#include <cfg/compiler.h>
#include <cpu/attr.h>
+#include <cpu/detect.h>
+#include <cpu/types.h>
+#include <cfg/macros.h>
/**
* Swap upper and lower bytes in a 16-bit value.
#define SWAB16(x) ((uint16_t)(ROTR((x), 8) + \
STATIC_ASSERT_EXPR(sizeof(x) == sizeof(uint16_t))))
-#if GNUC_PREREQ(4, 3)
+/*
+ * On Cortex-M3, GCC 4.4 builtin implementation is slower than our own
+ * rot-based implementation.
+ */
+#if GNUC_PREREQ(4, 3) && !CPU_CM3
#define SWAB32(x) ((uint32_t)(__builtin_bswap32((x) + \
STATIC_ASSERT_EXPR(sizeof(x) == sizeof(uint32_t)))))
-#define SWAB64(x) ((uint64_t)(__builtin_bswap64((x) + \
- STATIC_ASSERT_EXPR(sizeof(x) == sizeof(uint64_t)))))
#else
/**
* Reverse bytes in a 32-bit value (e.g.: 0x12345678 -> 0x78563412).
*/
-#define SWAB32(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) + \
- STATIC_ASSERT_EXPR(sizeof(x) == sizeof(uint32_t))))
+#define SWAB32(x) ((uint32_t)(( \
+ (ROTR(x, 8) & 0xFF00FF00) | \
+ (ROTL(x, 8) & 0x00FF00FF))) + \
+ STATIC_ASSERT_EXPR(sizeof(x) == sizeof(uint32_t)))
+#endif
+#if GNUC_PREREQ(4, 3)
+#define SWAB64(x) ((uint64_t)(__builtin_bswap64((x) + \
+ STATIC_ASSERT_EXPR(sizeof(x) == sizeof(uint64_t)))))
+#else
/**
* Reverse bytes in a 64-bit value.
*/
/* Avoid breaking strict aliasing rules. */
char *cx = (char *)(&x);
STATIC_ASSERT(sizeof(float) == 4);
- #define BYTEORDER_SWAP(a, b) ((a) ^= (b) ^= (a) ^= (b))
+ #define BYTEORDER_SWAP(a, b) do { (a) ^= (b); (b) ^= (a); (a) ^= (b); } while(0)
BYTEORDER_SWAP(cx[0], cx[3]);
BYTEORDER_SWAP(cx[1], cx[2]);
#undef BYTEORDER_SWAP
return be_float_to_cpu(x);
}
+#if CPU_ARM
+INLINE cpu_atomic_t
+cpu_atomic_xchg(volatile cpu_atomic_t *ptr, cpu_atomic_t val)
+{
+ cpu_atomic_t ret;
+
+ asm volatile(
+ "swp %0, %1, [%2]"
+
+ : "=&r" (ret)
+ : "r" (val), "r" (ptr)
+ : "memory", "cc");
+
+ return ret;
+}
+#else /* CPU_ARM */
+#include <cpu/irq.h>
+
+INLINE cpu_atomic_t
+cpu_atomic_xchg(volatile cpu_atomic_t *ptr, cpu_atomic_t val)
+{
+ cpu_atomic_t ret;
+
+ ATOMIC(
+ ret = *ptr;
+ *ptr = val;
+ );
+ return ret;
+}
+#endif /* CPU_ARM */
+
#ifdef __cplusplus
/// Type generic byte swapping.