb32(uint32_t(b)) : 32 + lsb32(uint32_t(b >> 32)));
+}
+
 #  else
 
 FORCE_INLINE Square lsb(Bitboard b) { // Assembly code by Heinz van Saanen
-- 
2.39.2