From eeeda9a11637c8093a0efaa5cdf9675c0b6da3b9 Mon Sep 17 00:00:00 2001
From: Solomon Peachy <pizza@shaftnet.org>
Date: Wed, 26 Feb 2020 16:00:51 -0500
Subject: [PATCH] misc:  Use GCC builtins for byteswapping.

(On most CPUs this translates to single-cycle operations.  1.2% binary
size shrink on x86_64, for example)
---
 backend_common.h                   | 40 +++++------------------------
 lib6145/libS6145ImageReProcess.c   | 41 +++++-------------------------
 lib70x/libMitsuD70ImageReProcess.c | 41 +++++-------------------------
 3 files changed, 18 insertions(+), 104 deletions(-)

diff --git a/backend_common.h b/backend_common.h
index fc88b06..221d23d 100644
--- a/backend_common.h
+++ b/backend_common.h
@@ -59,41 +59,13 @@
 #define le64_to_cpu(__x) __x
 #define le32_to_cpu(__x) __x
 #define le16_to_cpu(__x) __x
-#define be16_to_cpu(__x) ((uint16_t)(			\
-	(((uint16_t)(__x) & (uint16_t)0x00ff) <<  8) |  \
-	(((uint16_t)(__x) & (uint16_t)0xff00) >>  8) ))
-#define be32_to_cpu(__x) ((uint32_t)(			     \
-	(((uint32_t)(__x) & (uint32_t)0x000000ffUL) << 24) | \
-	(((uint32_t)(__x) & (uint32_t)0x0000ff00UL) <<  8) | \
-	(((uint32_t)(__x) & (uint32_t)0x00ff0000UL) >>  8) | \
-	(((uint32_t)(__x) & (uint32_t)0xff000000UL) >> 24) ))
-#define be64_to_cpu(__x) ((uint64_t)(                         \
-        (((uint64_t)(__x) & (uint64_t)0x00000000000000ffULL) << 56) |   \
-        (((uint64_t)(__x) & (uint64_t)0x000000000000ff00ULL) << 40) |   \
-        (((uint64_t)(__x) & (uint64_t)0x0000000000ff0000ULL) << 24) |   \
-        (((uint64_t)(__x) & (uint64_t)0x00000000ff000000ULL) <<  8) |   \
-        (((uint64_t)(__x) & (uint64_t)0x000000ff00000000ULL) >>  8) |   \
-        (((uint64_t)(__x) & (uint64_t)0x0000ff0000000000ULL) >> 24) |   \
-        (((uint64_t)(__x) & (uint64_t)0x00ff000000000000ULL) >> 40) |   \
-        (((uint64_t)(__x) & (uint64_t)0xff00000000000000ULL) >> 56) ))
+#define be16_to_cpu(__x) __builtin_bswap16(__x)
+#define be32_to_cpu(__x) __builtin_bswap32(__x)
+#define be64_to_cpu(__x) __builtin_bswap64(__x)
 #else
-#define le16_to_cpu(__x) ((uint16_t)(			\
-	(((uint16_t)(__x) & (uint16_t)0x00ff) <<  8) |  \
-	(((uint16_t)(__x) & (uint16_t)0xff00) >>  8) ))
-#define le32_to_cpu(__x) ((uint32_t)(			     \
-	(((uint32_t)(__x) & (uint32_t)0x000000ffUL) << 24) | \
-	(((uint32_t)(__x) & (uint32_t)0x0000ff00UL) <<  8) | \
-	(((uint32_t)(__x) & (uint32_t)0x00ff0000UL) >>  8) | \
-	(((uint32_t)(__x) & (uint32_t)0xff000000UL) >> 24) ))
-#define le64_to_cpu(__x) ((uint64_t)(                         \
-        (((uint64_t)(__x) & (uint64_t)0x00000000000000ffULL) << 56) |   \
-        (((uint64_t)(__x) & (uint64_t)0x000000000000ff00ULL) << 40) |   \
-        (((uint64_t)(__x) & (uint64_t)0x0000000000ff0000ULL) << 24) |   \
-        (((uint64_t)(__x) & (uint64_t)0x00000000ff000000ULL) <<  8) |   \
-        (((uint64_t)(__x) & (uint64_t)0x000000ff00000000ULL) >>  8) |   \
-        (((uint64_t)(__x) & (uint64_t)0x0000ff0000000000ULL) >> 24) |   \
-        (((uint64_t)(__x) & (uint64_t)0x00ff000000000000ULL) >> 40) |   \
-        (((uint64_t)(__x) & (uint64_t)0xff00000000000000ULL) >> 56)))
