summaryrefslogtreecommitdiffstats
path: root/include/asm-arm26/unaligned.h
diff options
context:
space:
mode:
Diffstat (limited to 'include/asm-arm26/unaligned.h')
-rw-r--r--include/asm-arm26/unaligned.h118
1 files changed, 118 insertions, 0 deletions
diff --git a/include/asm-arm26/unaligned.h b/include/asm-arm26/unaligned.h
new file mode 100644
index 00000000000..d992782089f
--- /dev/null
+++ b/include/asm-arm26/unaligned.h
@@ -0,0 +1,118 @@
+#ifndef __ASM_ARM_UNALIGNED_H
+#define __ASM_ARM_UNALIGNED_H
+
+#include <asm/types.h>
+
+extern int __bug_unaligned_x(void *ptr);
+
+/*
+ * What is the most efficient way of loading/storing an unaligned value?
+ *
+ * That is the subject of this file. Efficiency here is defined as
+ * minimum code size with minimum register usage for the common cases.
+ * It is currently not believed that long longs are common, so we
+ * trade efficiency for the chars, shorts and longs against the long
+ * longs.
+ *
+ * Current stats with gcc 2.7.2.2 for these functions:
+ *
+ * ptrsize get: code regs put: code regs
+ * 1 1 1 1 2
+ * 2 3 2 3 2
+ * 4 7 3 7 3
+ * 8 20 6 16 6
+ *
+ * gcc 2.95.1 seems to code differently:
+ *
+ * ptrsize get: code regs put: code regs
+ * 1 1 1 1 2
+ * 2 3 2 3 2
+ * 4 7 4 7 4
+ * 8 19 8 15 6
+ *
+ * which may or may not be more efficient (depending upon whether
+ * you can afford the extra registers). Hopefully the gcc 2.95
+ * is inteligent enough to decide if it is better to use the
+ * extra register, but evidence so far seems to suggest otherwise.
+ *
+ * Unfortunately, gcc is not able to optimise the high word
+ * out of long long >> 32, or the low word from long long << 32
+ */
+
+#define __get_unaligned_2_le(__p) \
+ (__p[0] | __p[1] << 8)
+
+#define __get_unaligned_4_le(__p) \
+ (__p[0] | __p[1] << 8 | __p[2] << 16 | __p[3] << 24)
+
+#define __get_unaligned_le(ptr) \
+ ({ \
+ __typeof__(*(ptr)) __v; \
+ __u8 *__p = (__u8 *)(ptr); \
+ switch (sizeof(*(ptr))) { \
+ case 1: __v = *(ptr); break; \
+ case 2: __v = __get_unaligned_2_le(__p); break; \
+ case 4: __v = __get_unaligned_4_le(__p); break; \
+ case 8: { \
+ unsigned int __v1, __v2; \
+ __v2 = __get_unaligned_4_le((__p+4)); \
+ __v1 = __get_unaligned_4_le(__p); \
+ __v = ((unsigned long long)__v2 << 32 | __v1); \
+ } \
+ break; \
+ default: __v = __bug_unaligned_x(__p); break; \
+ } \
+ __v; \
+ })
+
+static inline void __put_unaligned_2_le(__u32 __v, register __u8 *__p)
+{
+ *__p++ = __v;
+ *__p++ = __v >> 8;
+}
+
+static inline void __put_unaligned_4_le(__u32 __v, register __u8 *__p)
+{
+ __put_unaligned_2_le(__v >> 16, __p + 2);
+ __put_unaligned_2_le(__v, __p);
+}
+
+static inline void __put_unaligned_8_le(const unsigned long long __v, register __u8 *__p)
+{
+ /*
+ * tradeoff: 8 bytes of stack for all unaligned puts (2
+ * instructions), or an extra register in the long long
+ * case - go for the extra register.
+ */
+ __put_unaligned_4_le(__v >> 32, __p+4);
+ __put_unaligned_4_le(__v, __p);
+}
+
+/*
+ * Try to store an unaligned value as efficiently as possible.
+ */
+#define __put_unaligned_le(val,ptr) \
+ ({ \
+ switch (sizeof(*(ptr))) { \
+ case 1: \
+ *(ptr) = (val); \
+ break; \
+ case 2: __put_unaligned_2_le((val),(__u8 *)(ptr)); \
+ break; \
+ case 4: __put_unaligned_4_le((val),(__u8 *)(ptr)); \
+ break; \
+ case 8: __put_unaligned_8_le((val),(__u8 *)(ptr)); \
+ break; \
+ default: __bug_unaligned_x(ptr); \
+ break; \
+ } \
+ (void) 0; \
+ })
+
+/*
+ * Select endianness
+ */
+#define get_unaligned __get_unaligned_le
+#define put_unaligned __put_unaligned_le
+
+#endif