headers
Horacio | 1 Jun 02:02
Picon

Re: Compile flags for iCore7?


木曜日 28 5月 2009 20:00:02 に Robert Swain さんは書きました:
> Horacio wrote:
> > I got nice Nehalem CPU and I am compiling FFMpeg and friends (x264,
> > flaac, faad, theora, etc..) on MinGW in a Vista Ultimate 64 bits OS.
> >
> > I was wondering if there is any particular gcc-4.3 flags to optimize all
> > libraries and get the most of the Nethalem power??  I have seen  -mssse3,
> > - pipe -O3.
>
> In Linux with gcc 4.3.3 I use:
>
> -march=core2 -msse4 -msse4.1 -msse4.2
>
>
Thanks a lot, I tried these on MinGW TDS but I got no improvement. I am trying 
to encode x264 at 1920x1080 and it goes at 1 fps using the hq preset.  Any 
tips to accelerate this would be greatly appreciated.

regards
Horacio

> Regards,
> Rob
> _______________________________________________
> ffmpeg-user mailing list
> ffmpeg-user <at> mplayerhq.hu
> https://lists.mplayerhq.hu/mailman/listinfo/ffmpeg-user

_______________________________________________
(Continue reading)

Permalink | Reply | 5 comments |
headers
ffmpeg | 1 Jun 02:37
Picon
Favicon

Same quality without huge file size

I would like to convert files without quality loss but "-sameq" generate huge files.

I have WMF file with 4.76 MB of size...

Input #0, asf, from 'xxx.wmv':
  Duration: 00:10:30.26, start: 3.000000, bitrate: 63 kb/s
    Stream #0.0: Audio: wmav2, 16000 Hz, mono, s16, 16 kb/s
    Stream #0.1: Video: wmv1, yuv420p, 320x240, 80 kb/s, 1k tbr, 1k tbn, 1k tbc
Output #0, flv, to 'xxx.flv':
    Stream #0.0: Video: flv, yuv420p, 320x240, q=2-31, 10 kb/s, 1k tbn, 25 tbc
    Stream #0.1: Audio: libmp3lame, 11025 Hz, mono, s16, 16 kb/s

After conversion with "-sameq" and reduction frame rate from 40 to 25 it has... 51 MB!
After conversion with "-b 10000" and same reduction of frame rate it has... 7.99 MB - still more than the
original file...
After conversion with "-b 1000" (!) and same reduction of frame rate it has 7.96!

Whan can I do more to convert without quality loss
but with filesize equal (or smaller) to original file?
Permalink | Reply | 4 comments |
headers
Doug Baggett | 1 Jun 02:48
Picon

Re: Same quality without huge file size

I think there has been some mis-statements in the documentation concerning
that switch. You are not the first to encounter this (same thing happened
with me the first time I tried it)

On Sun, May 31, 2009 at 8:37 PM, ffmpeg <ffmpeg <at> o2.pl> wrote:

> I would like to convert files without quality loss but "-sameq" generate
> huge files.
>
> I have WMF file with 4.76 MB of size...
>
> Input #0, asf, from 'xxx.wmv':
>  Duration: 00:10:30.26, start: 3.000000, bitrate: 63 kb/s
>    Stream #0.0: Audio: wmav2, 16000 Hz, mono, s16, 16 kb/s
>    Stream #0.1: Video: wmv1, yuv420p, 320x240, 80 kb/s, 1k tbr, 1k tbn, 1k
> tbc
> Output #0, flv, to 'xxx.flv':
>    Stream #0.0: Video: flv, yuv420p, 320x240, q=2-31, 10 kb/s, 1k tbn, 25
> tbc
>    Stream #0.1: Audio: libmp3lame, 11025 Hz, mono, s16, 16 kb/s
>
>
> After conversion with "-sameq" and reduction frame rate from 40 to 25 it
> has... 51 MB!
> After conversion with "-b 10000" and same reduction of frame rate it has...
> 7.99 MB - still more than the original file...
> After conversion with "-b 1000" (!) and same reduction of frame rate it has
> 7.96!
>
> Whan can I do more to convert without quality loss
(Continue reading)

Permalink | Reply | 0 comments |
headers
Jason Garrett-Glaser | 1 Jun 02:56
Picon

Re: Compile flags for iCore7?

