From 2afee60a7c37aaf47ffdb3b59dee41b2af95ed2a Mon Sep 17 00:00:00 2001
From: skal <pascal.massimino@gmail.com>
Date: Mon, 10 Sep 2012 13:13:51 +0200
Subject: [PATCH] speed up for ARM using 8bit for boolean decoder

SBITS=8 is reported 20-30% faster on ARM (where 64bit ops
are expensive).

Also use 32bits for i32.

Change-Id: Id6a7197d805061aeb8832f20432512d0d930ebfa
---
 src/utils/bit_reader.h | 19 ++++++++++++++++---
 1 file changed, 16 insertions(+), 3 deletions(-)

diff --git a/src/utils/bit_reader.h b/src/utils/bit_reader.h
index 1ef501a5..7ac5f320 100644
--- a/src/utils/bit_reader.h
+++ b/src/utils/bit_reader.h
@@ -24,7 +24,20 @@
 extern "C" {
 #endif
 
-#define BITS 32     // can be 32, 16 or 8
+//------------------------------------------------------------------------------
+// BITS can be either 32, 16 or 8. Pick values that fit natural register size.
+
+#if defined(__i386__) || defined(_M_IX86)      // x86 32bit
+#define BITS 16
+#elif defined(__arm__) || defined(_M_ARM)     // ARM
+#define BITS 8
+#else                      // reasonable default
+#define BITS 32
+#endif
+
+//------------------------------------------------------------------------------
+// Derived types and constants
+
 #define MASK ((((bit_t)1) << (BITS)) - 1)
 #if (BITS == 32)
 typedef uint64_t bit_t;   // natural register type
@@ -38,7 +51,7 @@ typedef uint8_t lbit_t;
 #endif
 
 //------------------------------------------------------------------------------
-// Bitreader and code-tree reader
+// Bitreader
 
 typedef struct VP8BitReader VP8BitReader;
 struct VP8BitReader {
@@ -151,7 +164,7 @@ static WEBP_INLINE int VP8GetSigned(VP8BitReader* const br, int v) {
 
 
 // -----------------------------------------------------------------------------
-// Bitreader
+// Bitreader for lossless format
 
 typedef struct {
   uint64_t       val_;