[MEncoder-users] x264 encoding slower after updating libx264

[belcampo]

belcampo wrote:
> Loren Merritt wrote:
>> On Sat, 14 Nov 2009, Dâniel Fraga wrote:
>>> On Sat, 14 Nov 2009, Loren Merritt wrote:
>>>
>>>> MB-tree is enabled by default now, didn't exist in api67, and takes a
>>>> bunch of memory.
>>>> B-frames and multiref are enabled by default now, weren't in api67, and
>>>> take some memory and cpu-time.
>>>> RDO, trellis, and several other options are enabled by default
>>>> now, weren't in api67, and take cpu-time.
>>>> All of the above also affect filesize.
>>>
>>>     And all of these options result in better quality? Because I
>>> didn't notice any improvement in the quality... of course I'm just
>>> looking at the video (I didn't do any scientific comparison).
>>
>> All of these options result in better compression ratio. That can be 
>> either better quality or lower bitrate, depending on what ratecontrol 
>> mode you choose.
> So if I understand correct, given a constant quality, it would result in 
> a smaller file ?
>>
>> --Loren Merritt
>>
>>
>> ------------------------------------------------------------------------
>>
>> _______________________________________________
>> MEncoder-users mailing list
>> MEncoder-users at mplayerhq.hu
>> https://lists.mplayerhq.hu/mailman/listinfo/mencoder-users
> 
> _______________________________________________
> MEncoder-users mailing list
> MEncoder-users at mplayerhq.hu
> https://lists.mplayerhq.hu/mailman/listinfo/mencoder-users
I did some testing with current ffmpeg/x264 in comparison with ffmpeg 
SVN-r17923/x264-65
convert 400frames from BBC-HD source to 720x576

Current version
[libx264 @ 0x8a537d0]frame I:5     Avg QP:22.00  size: 38218
[libx264 @ 0x8a537d0]frame P:395   Avg QP:22.00  size:  6277
size 2670643
time 0:30.10
Memory use 155m  95m ffmpeg

SVN-r17923/x264-65
[libx264 @ 0x89fc430]slice I:5     Avg QP:22.00  size: 38065
[libx264 @ 0x89fc430]slice P:395   Avg QP:22.00  size:  6565
size 2784070
time 0:25.31
Memory use 94428  42m ffmpeg

Indeed size is better by 4.08%
But time is worse by almost 19%
Memory needed more than doubles