Adds a flag context_reinit to MpegEncContext to relieable keep track
of frame parameter changes which require a context reinitialization.
This is required for broken inputs which change the frame size but
error out before the context can be reinitialized.
This is mainly required for frame parameter changes during frame based
multithreading but single threaded usage profits too from avoiding
ff_MPV_common_end()/ff_MPV_common_init() cycles.
Prevents dangling pointers and makes access after free more obvious.
Setting AVFrame.qscale_table to NULL is required for successfully
allocating a previously freed Picture with ff_alloc_picture().
this crashes otherwise, and can happen from try_decode_frame() in the case of decoding errors
Fixes Ticket1602
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
This file defines a single, huge function, MPV_motion(), which
although being declared inline is not actually inlined by the
compiler (for good reason). There is thus no sense in defining
this function in a header file, resulting in multiple copies of
it in the final library.
Signed-off-by: Mans Rullgard <mans@mansr.com>
Defining restrict results - for some compilers - in changing other
uses of the restrict keyword also, e.g. __declspec(restrict) gets
changed to __declspec(__restrict) on MSVC. This causes compilation
failures. Therefore, using a private namespace macro instead is
more reliable and robust.
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
There have been multiple user complaints about loosing this feature
while its not clear the 3% speedloss claims where real or fabricated.
My own testing indicates no statistically significant speed difference
both with mpeg2 and mpeg4, and if at all the code with lowres support
is a tiny bit faster than without.
This reverts commit 92ef4be4ab9fbb7d901b22e0036a4ca90b00a476, reversing
changes made to 2e07f42957.
Conflicts:
cmdutils.c
libavcodec/arm/vp8dsp_init_arm.c
libavcodec/mpegvideo.c
libavcodec/mpegvideo.h
libavutil/arm/Makefile
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
The new lowres support is limited to decoders where lowres decoding
is possible in high quality.
I was not able to measure any speed difference, but if one is found
the 2-3 lines that might affect speed can be made compile time conditional
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
This fixes race conditions that ultimately lead to memory corruption.
Found-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
If the dummy frame are not created from a reference frame they could
be deleted untimely resulting in multithreaded decoder waiting on
the current frame to finish.
Noticed by Ronald S. Bultje in the RV34 decoder with a broken file.
The DC coefficient should be included, too.
This probably was missed because DC quantizer is always
even for MPEG-1/2 but this function is also used for MPEG-4.
Signed-off-by: Reimar Döffinger <Reimar.Doeffinger@gmx.de>
When either video dimension is only one macroblock, subtractions
based on v_edge_pos and the macroblock size may be negative. In
that situation, an unsigned comparison isn't sufficent to test for
MV overruns, because a limit of (unsigned)-1 will let any other
value pass.
Signed-off-by: Anton Khirnov <anton@khirnov.net>
Under certain conditions pictures could be released before they were
returned with frame-threading. Broken mv computation in the upcoming
rv34 frame-threading patch was caused by this.
To prevent contexts from running out of available pictures the loop
releasing "unused" pictures has to be run for B frames too.
Adds a new member to MpegEncContext to hold the number of used slice
contexts. Fixes segfaults with '-threads 17 -thread_type slice' and
fate-vsynth{1,2}-mpeg{2,4}thread{,_ilace} with --disable-pthreads.
This was intended as an optimisation for skipped blocks in MPEG2
P-frames and never used elsewhere. Removing this "optimisation"
speeds up MPEG2 decoding by 1-2% (ARM Cortex-A9).
Signed-off-by: Mans Rullgard <mans@mansr.com>
This reverts commit 4f820131fa.
It is better to abort() than to have remotly exploitable arbitrary code
execution bugs. Even more so that this abort has never been triggered
by any input people threw at it.
If after more extensive testing its removial is found safe we can remove
the abort() later.
Janne Grunau
Michael Niedermayer
Alberto Delmás
Diego Biurrun
Mans Rullgard
Anton Khirnov
Ronald S. Bultje
Lou Logan
Martin Storsjö
Reimar Döffinger
John Brooks
Konstantin Todorov
Andrey Utkin
Dustin Brody