On 11/8/11 1:40 PM, Sebastien Zwickert wrote:
Hi,
Updated patch in attachment. I've worked on the latest patch sent in this
thread.
I dropped changes made in doc/general.texi cause of HW-accels are not listed.
In configure, I added vda deps (h264_decoder and pthreads).
I added vda to CONFIG_LIST and updated activation behavior .
Now, by default vda is enable if it is not explicitly disabled in configure
line and if header is found.
In mpegvideo.c, I added again vda pix fmt that was removed from my initial
patch. In fact, in
h264 decode_slice_header when the context calls get_format the PixelFormat list
passed as argument
is always the list from mpeg (because color range is always mpeg).
I reordered the headers inclusion (make checkheaders returns without error).
In vda.h, I undefined Picture after the inclusion of VDADecoder.h (Picture
struct is redefined in QD.h
which get included from VDADecoder.h).
Unfortunately, I have no better workaround for the emmintrin.h and gcc-4.2.1
issue. (emmintrin.h get
included from VDADecoder.h).
Finally, I set pixel format buffer attributes on decoder creation
(ff_vda_create_decoder).
I tested the patch with these configure lines :
./configure --disable-everything --enable-vda
./configure --disable-everything --enable-hwaccel=vda
./configure --enable-vda
./configure --disable-vda
./configure
All these configures succeed in compiling.
I built a vlc release with libav and vda enabled and it works as expected.
I confirm that builds properly, we discussed about using directly
pthreads call instead of the lock manager so it will be updated again
shortly =)
lu
_______________________________________________
libav-devel mailing list
[email protected]
https://lists.libav.org/mailman/listinfo/libav-devel