On Sun, May 31, 2009 at 8:02 PM, Horacio <hsanson <at> gmail.com> wrote:
>
>
> 木曜日 28 5月 2009 20:00:02 に Robert Swain さんは書きました:
>> Horacio wrote:
>> > I got nice Nehalem CPU and I am compiling FFMpeg and friends (x264,
>> > flaac, faad, theora, etc..) on MinGW in a Vista Ultimate 64 bits OS.
>> >
>> > I was wondering if there is any particular gcc-4.3 flags to optimize all
>> > libraries and get the most of the Nethalem power??  I have seen  -mssse3,
>> > - pipe -O3.
>>
>> In Linux with gcc 4.3.3 I use:
>>
>> -march=core2 -msse4 -msse4.1 -msse4.2
>>
>>
> Thanks a lot, I tried these on MinGW TDS but I got no improvement. I am trying
> to encode x264 at 1920x1080 and it goes at 1 fps using the hq preset.  Any
> tips to accelerate this would be greatly appreciated.

Did you forget -threads 0 (or compile without asm?)

Dark Shikari
Permalink | Reply | 4 comments |
headers
Baptiste Coudurier | 1 Jun 04:51
Picon

Re: XDCAM HD Codec?

Hi,

On 5/31/2009 10:49 AM, subs <at> mxr.at wrote:
> Hi!
> 
> I am trying to process my XDCAM HD (imported with XDCAM Transfer from MXF
> to MOV) files with ffmpeg but I recieve an input error: Unsupported codec
> (id=0) for input stream #0.0.
> 
> When I open the files in QT the format is: Apple XDCAM HD 1080p25 (35 Mb/s
> VBR), 1440 x 1080
> 
> Anyone has a solution?

Use latest svn, it should be supported. If not, please upload a sample
somewhere I'll have a look.

--

-- 
Baptiste COUDURIER                              GnuPG Key Id: 0x5C1ABAAA
Key fingerprint                 8D77134D20CC9220201FC5DB0AC9325C5C1ABAAA
FFmpeg maintainer                                  http://www.ffmpeg.org
Permalink | Reply | 1 comment |
headers
Baptiste Coudurier | 1 Jun 04:53
Picon

Re: Same quality without huge file size

On 5/31/2009 5:37 PM, ffmpeg wrote:
> [...]
> 
> Whan can I do more to convert without quality loss
> but with filesize equal (or smaller) to original file?

Please understand that what you are asking for is fairy powder.

--

-- 
Baptiste COUDURIER                              GnuPG Key Id: 0x5C1ABAAA
Key fingerprint                 8D77134D20CC9220201FC5DB0AC9325C5C1ABAAA
FFmpeg maintainer                                  http://www.ffmpeg.org
Permalink | Reply | 2 comments |
headers
Baptiste Coudurier | 1 Jun 04:57
Picon

Re: field extract

Hi,

On 5/31/2009 4:39 AM, belcampo wrote:
>> vob file has interlaced MPEG2 video stream. I need one field of every
>> interlaced frame. a stream without distortion made by
>> deinterlacing and without artifacts on moving object on interlaced
>> frames. What I need is 2CIF stream - one field of
>> interlaced vob file. I know it is possible to extract field stream (or
>> series of jpegs) with ffmpeg but it is not obvious
>> and I cannot figure out now.
>>
> http://lists.mplayerhq.hu/pipermail/ffmpeg-user/2007-July/010059.html it
> says: ffmpeg doesn't interlace fields in frame, take a look to the mjpeg
> tools.

Humm, what is this supposed to mean ? "doesn't interlace fields in frame" ?

Ffmpeg can encode material in interlaced mode and decoder interlaced
material.

PS: I don't if this is you or not, but someone really screw threading.
Please post in a clean manner and avoid breaking threads.

--

-- 
Baptiste COUDURIER                              GnuPG Key Id: 0x5C1ABAAA
Key fingerprint                 8D77134D20CC9220201FC5DB0AC9325C5C1ABAAA
FFmpeg maintainer                                  http://www.ffmpeg.org
Permalink | Reply | 0 comments |
headers
Horacio | 1 Jun 06:39
Picon

Re: Compile flags for iCore7?