+#define le16_to_cpu(__x) __builtin_bswap16(__x)
+#define le32_to_cpu(__x) __builtin_bswap32(__x)
+#define le64_to_cpu(__x) __builtin_bswap64(__x)
 #define be64_to_cpu(__x) __x
 #define be32_to_cpu(__x) __x
 #define be16_to_cpu(__x) __x
diff --git a/lib6145/libS6145ImageReProcess.c b/lib6145/libS6145ImageReProcess.c
index 2428ab6..2bb9c73 100644
--- a/lib6145/libS6145ImageReProcess.c
+++ b/lib6145/libS6145ImageReProcess.c
@@ -224,46 +224,17 @@ static void LeadEdgeCorrection(void);
 
 //-------------------------------------------------------------------------
 // Endian Manipulation macros
-
 #if (__BYTE_ORDER == __LITTLE_ENDIAN)
 #define le64_to_cpu(__x) __x
 #define le32_to_cpu(__x) __x
 #define le16_to_cpu(__x) __x
-#define be16_to_cpu(__x) ((uint16_t)(			\
-	(((uint16_t)(__x) & (uint16_t)0x00ff) <<  8) |  \
-	(((uint16_t)(__x) & (uint16_t)0xff00) >>  8) ))
-#define be32_to_cpu(__x) ((uint32_t)(			     \
-	(((uint32_t)(__x) & (uint32_t)0x000000ffUL) << 24) | \
-	(((uint32_t)(__x) & (uint32_t)0x0000ff00UL) <<  8) | \
-	(((uint32_t)(__x) & (uint32_t)0x00ff0000UL) >>  8) | \
-	(((uint32_t)(__x) & (uint32_t)0xff000000UL) >> 24) ))
-#define be64_to_cpu(__x) ((uint64_t)(                         \
-        (((uint64_t)(__x) & (uint64_t)0x00000000000000ffULL) << 56) |   \
-        (((uint64_t)(__x) & (uint64_t)0x000000000000ff00ULL) << 40) |   \
-        (((uint64_t)(__x) & (uint64_t)0x0000000000ff0000ULL) << 24) |   \
-        (((uint64_t)(__x) & (uint64_t)0x00000000ff000000ULL) <<  8) |   \
-        (((uint64_t)(__x) & (uint64_t)0x000000ff00000000ULL) >>  8) |   \
-        (((uint64_t)(__x) & (uint64_t)0x0000ff0000000000ULL) >> 24) |   \
-        (((uint64_t)(__x) & (uint64_t)0x00ff000000000000ULL) >> 40) |   \
-        (((uint64_t)(__x) & (uint64_t)0xff00000000000000ULL) >> 56) ))
+#define be16_to_cpu(__x) __builtin_bswap16(__x)
+#define be32_to_cpu(__x) __builtin_bswap32(__x)
+#define be64_to_cpu(__x) __builtin_bswap64(__x)
 #else
-#define le16_to_cpu(__x) ((uint16_t)(			\
-	(((uint16_t)(__x) & (uint16_t)0x00ff) <<  8) |  \
-	(((uint16_t)(__x) & (uint16_t)0xff00) >>  8) ))
-#define le32_to_cpu(__x) ((uint32_t)(			     \
-	(((uint32_t)(__x) & (uint32_t)0x000000ffUL) << 24) | \
-	(((uint32_t)(__x) & (uint32_t)0x0000ff00UL) <<  8) | \
-	(((uint32_t)(__x) & (uint32_t)0x00ff0000UL) >>  8) | \
-	(((uint32_t)(__x) & (uint32_t)0xff000000UL) >> 24) ))
-#define le64_to_cpu(__x) ((uint64_t)(                         \
-        (((uint64_t)(__x) & (uint64_t)0x00000000000000ffULL) << 56) |   \
-        (((uint64_t)(__x) & (uint64_t)0x000000000000ff00ULL) << 40) |   \
-        (((uint64_t)(__x) & (uint64_t)0x0000000000ff0000ULL) << 24) |   \
-        (((uint64_t)(__x) & (uint64_t)0x00000000ff000000ULL) <<  8) |   \
-        (((uint64_t)(__x) & (uint64_t)0x000000ff00000000ULL) >>  8) |   \
-        (((uint64_t)(__x) & (uint64_t)0x0000ff0000000000ULL) >> 24) |   \
-        (((uint64_t)(__x) & (uint64_t)0x00ff000000000000ULL) >> 40) |   \
-        (((uint64_t)(__x) & (uint64_t)0xff00000000000000ULL) >> 56)))
+#define le16_to_cpu(__x) __builtin_bswap16(__x)
+#define le32_to_cpu(__x) __builtin_bswap32(__x)
