[libav-devel] [PATCH 6/6] lavr: Add x86-optimized function for flt to s32 conversion

Justin Ruggles justin.ruggles at gmail.com
Wed May 2 23:18:58 CEST 2012


On 05/02/2012 05:01 PM, Vitor Sessak wrote:

> On 05/02/2012 08:53 PM, Justin Ruggles wrote:
>> On 05/02/2012 02:46 PM, Vitor Sessak wrote:
>>
>>> On 05/01/2012 10:49 PM, Justin Ruggles wrote:
>>>> ---
>>>> Athlon64
>>>>      C - 8176
>>>> SSE2 - 1784
>>>>
>>>> SandyBridge
>>>>      C - 5578
>>>> SSE2 -  418
>>>
>>> Is there any reason why not to do an AVX (YMM) version?
>>
>>
>> It was actually slower when I tested it on sandy bridge. If some other
>> system runs it faster it's easy to add.
> 
> Strange. And what about 128-bit 3-op?


slower also.

-Justin


More information about the libav-devel mailing list