This is something of a hack. It allocates a new hwframe context for
the target format, then maps it back to the source link and overwrites
the input link hw_frames_ctx so that the previous filter will receive
the frames we want from ff_get_video_buffer(). It may fail if
the previous filter imposes any additional constraints on the frames
it wants to use as output.
(cherry picked from commit 81a4cb8e58)
Use the flags argument of av_hwframe_ctx_create_derived() to pass the
mapping flags which will be used on allocation. Also, set the format
and hardware context on the allocated frame automatically - the user
should not be required to do this themselves.
(cherry picked from commit c5714b51aa)
Some frames contexts are not usable without additional format-specific
state in hwctx. This change adds new functions frames_derive_from and
frames_derive_to to initialise this state appropriately when deriving
a frames context which will require it to be set.
(cherry picked from commit 27978155bc)
This only supports one device globally, but more can be used by
passing them with input streams in hw_frames_ctx or by deriving new
devices inside a filter graph with hwmap.
(cherry picked from commit e669db7610)
The non-H.26[45] codecs already use this form. Since we don't
currently generate I frames for codecs which support them separately
to IDR, the p_per_i variable is set to infinity by default so that it
doesn't interfere with any other calculation. (All the code for I
frames still exists, and it works for H.264 if set manually.)
(cherry picked from commit 6af014f402)
In order to work correctly with the i965 driver, this also fixes the
direction of forward/backward references - forward references are
intended to be those from the past to the current frame, not from the
current frame to the future.
(cherry picked from commit 9aa251c98c)
Adds functions to convert to/from strings and a function to iterate
over all supported device types. Also adds a new invalid type
AV_HWDEVICE_TYPE_NONE, which acts as a sentinel value.
(cherry picked from commit b7487f4f3c)
Previously this was leaking, though it actually hit an assert making
sure that the buffer had already been cleared when freeing the picture.
(cherry picked from commit 17aeee5832)
The driver is somewhat bitrotten (not updated for years) but is still
usable for decoding with this change. To support it, this adds a new
driver quirk to indicate no support at all for surface attributes.
Based on a patch by wm4 <nfxjfg@googlemail.com>.
(cherry picked from commit e791b915c7)
* commit 'f8f7ad758d0e1f36915467567f4d75541d98c12f':
qsv: Set the correct range for la_depth
This commit is a noop. There is a separate parameter to enable
lookahead, so overloading the depth is unnecessary.
Merged-by: Mark Thompson <sw@jkqxz.net>
* commit '00aeedd84105a17f414185bd33ecadebeddb3a27':
qsv{dec,enc}: use a struct as a memory id with internal memory allocator
Merged-by: Mark Thompson <sw@jkqxz.net>
* commit '404e51478ecad060249d5b9bee6ab39a8a9d8c1c':
qsv{dec,enc}: always use an internal mfxFrameSurface1
Minor fixups for differences in the QSV encoder because of a53cc.
Merged-by: Mark Thompson <sw@jkqxz.net>
* commit '8ea15afbf2c1ec89b5d4bac1f0b8345e4b906a5d':
hwcontext_qsv: transfer data through the child context when VPP fails
Merged-by: Mark Thompson <sw@jkqxz.net>
* commit 'b91ce4860054430d3712deb0d9487cac2fcb7d68':
hwcontext_qsv: do not fail when download/upload VPP session creation fails
Merged-by: Mark Thompson <sw@jkqxz.net>
This was skipped in c17563c5d3 because
it depended on the filter setup merge, but was forgotten after that
actually happened.
Fixes hwaccel fate for stream size change tests.
When decoding with threads enabled, the get_format callback will be
called with one of the per-thread codec contexts rather than with the
outer context. If a hwaccel is in use too, this will add a reference
to the hardware frames context on that codec context, which will then
propagate to all of the other per-thread contexts for decoding. Once
the decoder finishes, however, the per-thread contexts are not freed
normally, so these references leak.
Merges Libav commit fd0fae60.
Signed-off-by: wm4 <nfxjfg@googlemail.com>
* commit 'ad71d3276fef0ee7e791e62bbfe9c4e540047417':
lavfi: add a QSV deinterlacing filter
Minor fixup for lavfi differences.
Merged-by: Mark Thompson <sw@jkqxz.net>
* commit '0956fd460681e8ccbdae19f135f0d3970bf95c2f':
qsvenc: do not re-execute encoding on all positive status codes
Noop, see fb240a6276.
Merged-by: Mark Thompson <sw@jkqxz.net>
* commit 'd9ec3c60143babe1bb77c268e1d5547d15acd69b':
qsvenc: take only the allocated dimensions from the frames context
Merged-by: Mark Thompson <sw@jkqxz.net>
Only do this when building for a recent VAAPI version - initial
driver implementations were confused about the interpretation of the
framerate field, but hopefully this will be consistent everywhere
once 0.40.0 is released.
(cherry picked from commit ff35aa8ca4)
Default to using VBR when a target bitrate is set, unless the max rate
is also set and matches the target. Changes to the Intel driver mean
that min_qp is also respected in this case, so set a codec default to
unset the value rather than using the current default inherited from
the MPEG-4 part 2 encoder.
(cherry picked from commit eddfb57210)
This includes a backward-compatibility hack to choose CBR anyway on
old drivers which have no CBR support, so that existing programs will
continue to work their options now map to VBR.
(cherry picked from commit f033ba470f)
Before this change, it was possible to overflow pic_order_cnt_lsb and
generate a stream with invalid POC numbering. This makes sure that
the field is large enough that a single IDR B* P sequence uses fewer
than half the available POC lsb values.
(cherry picked from commit 89725a8512)
This change makes the configured GOP size be respected exactly -
previously the value could be exceeded slightly due to flaws in the
frame type selection logic.
(cherry picked from commit 37fab0661a)
The original code is correctly following the API - vaTerminate() must
be called to free the resources of a VADisplay after it is created by
any of the vaGetDisplay*() calls; it is not necessary to have
successfully called vaInitialize() on it. The segfaults which
prompted this change must therefore be bugs in libva or the driver it
loads.
This reverts commit 3606602f11.
Moves much of the setup logic for VAAPI decoding into lavc; the user
now need only provide the hw_frames_ctx.
(cherry picked from commit 123ccd07c5)
(cherry picked from commit 5e879b54a3)
(cherry picked from commit 0aec37e625)
(cherry picked from commit cfa4eb4fba)
Before this change, it was possible to overflow pic_order_cnt_lsb and
generate a stream with invalid POC numbering. This makes sure that
the field is large enough that a single IDR B* P sequence uses fewer
than half the available POC lsb values.
This change makes the configured GOP size be respected exactly -
previously the value could be exceeded slightly due to flaws in the
frame type selection logic.
This was not observed earlier because the only syntax element which
it normally misses with the current setup is slice_qp_delta, but that
is always going to be zero (in IDR frames QP isn't varied on the
slice) which will always exp-golomb code as a single 1 bit. The
immediately following part is the byte alignment, which is always a 1
bit followed by 0s which are ignored, so as long as the bitstream is
never aligned at that point we will never notice because the only
difference is that an ignored bit is a 1 instead of a 0.
(cherry picked from commit fc30a90898)
Mark Thompson
Mark ThompsonandMark Thompson