CPU: introduce cpu_atomic_xchg()

[bertos.git] / bertos / cpu / byteorder.h

diff --git a/bertos/cpu/byteorder.h b/bertos/cpu/byteorder.h
index b560559c4342f95cfe49c3c3bda698f19459c1cb..721f81e51435b9c1554728e1be917ebf98bf85de 100644 (file)
--- a/bertos/cpu/byteorder.h
+++ b/bertos/cpu/byteorder.h
@@ -40,6 +40,9 @@
 
 #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.
@@ -47,22 +50,27 @@
 #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.
  */
@@ -161,6 +169,37 @@ INLINE float net_to_host_float(float x)
        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.