[Libva] mpeg4 decoding
Gwenole Beauchesne
gb.devel at gmail.com
Wed Jan 14 12:48:28 PST 2015
Hi,
2015-01-14 17:50 GMT+01:00 Varga, Michael <Michael.Varga at amd.com>:
> Yes, the VOP header is needed for decoding. If the upper layer ( possibly ffmpeg ) would stop truncating the VOP header this would solve my problem, in which case I would not need modulo_time_base.
I think this would be the better option, but if I remember correctly,
that was not quite possible without refactoring other code in FFmpeg
at the risk of breaking the various MPEG-4:2 derivatives from which I
had not clue about.
> I need a value for modulo_time_base to regenerate the VOP header. modulo_time_base can be either generated auto-magically or passed in through the API.
>
> Any suggestions?
We could add a field at the end of the struct, as usual for API
compatibility, but this still offers the risk of ABI breakage if the
the VA struct is allocated on stack. An API & ABI compatible way could
be achieve to insert that field into the existing vop_fields
bitfields, assuming modulo_time_base requires up to 7 bits based on
quick & initial calculation.
> -----Original Message-----
> From: Xiang, Haihao [mailto:haihao.xiang at intel.com]
> Sent: Tuesday, January 13, 2015 10:41 PM
> To: Varga, Michael
> Cc: Emil Velikov; Gwenole Beauchesne; libva at lists.freedesktop.org
> Subject: RE: [Libva] mpeg4 decoding
>
> On Mon, 2015-01-12 at 21:38 +0000, Varga, Michael wrote:
>> Haihao,
>>
>> I use modulo_time_base when reconstructing the header VideoObjectPlane header.
>
> Does HW need the VOP header for decode ?
>
>> When modulo_time_base is left blank frames do not decode correctly.
>> From your comment your refering to vop_time_increment?
>
> No, I didn't refer to vop_time_increment or modulo_time_base. I just wonder it is really needed to add a new field in
> VAPictureParameterBufferMPEG4 or not.
>
>
>> For gpu acceleration vop_time_increment can be set to "0" AFAIK.
>>
>> MPEG 4 spec:
>> modulo_time_base: This value represents the local time base in one second resolution units (1000 milliseconds).
>> It consists of a number of consecutive '1' followed by a '0'. Each '1'
>> represents a duration of one second that have elapsed. For I-,
>> S(GMC)-, and P-VOPs of a non scalable bitstream and the base layer of
>> a scalable bitstream, the number of '1's indicate the number of
>> seconds elapsed since the synchronization point marked by time_code of
>> the previous GOV header or by modulo_time_base of the previously
>> decoded I-, S(GMC)-, or P-VOP, in decoding order. For B-VOP of a non
>> scalable bitstream and a base layer of a scalable bitstream, the
>> number of '1's indicates the number of seconds elapsed since the synchronization point marked in the previous GOV header, or I-VOP, S(GMC)-VOP, or P-VOP, in display order. For I-, P-, or B-VOPs of enhancement layer of scalable bitstream, the number of '1's indicate the number of seconds elapsed since the synchronization point marked in the previous GOV header, I-VOP, P-VOP, or B-VOP, in display order.
>>
>> vop_time_increment: This value represents the absolute
>> vop_time_increment from the synchronization point marked by the
>> modulo_time_base measured in the number of clock ticks. It can take a
>> value in the range of [0,vop_time_increment_resolution). The number of
>> bits representing the value is calculated as the minimum number of unsigned integer bits required to represent the above range. The local time base in the units of seconds is recovered by dividing this value by the vop_time_increment_resolution.
>>
>>
>> Thanks
>> Michael
>>
>> -----Original Message-----
>> From: Xiang, Haihao [mailto:haihao.xiang at intel.com]
>> Sent: Wednesday, January 07, 2015 11:51 PM
>> To: Varga, Michael
>> Cc: Emil Velikov; Gwenole Beauchesne; libva at lists.freedesktop.org
>> Subject: Re: [Libva] mpeg4 decoding
>>
>>
>> Hi Michael,
>>
>> I am not familiar with MPEG4, I think modulo_time_base is used to calculate time stamp etc, is modulo_time_base really required for HW decode ?
