[MPlayer-dev-eng] [PATCH]: af_resample SIMD optimization (MMX/SSE2/3DNOW!/SSE/SSE3)

[Loren Merritt]

On Tue, 31 Mar 2009, Zhou Zongyi wrote:

>+static inline uint32_t fastmod(uint32_t a, uint32_t b, uint32_t c)
>+{
>+    uint32_t ret;
>+    __asm__
>+        (
>+        "mull %2 \n\t"
>+        "mull %3 \n\t"
>+        :"=d"(ret)
>+        :"a"(a),"g"(c),"g"(b)
>+        );
>+    return ret;
>+}

mull modifies eax, so mark a as modified.
Both a and ret must be earlyclobber, otherwise gcc can put b in the same reg.

--Loren Merritt