[FFmpeg-devel] [PATCH] ARM: NEON optimised {put, avg}_h264_chroma_mc[48]

[Ian Caulfield]

2008/8/25 M?ns Rullg?rd <mans at mansr.com>:
> "Ian Caulfield" <ian.caulfield at gmail.com> writes:
>
>> 2008/8/25 M?ns Rullg?rd <mans at mansr.com>:
>>> "Ian Caulfield" <ian.caulfield at gmail.com> writes:
>>>
>>>> 2008/8/25 Mans Rullgard <mans at mansr.com>
>>>>>
>>>>> +        vld1.64         {d4, d5}, [r1], r4
>>>>
>>>> Can you guarantee any kind of alignment on the address in r1? If so,
>>>> potentially better performance can be achieved by specifying the
>>>> alignment in the opcode...
>>>
>>> I know, and I've specified it everywhere I know of a guaranteed
>>> alignment.  I don't know which line you're quoting, but if it's the
>>> source address for an h264_chroma_mc function, there are no promises.
>>
>> Fair enough - hadn't looked at your other patches.
>
> If you see anything else that looks suboptimal, please tell.
>
>>> Do you have something ARMv7 based?
>>
>> No, but I'm quite familiar with the Neon units on the Cortex-A8 and -A9.
>
> The A9 hasn't even made it to silicon yet.  Do you know something the
> rest of us don't?

I work for ARM, so probably ;)

Ian