"fast unaligned" bytestream functions
[ffmpeg.git] / libavutil / intreadwrite.h
index 0e11598..b59c481 100644 (file)
@@ -44,40 +44,89 @@ struct unaligned_16 { uint16_t l; } __attribute__((packed));
 #endif /* !__GNUC__ */
 
 /* endian macros */
-#define AV_RB8(x)  (((uint8_t*)(x))[0])
+#define AV_RB8(x)     (((uint8_t*)(x))[0])
 #define AV_WB8(p, d)  { ((uint8_t*)(p))[0] = (d); }
 
-#define AV_RB16(x) ((((uint8_t*)(x))[0] << 8) | ((uint8_t*)(x))[1])
+#define AV_RL8(x)     AV_RB8(x)
+#define AV_WL8(p, d)  AV_WB8(p, d)
+
+#ifdef HAVE_FAST_UNALIGNED
+# ifdef WORDS_BIGENDIAN
+#  define AV_RB16(x)    LD16(x)
+#  define AV_WB16(p, d) ST16(p, d)
+
+#  define AV_RL16(x)    bswap_16(LD16(x))
+#  define AV_WL16(p, d) ST16(p, bswap_16(d))
+# else /* WORDS_BIGENDIAN */
+#  define AV_RB16(x)    bswap_16(LD16(x))
+#  define AV_WB16(p, d) ST16(p, bswap_16(d))
+
+#  define AV_RL16(x)    LD16(x)
+#  define AV_WL16(p, d) ST16(p, d)
+# endif
+#else /* HAVE_FAST_UNALIGNED */
+#define AV_RB16(x)  ((((uint8_t*)(x))[0] << 8) | ((uint8_t*)(x))[1])
 #define AV_WB16(p, d) { \
                     ((uint8_t*)(p))[1] = (d); \
                     ((uint8_t*)(p))[0] = (d)>>8; }
 
-#define AV_RB32(x) ((((uint8_t*)(x))[0] << 24) | \
-                   (((uint8_t*)(x))[1] << 16) | \
-                   (((uint8_t*)(x))[2] << 8) | \
-                    ((uint8_t*)(x))[3])
+#define AV_RL16(x)  ((((uint8_t*)(x))[1] << 8) | \
+                      ((uint8_t*)(x))[0])
+#define AV_WL16(p, d) { \
+                    ((uint8_t*)(p))[0] = (d); \
+                    ((uint8_t*)(p))[1] = (d)>>8; }
+#endif
+
+#define AV_RB24(x)  ((((uint8_t*)(x))[0] << 16) | \
+                     (((uint8_t*)(x))[1] <<  8) | \
+                      ((uint8_t*)(x))[2])
+#define AV_WB24(p, d) { \
+                    ((uint8_t*)(p))[2] = (d); \
+                    ((uint8_t*)(p))[1] = (d)>>8; \
+                    ((uint8_t*)(p))[0] = (d)>>16; }
+
+#define AV_RL24(x)  ((((uint8_t*)(x))[2] << 16) | \
+                     (((uint8_t*)(x))[1] <<  8) | \
+                      ((uint8_t*)(x))[0])
+#define AV_WL24(p, d) { \
+                    ((uint8_t*)(p))[0] = (d); \
+                    ((uint8_t*)(p))[1] = (d)>>8; \
+                    ((uint8_t*)(p))[2] = (d)>>16; }
+
+#ifdef HAVE_FAST_UNALIGNED
+# ifdef WORDS_BIGENDIAN
+#  define AV_RB32(x)    LD32(x)
+#  define AV_WB32(p, d) ST32(p, d)
+
+#  define AV_RL32(x)    bswap_32(LD32(x))
+#  define AV_WL32(p, d) ST32(p, bswap_32(d))
+# else /* WORDS_BIGENDIAN */
+#  define AV_RB32(x)    bswap_32(LD32(x))
+#  define AV_WB32(p, d) ST32(p, bswap_32(d))
+
+#  define AV_RL32(x)    LD32(x)
+#  define AV_WL32(p, d) ST32(p, d)
+# endif
+#else /* HAVE_FAST_UNALIGNED */
+#define AV_RB32(x)  ((((uint8_t*)(x))[0] << 24) | \
+                     (((uint8_t*)(x))[1] << 16) | \
+                     (((uint8_t*)(x))[2] <<  8) | \
+                      ((uint8_t*)(x))[3])
 #define AV_WB32(p, d) { \
                     ((uint8_t*)(p))[3] = (d); \
                     ((uint8_t*)(p))[2] = (d)>>8; \
                     ((uint8_t*)(p))[1] = (d)>>16; \
                     ((uint8_t*)(p))[0] = (d)>>24; }
 
-#define AV_RL8(x)  AV_RB8(x)
-#define AV_WL8(p, d)  AV_WB8(p, d)
-
-#define AV_RL16(x) ((((uint8_t*)(x))[1] << 8) | ((uint8_t*)(x))[0])
-#define AV_WL16(p, d) { \
-                    ((uint8_t*)(p))[0] = (d); \
-                    ((uint8_t*)(p))[1] = (d)>>8; }
-
 #define AV_RL32(x) ((((uint8_t*)(x))[3] << 24) | \
-                   (((uint8_t*)(x))[2] << 16) | \
-                   (((uint8_t*)(x))[1] << 8) | \
-                    ((uint8_t*)(x))[0])
+                    (((uint8_t*)(x))[2] << 16) | \
+                    (((uint8_t*)(x))[1] <<  8) | \
+                     ((uint8_t*)(x))[0])
 #define AV_WL32(p, d) { \
                     ((uint8_t*)(p))[0] = (d); \
                     ((uint8_t*)(p))[1] = (d)>>8; \
                     ((uint8_t*)(p))[2] = (d)>>16; \
                     ((uint8_t*)(p))[3] = (d)>>24; }
+#endif
 
 #endif /* INTREADWRITE_H */