[x265] [PATCH] asm: dct8 sse2 1.88x improvement over c code

[Steve Borho]

On Fri, Feb 20, 2015 at 10:06 AM, dave <dtyx265 at gmail.com> wrote:
> On 02/19/2015 04:58 PM, Steve Borho wrote:
>>
>> On 02/19, dtyx265 at gmail.com wrote:
>>>
>>> # HG changeset patch
>>> # User David T Yuen <dtyx265 at gmail.com>
>>> # Date 1424385856 28800
>>> # Node ID 28287b57013e9c43488bfba1570ded5cfb4af16d
>>> # Parent  039ea966d5ebccab1de2c3766fb7b4f125d2020a
>>> asm: dct8 sse2 1.88x improvement over c code
>>>
>>> This is backported from dct8 sse4
>>
>> it would be helpful for reviewers if you could tell us which CPUs this
>> will help and by how much.
>
> This is intended for cpus that support up to sse3 and until an ssse3 dct8
> primitive is developed, also ssse3.  The following is from testing on my
> sse3 system
>
> dct8x8            1.88x      11202.50      21044.87
>
>> If your CPU does not have SSE4 would the
>> encoder use the C reference if this primitive were not present?
>
> Yes, this replaces the C reference on systems with sse2 to ssse3.

thanks, queued with this info added to the commit message

-- 
Steve Borho