+#define le64_to_cpu(__x) __builtin_bswap64(__x)
 #define be64_to_cpu(__x) __x
 #define be32_to_cpu(__x) __x
 #define be16_to_cpu(__x) __x
diff --git a/lib70x/libMitsuD70ImageReProcess.c b/lib70x/libMitsuD70ImageReProcess.c
index fba91aa..2acc214 100644
--- a/lib70x/libMitsuD70ImageReProcess.c
+++ b/lib70x/libMitsuD70ImageReProcess.c
@@ -65,46 +65,17 @@
 
 //-------------------------------------------------------------------------
 // Endian Manipulation macros
-
 #if (__BYTE_ORDER == __LITTLE_ENDIAN)
 #define le64_to_cpu(__x) __x
 #define le32_to_cpu(__x) __x
 #define le16_to_cpu(__x) __x
-#define be16_to_cpu(__x) ((uint16_t)(			\
-	(((uint16_t)(__x) & (uint16_t)0x00ff) <<  8) |  \
-	(((uint16_t)(__x) & (uint16_t)0xff00) >>  8) ))
-#define be32_to_cpu(__x) ((uint32_t)(			     \
-	(((uint32_t)(__x) & (uint32_t)0x000000ffUL) << 24) | \
-	(((uint32_t)(__x) & (uint32_t)0x0000ff00UL) <<  8) | \
-	(((uint32_t)(__x) & (uint32_t)0x00ff0000UL) >>  8) | \
-	(((uint32_t)(__x) & (uint32_t)0xff000000UL) >> 24) ))
-#define be64_to_cpu(__x) ((uint64_t)(                         \
-        (((uint64_t)(__x) & (uint64_t)0x00000000000000ffULL) << 56) |   \
-        (((uint64_t)(__x) & (uint64_t)0x000000000000ff00ULL) << 40) |   \
-        (((uint64_t)(__x) & (uint64_t)0x0000000000ff0000ULL) << 24) |   \
-        (((uint64_t)(__x) & (uint64_t)0x00000000ff000000ULL) <<  8) |   \
-        (((uint64_t)(__x) & (uint64_t)0x000000ff00000000ULL) >>  8) |   \
-        (((uint64_t)(__x) & (uint64_t)0x0000ff0000000000ULL) >> 24) |   \
-        (((uint64_t)(__x) & (uint64_t)0x00ff000000000000ULL) >> 40) |   \
-        (((uint64_t)(__x) & (uint64_t)0xff00000000000000ULL) >> 56) ))
+#define be16_to_cpu(__x) __builtin_bswap16(__x)
+#define be32_to_cpu(__x) __builtin_bswap32(__x)
+#define be64_to_cpu(__x) __builtin_bswap64(__x)
 #else
-#define le16_to_cpu(__x) ((uint16_t)(			\
-	(((uint16_t)(__x) & (uint16_t)0x00ff) <<  8) |  \
-	(((uint16_t)(__x) & (uint16_t)0xff00) >>  8) ))
-#define le32_to_cpu(__x) ((uint32_t)(			     \
-	(((uint32_t)(__x) & (uint32_t)0x000000ffUL) << 24) | \
-	(((uint32_t)(__x) & (uint32_t)0x0000ff00UL) <<  8) | \
-	(((uint32_t)(__x) & (uint32_t)0x00ff0000UL) >>  8) | \
-	(((uint32_t)(__x) & (uint32_t)0xff000000UL) >> 24) ))
-#define le64_to_cpu(__x) ((uint64_t)(                         \
-        (((uint64_t)(__x) & (uint64_t)0x00000000000000ffULL) << 56) |   \
-        (((uint64_t)(__x) & (uint64_t)0x000000000000ff00ULL) << 40) |   \
-        (((uint64_t)(__x) & (uint64_t)0x0000000000ff0000ULL) << 24) |   \
-        (((uint64_t)(__x) & (uint64_t)0x00000000ff000000ULL) <<  8) |   \
-        (((uint64_t)(__x) & (uint64_t)0x000000ff00000000ULL) >>  8) |   \
-        (((uint64_t)(__x) & (uint64_t)0x0000ff0000000000ULL) >> 24) |   \
-        (((uint64_t)(__x) & (uint64_t)0x00ff000000000000ULL) >> 40) |   \
-        (((uint64_t)(__x) & (uint64_t)0xff00000000000000ULL) >> 56)))
+#define le16_to_cpu(__x) __builtin_bswap16(__x)
+#define le32_to_cpu(__x) __builtin_bswap32(__x)
+#define le64_to_cpu(__x) __builtin_bswap64(__x)
 #define be64_to_cpu(__x) __x
 #define be32_to_cpu(__x) __x
 #define be16_to_cpu(__x) __x