On Monday 01 June 2009 09:56:57 Jason Garrett-Glaser wrote:
> On Sun, May 31, 2009 at 8:02 PM, Horacio <hsanson <at> gmail.com> wrote:
> > 木曜日 28 5月 2009 20:00:02 に Robert Swain さんは書きました:
> >> Horacio wrote:
> >> > I got nice Nehalem CPU and I am compiling FFMpeg and friends (x264,
> >> > flaac, faad, theora, etc..) on MinGW in a Vista Ultimate 64 bits OS.
> >> >
> >> > I was wondering if there is any particular gcc-4.3 flags to optimize
> >> > all libraries and get the most of the Nethalem power??  I have seen 
> >> > -mssse3, - pipe -O3.
> >>
> >> In Linux with gcc 4.3.3 I use:
> >>
> >> -march=core2 -msse4 -msse4.1 -msse4.2
> >
> > Thanks a lot, I tried these on MinGW TDS but I got no improvement. I am
> > trying to encode x264 at 1920x1080 and it goes at 1 fps using the hq
> > preset.  Any tips to accelerate this would be greatly appreciated.
>
> Did you forget -threads 0 (or compile without asm?)
>

x264 configure shows asm support yes but I have to disable pthreads because 
ffmpeg wont compile if enabled.

Are the iCore7 optimizations functional in Windows? shall I expect a higher 
encoding speed?

thanks,
(Continue reading)

Permalink | Reply | 3 comments |
headers
Jason Garrett-Glaser | 1 Jun 06:50
Picon

Re: Compile flags for iCore7?

2009/6/1 Horacio <hsanson <at> gmail.com>:
>
>
> On Monday 01 June 2009 09:56:57 Jason Garrett-Glaser wrote:
>> On Sun, May 31, 2009 at 8:02 PM, Horacio <hsanson <at> gmail.com> wrote:
>> > 木曜日 28 5月 2009 20:00:02 に Robert Swain さんは書きました:
>> >> Horacio wrote:
>> >> > I got nice Nehalem CPU and I am compiling FFMpeg and friends (x264,
>> >> > flaac, faad, theora, etc..) on MinGW in a Vista Ultimate 64 bits OS.
>> >> >
>> >> > I was wondering if there is any particular gcc-4.3 flags to optimize
>> >> > all libraries and get the most of the Nethalem power??  I have seen
>> >> > -mssse3, - pipe -O3.
>> >>
>> >> In Linux with gcc 4.3.3 I use:
>> >>
>> >> -march=core2 -msse4 -msse4.1 -msse4.2
>> >
>> > Thanks a lot, I tried these on MinGW TDS but I got no improvement. I am
>> > trying to encode x264 at 1920x1080 and it goes at 1 fps using the hq
>> > preset.  Any tips to accelerate this would be greatly appreciated.
>>
>> Did you forget -threads 0 (or compile without asm?)
>>
>
> x264 configure shows asm support yes but I have to disable pthreads because
> ffmpeg wont compile if enabled.

You need to enable pthreads in ffmpeg as well.  You cannot
meaningfully complain about speed if you refuse to use multithreading.
(Continue reading)

Permalink | Reply | 2 comments |
headers
Rodney Baker | 1 Jun 08:21
Picon
Picon
Favicon
Gravatar

Re: Compile flags for iCore7?

On Mon, 1 Jun 2009 14:20:56 Jason Garrett-Glaser wrote:
> 2009/6/1 Horacio <hsanson <at> gmail.com>:
> > On Monday 01 June 2009 09:56:57 Jason Garrett-Glaser wrote:
> >> On Sun, May 31, 2009 at 8:02 PM, Horacio <hsanson <at> gmail.com> wrote:
> >> > 木曜日 28 5月 2009 20:00:02 に Robert Swain さんは書きました:
> >> >> Horacio wrote:
> >> >> > I got nice Nehalem CPU and I am compiling FFMpeg and friends (x264,
> >> >> > flaac, faad, theora, etc..) on MinGW in a Vista Ultimate 64 bits
> >> >> > OS.
> >> >> >
> >> >> > I was wondering if there is any particular gcc-4.3 flags to
> >> >> > optimize all libraries and get the most of the Nethalem power??  I
> >> >> > have seen -mssse3, - pipe -O3.
> >> >>
> >> >> In Linux with gcc 4.3.3 I use:
> >> >>
> >> >> -march=core2 -msse4 -msse4.1 -msse4.2
> >> >
> >> > Thanks a lot, I tried these on MinGW TDS but I got no improvement. I
> >> > am trying to encode x264 at 1920x1080 and it goes at 1 fps using the
> >> > hq preset.  Any tips to accelerate this would be greatly appreciated.
> >>
> >> Did you forget -threads 0 (or compile without asm?)
> >
> > x264 configure shows asm support yes but I have to disable pthreads
> > because ffmpeg wont compile if enabled.
>
> You need to enable pthreads in ffmpeg as well.  You cannot
> meaningfully complain about speed if you refuse to use multithreading.
>
(Continue reading)

Permalink | Reply | 1 comment |

Gmane