[FFmpeg-devel] [PATCH 1/3] dnn: introduce dnn operand (in c code) to hold operand infos within network

[Guo, Yejun]

> -----Original Message-----
> From: ffmpeg-devel [mailto:ffmpeg-devel-bounces at ffmpeg.org] On Behalf Of
> Pedro Arthur
> Sent: Tuesday, August 27, 2019 10:24 PM
> To: FFmpeg development discussions and patches <ffmpeg-devel at ffmpeg.org>
> Subject: Re: [FFmpeg-devel] [PATCH 1/3] dnn: introduce dnn operand (in c code)
> to hold operand infos within network
> 
> Hi,
> 
> 
> Em ter, 20 de ago de 2019 às 05:54, Guo, Yejun <yejun.guo at intel.com>
> escreveu:
> >
> > the info can be saved in dnn operand object without regenerating again and
> again,
> > and it is also needed for layer split/merge, and for memory reuse.
> >
> > to make things step by step, this patch just focuses on c code,
> > the change within python script will be added later.
> >
> > Signed-off-by: Guo, Yejun <yejun.guo at intel.com>
> > ---
> >  libavfilter/dnn/dnn_backend_native.c           | 226
> ++++++++++++-------------
> >  libavfilter/dnn/dnn_backend_native.h           |  54 +++++-
> >  libavfilter/dnn/dnn_backend_native_layer_pad.c |  24 ++-
> >  libavfilter/dnn/dnn_backend_native_layer_pad.h |   4 +-
> >  tests/dnn/Makefile                             |   2 +-
> >  tests/dnn/dnn-layer-pad-test.c                 |  60 +++++--
> >  6 files changed, 236 insertions(+), 134 deletions(-)
> >
> > diff --git a/libavfilter/dnn/dnn_backend_native.c
> b/libavfilter/dnn/dnn_backend_native.c
> > index d52abc6..78227a5 100644
> > --- a/libavfilter/dnn/dnn_backend_native.c
> > +++ b/libavfilter/dnn/dnn_backend_native.c
> >

> > +int32_t calculate_operand_data_length(DnnOperand* operand)
> > +{
> > +    //av_assert0(operand->data_type == DNN_FLOAT);
> Please remove the whole commented line or uncomment the assert if it
> is relevant.

thanks, I'll remove it and add comments.

> 
> > +    return operand->dims[0] * operand->dims[1] * operand->dims[2] *
> operand->dims[3] * sizeof(float);
> > +}
> > +
> > +
> > +    /**
> > +     * to avoid possible memory leak, do not use char *name
> > +     */
> > +    char name[512];
> 512 bytes seems a bit too large for a name, but if you think it is
> really necessary I'm ok with it.
> 

good point, I'll change to 128.