ARM: optimised MUL64

Originally committed as revision 14244 to svn://svn.ffmpeg.org/ffmpeg/trunk
This commit is contained in:
Måns Rullgård 2008-07-15 19:06:29 +00:00
parent 7995962d4b
commit 6865e04620
1 changed files with 9 additions and 0 deletions

View File

@ -48,6 +48,15 @@ static inline av_const int MULH(int a, int b)
hi; })
#endif
static inline av_const int64_t MUL64(int a, int b)
{
union { uint64_t x; unsigned hl[2]; } x;
asm ("smull %0, %1, %2, %3"
: "=r"(x.hl[0]), "=r"(x.hl[1]) : "r"(a), "r"(b));
return x.x;
}
#define MUL64 MUL64
#if defined(HAVE_ARMV5TE)
/* signed 16x16 -> 32 multiply add accumulate */