>>
>> Thanks
>> Haihao
>>
>> > I only really need modulo_time_base. It would appear from my tests vop_time_increment is not needed. Could modulo_time_base be added?
>> >
>> > The core issue I am having is I need the modulo_time_base value. I noticed Gwenole wrote a reference driver which wraps vdpau. Gwenole's code has a fix to compute modulo_time_base. I cannot use code from Gwenole's driver due to licensing issues, his code is GPL and MESA is MIT.
>> >
>> > -----Original Message-----
>> > From: Emil Velikov [mailto:emil.l.velikov at gmail.com]
>> > Sent: Saturday, November 08, 2014 9:58 AM
>> > To: Varga, Michael; Gwenole Beauchesne
>> > Cc: emil.l.velikov at gmail.com; libva at lists.freedesktop.org
>> > Subject: Re: [Libva] mpeg4 decoding
>> >
>> > Hi guys.
>> >
>> > Don't mean to be picky or anything but won't this change cause a non-backward compat ABI/API breakage ?
>> >
>> > While adding modulo_time_base might be ok (highly dependant on the implementation which may use the vop_fields union without sanitising it), vop_time_increment is definitely going to cause issues.
>> >
>> >
>> > Cheers,
>> > Emil
>> >
>> > On 31/10/14 14:32, Varga, Michael wrote:
>> > > Hi Gwenole,
>> > >
>> > > I double checked vop_time_increment and its not strictly needed,
>> > > setting this value to 0 works but if you'll add it I'll use anyway.
>> > > My patch is below. I declared modulo_time_base with 4 bits, the
>> > > MPEG4 spec does not specify an upper bound for the size of this
>> > > variable. If you think this is too big I can reduce it. Also, is
>> > > there anything you think I should know about your process for submitting patches? Thanks!
>> > > :)
>> > >
>> > >
>> > > From 44aeb5d81789c42404d6a2398ef06699aabc6d73 Mon Sep 17 00:00:00
>> > > 2001
>> > > From: Michael Varga <Michael.Varga at amd.com>
>> > > Date: Fri, 31 Oct 2014 08:58:34 -0500
>> > > Subject: [PATCH] va: added two variables to
>> > > VAPictureParameterBufferMPEG4
>> > >
>> > > Added modulo_time_base and vop_time_increment so the VOP header
>> > > may be reconstructed if it has been truncated.
>> > >
>> > > Signed-off-by: Michael Varga <Michael.Varga at amd.com>
>> > > ---
>> > > va/va.h | 2 ++
>> > > 1 file changed, 2 insertions(+)
>> > >
>> > > diff --git a/va/va.h b/va/va.h
>> > > index 01694a9..c21770c 100644
>> > > --- a/va/va.h
>> > > +++ b/va/va.h
>> > > @@ -1312,12 +1312,14 @@ typedef struct _VAPictureParameterBufferMPEG4
>> > > unsigned int intra_dc_vlc_thr : 3;
>> > > unsigned int top_field_first : 1;
>> > > unsigned int alternate_vertical_scan_flag : 1;
>> > > + unsigned int modulo_time_base : 4;
>> > > } bits;
>> > > unsigned int value;
>> > > } vop_fields;
>> > > unsigned char vop_fcode_forward;
>> > > unsigned char vop_fcode_backward;
>> > > unsigned short vop_time_increment_resolution;
>> > > + unsigned short vop_time_increment;
>> > > /* short header related */
>> > > unsigned char num_gobs_in_vop;
>> > > unsigned char num_macroblocks_in_gob;
>> > >
>> >
>> > _______________________________________________
>> > Libva mailing list
>> > Libva at lists.freedesktop.org
>> > http://lists.freedesktop.org/mailman/listinfo/libva
>>
>>
>
>
Regards,
--
Gwenole Beauchesne
Intel Corporation SAS / 2 rue de Paris, 92196 Meudon Cedex, France
Registration Number (RCS): Nanterre B 302 456 199
More information about the Libva
mailing list