#ifdef __GNUC__
#define INLINE __inline
+#define PACKED __attribute__((packed))
#elif defined(__WATCOMC__)
#define INLINE __inline
+#define PACKED
#else
#define INLINE
+#define PACKED
#endif
+#define BSWAP16(x) ((((x) >> 8) & 0xff) | (((x) & 0xff) << 8))
+#define BSWAP32(x) \
+ ((((x) >> 24) & 0xff) | \
+ (((x) >> 8) & 0xff00) | \
+ (((x) << 8) & 0xff0000) | \
+ ((x) << 24))
+
+#if defined(__i386__) || defined(__x86_64__) || defined(__386__) || defined(MSDOS)
/* fast conversion of double -> 32bit int
* for details see:
* - http://chrishecker.com/images/f/fb/Gdmfp.pdf
val += 6755399441055744.0;
return *(int32_t*)&val;
}
+#else
+#define cround64(x) ((int32_t)(x))
+#endif
+
+static INLINE float rsqrt(float x)
+{
+ float xhalf = x * 0.5f;
+ int32_t i = *(int32_t*)&x;
+ i = 0x5f3759df - (i >> 1);
+ x = *(float*)&i;
+ x = x * (1.5f - xhalf * x * x);
+ return x;
+}
extern uint32_t perf_start_count, perf_interval_count;
#ifdef __WATCOMC__
+void memset16(void *dest, uint16_t val, int count);
+#pragma aux memset16 = \
+ "cld" \
+ "test ecx, 1" \
+ "jz memset16_dwords" \
+ "rep stosw" \
+ "jmp memset16_done" \
+ "memset16_dwords:" \
+ "shr ecx, 1" \
+ "push ax" \
+ "shl eax, 16" \
+ "pop ax" \
+ "rep stosd" \
+ "memset16_done:" \
+ parm[edi][ax][ecx];
+
+#ifdef USE_MMX
+void memcpy64(void *dest, void *src, int count);
+#pragma aux memcpy64 = \
+ "cploop:" \
+ "movq mm0, [edx]" \
+ "movq [ebx], mm0" \
+ "add edx, 8" \
+ "add ebx, 8" \
+ "dec ecx" \
+ "jnz cploop" \
+ "emms" \
+ parm[ebx][edx][ecx] \
+ modify[8087];
+#else
+#define memcpy64(dest, src, count) memcpy(dest, src, (count) << 3)
+#endif
+
void perf_start(void);
#pragma aux perf_start = \
"xor eax, eax" \
void debug_break(void);
#pragma aux debug_break = "int 3";
+
+void halt(void);
+#pragma aux halt = "hlt";
#endif
#ifdef __GNUC__
+#if defined(__i386__) || defined(__x86_64__)
+#define memset16(dest, val, count) asm volatile ( \
+ "cld\n\t" \
+ "test $1, %2\n\t" \
+ "jz 0f\n\t" \
+ "rep stosw\n\t" \
+ "jmp 1f\n\t" \
+ "0:\n\t" \
+ "shr $1, %2\n\t" \
+ "push %%ax\n\t" \
+ "shl $16, %%eax\n\t" \
+ "pop %%ax\n\t" \
+ "rep stosl\n\t" \
+ "1:\n\t"\
+ :: "D"(dest), "a"((uint16_t)(val)), "c"(count) \
+ : "memory")
+#else
+static void INLINE memset16(void *dest, uint16_t val, int count)
+{
+ uint16_t *ptr = dest;
+ while(count--) *ptr++ = val;
+}
+#endif
+
+#ifdef USE_MMX
+#define memcpy64(dest, src, count) asm volatile ( \
+ "0:\n\t" \
+ "movq (%1), %%mm0\n\t" \
+ "movq %%mm0, (%0)\n\t" \
+ "add $8, %1\n\t" \
+ "add $8, %0\n\t" \
+ "dec %2\n\t" \
+ "jnz 0b\n\t" \
+ "emms\n\t" \
+ :: "r"(dest), "r"(src), "r"(count) \
+ : "%mm0")
+#else
+#define memcpy64(dest, src, count) memcpy(dest, src, (count) << 3)
+#endif
+
#define perf_start() asm volatile ( \
"xor %%eax, %%eax\n" \
"cpuid\n" \
#define debug_break() \
asm volatile ("int $3")
+
+#define halt() \
+ asm volatile("hlt")
#endif
#ifdef _MSC_VER
+void __inline memset16(void *dest, uint16_t val, int count)
+{
+ __asm {
+ cld
+ mov ax, val
+ mov edi, dest
+ mov ecx, count
+ test ecx, 1
+ jz memset16_dwords
+ rep stosw
+ jmp memset16_done
+ memset16_dwords:
+ shr ecx, 1
+ push ax
+ shl eax, 16
+ pop ax
+ rep stosd
+ memset16_done:
+ }
+}
+
#define perf_start() \
do { \
__asm { \