Hi,

On Fri, Apr 22, 2011 at 6:55 PM, Alexander Strange
<[email protected]> wrote:
> On Fri, Apr 22, 2011 at 6:01 PM, Ronald S. Bultje <[email protected]> wrote:
>> From: Jason Garrett-Glaser <[email protected]>
>>
>> This prevents a call to bytestream_get_be16() using a movzwl both before
>> and after the ror instruction, which is obviously inefficient. Arm uses
>> the same trick also.
>>
>> Sintel decoding goes from (avg+SD) 9.856 +/- 0.003 to 9.797 +/- 0.003 sec.
>>
>
> We should replace most inlined uses of asm with compiler intrinsics.

What if we don't use gcc, though?

> The code generated is good for most of them on x86.
> This includes the SSE code I introduced in intreadwrite.
>
> Using asm slightly increases code size because all asm statements are
> assumed to degrade aliasing and to clobber eflags.
>
> Of course this isn't true for other architectures like arm, where we
> should probably keep them.
>
> That said, patch OK.

Applied.

Ronald
_______________________________________________
libav-devel mailing list
[email protected]
https://lists.libav.org/mailman/listinfo/libav-devel

Reply via email to