X-Git-Url: https://wimlib.net/git/?a=blobdiff_plain;f=include%2Fwimlib%2Funaligned.h;h=d25e76f6d909fb1e61ccd507956d12e56896daff;hb=4109848cdf4c31d8ea1a410b183478ad4124f15b;hp=ccc86f016a3ab9a1638e49a1bf42891e437d0a8e;hpb=0ecb0529b5fcacc1abafa1f3f02a40c44783ada8;p=wimlib

diff --git a/include/wimlib/unaligned.h b/include/wimlib/unaligned.h
index ccc86f01..d25e76f6 100644
--- a/include/wimlib/unaligned.h
+++ b/include/wimlib/unaligned.h
@@ -3,6 +3,9 @@
  *
  * Inline functions for unaligned memory accesses.
  *
+ * Author:	Eric Biggers
+ * Year:	2014, 2015
+ *
  * The author dedicates this file to the public domain.
  * You can do whatever you want with this file.
  */
@@ -46,84 +49,85 @@ DEFINE_UNALIGNED_TYPE(machine_word_t);
 #define load_word_unaligned	load_machine_word_t_unaligned
 #define store_word_unaligned	store_machine_word_t_unaligned
 
-static inline void
-copy_word_unaligned(const void *src, void *dst)
-{
-	store_word_unaligned(load_word_unaligned(src), dst);
-}
-
-static inline machine_word_t
-repeat_byte(u8 b)
-{
-	machine_word_t v;
-
-	BUILD_BUG_ON(WORDSIZE != 4 && WORDSIZE != 8);
-
-	v = b;
-	v |= v << 8;
-	v |= v << 16;
-	v |= v << ((WORDSIZE == 8) ? 32 : 0);
-	return v;
-}
-
 static inline u16
-get_unaligned_u16_le(const void *p)
+get_unaligned_le16(const u8 *p)
 {
-	u16 v;
-
-	if (UNALIGNED_ACCESS_IS_FAST) {
-		v = le16_to_cpu(load_le16_unaligned(p));
-	} else {
-		const u8 *p8 = p;
-		v = 0;
-		v |= (u16)p8[0] << 0;
-		v |= (u16)p8[1] << 8;
-	}
-	return v;
+	if (UNALIGNED_ACCESS_IS_FAST)
+		return le16_to_cpu(load_le16_unaligned(p));
+	else
+		return ((u16)p[1] << 8) | p[0];
 }
 
 static inline u32
-get_unaligned_u32_le(const void *p)
+get_unaligned_le32(const u8 *p)
 {
-	u32 v;
-
-	if (UNALIGNED_ACCESS_IS_FAST) {
-		v = le32_to_cpu(load_le32_unaligned(p));
-	} else {
-		const u8 *p8 = p;
-		v = 0;
-		v |= (u32)p8[0] << 0;
-		v |= (u32)p8[1] << 8;
-		v |= (u32)p8[2] << 16;
-		v |= (u32)p8[3] << 24;
-	}
-	return v;
+	if (UNALIGNED_ACCESS_IS_FAST)
+		return le32_to_cpu(load_le32_unaligned(p));
+	else
+		return ((u32)p[3] << 24) | ((u32)p[2] << 16) |
+			((u32)p[1] << 8) | p[0];
 }
 
 static inline void
-put_unaligned_u16_le(u16 v, void *p)
+put_unaligned_le16(u16 v, u8 *p)
 {
 	if (UNALIGNED_ACCESS_IS_FAST) {
 		store_le16_unaligned(cpu_to_le16(v), p);
 	} else {
-		u8 *p8 = p;
-		p8[0] = (v >> 0) & 0xFF;
-		p8[1] = (v >> 8) & 0xFF;
+		p[0] = (u8)(v >> 0);
+		p[1] = (u8)(v >> 8);
 	}
 }
 
 static inline void
-put_unaligned_u32_le(u32 v, void *p)
+put_unaligned_le32(u32 v, u8 *p)
 {
 	if (UNALIGNED_ACCESS_IS_FAST) {
 		store_le32_unaligned(cpu_to_le32(v), p);
 	} else {
-		u8 *p8 = p;
-		p8[0] = (v >> 0) & 0xFF;
-		p8[1] = (v >> 8) & 0xFF;
-		p8[2] = (v >> 16) & 0xFF;
-		p8[3] = (v >> 24) & 0xFF;
+		p[0] = (u8)(v >> 0);
+		p[1] = (u8)(v >> 8);
+		p[2] = (u8)(v >> 16);
+		p[3] = (u8)(v >> 24);
 	}
 }
 
+/*
+ * Given a 32-bit value that was loaded with the platform's native endianness,
+ * return a 32-bit value whose high-order 8 bits are 0 and whose low-order 24
+ * bits contain the first 3 bytes, arranged in octets in a platform-dependent
+ * order, at the memory location from which the input 32-bit value was loaded.
+ */
+static inline u32
+loaded_u32_to_u24(u32 v)
+{
+	if (CPU_IS_LITTLE_ENDIAN)
+		return v & 0xFFFFFF;
+	else
+		return v >> 8;
+}
+
+/*
+ * Load the next 3 bytes from the memory location @p into the 24 low-order bits
+ * of a 32-bit value.  The order in which the 3 bytes will be arranged as octets
+ * in the 24 bits is platform-dependent.  At least LOAD_U24_REQUIRED_NBYTES
+ * bytes must be available at @p; note that this may be more than 3.
+ */
+static inline u32
+load_u24_unaligned(const u8 *p)
+{
+#if UNALIGNED_ACCESS_IS_FAST
+#  define LOAD_U24_REQUIRED_NBYTES 4
+	return loaded_u32_to_u24(load_u32_unaligned(p));
+#else
+#  define LOAD_U24_REQUIRED_NBYTES 3
+#  if CPU_IS_BIG_ENDIAN
+	return ((u32)p[2] << 0) | ((u32)p[1] << 8) | ((u32)p[0] << 16);
+#  else
+	return ((u32)p[0] << 0) | ((u32)p[1] << 8) | ((u32)p[2] << 16);
+#  endif
+#endif
+}
+
+
 #endif /* _WIMLIB_UNALIGNED_H */