Skip to content

Commits

History for gst-dsp/tidsp

Commits on Feb 16, 2012

  1. jpeg: fixes for old socket-node 

    Looks like there's a difference between L23.i3.3 and L23.i3.4, but since
the latter was never released publicly, let's assume SN_API=1 means the
former.

Unfortunately, the socket-node doesn't allow more than one buffer, which
would hang some pipelines. We could disable use_pinned, but that's going
backwards, as use_pinned must go away.

Maybe there's a way to 'unhang' the pipeline somehow, but for now this
is all we can do.

Signed-off-by: Felipe Contreras <felipe.contreras@nokia.com>
    Felipe Contreras committed Feb 16, 2012
    20d0895

Commits on Jan 31, 2012

  1. Trivial cleanups for 'len' 

    It's updated now by dmm buffer allocators.

Signed-off-by: Felipe Contreras <felipe.contreras@nokia.com>
    Felipe Contreras committed Jan 31, 2012
    e9c2e94

  2. Change dmm_buffer size/len semantics a bit 

    Basically, 'len' is supposed to have the size of the data contained,
whereas 'size' is the size of the actual buffer (with padding).

The problem is that before 'size' didn't have any padding, so 'len' was
automatically filled right before sending the buffer based on this
'size', but now we don't have the unpadded 'size' any more, thanks to
f2b6efb[1].

We really need this padded size, so instead of using yet another field,
use 'len' which is unused from buffer allocation up to sending the
buffer.

This means that some code that used 'size' now should be using 'len'
instead.

Obviously this is a tricky change prone to trigger many issues, so
should be tested thorougly.

[1] dmm_buffer: store correct size (aligned)

Signed-off-by: Felipe Contreras <felipe.contreras@nokia.com>
    Felipe Contreras committed Jan 31, 2012
    7e8be51

  3. tidsp/*dec: fix all the lengths of decoder output 

    Should help with framesizes that are not multiples of 16.

Signed-off-by: Felipe Contreras <felipe.contreras@nokia.com>
    Felipe Contreras committed Jan 31, 2012
    6a1fbcb

  4. Add gstdsp_vdec_len_fixup() helper 

    Other codecs would/should benefit from this.

Signed-off-by: Felipe Contreras <felipe.contreras@nokia.com>
    Felipe Contreras committed Jan 31, 2012
    088504d

  5. tidsp: mp4venc: ensure availability of codec data 

    Use malloc and memcpy to be sure of the availability of the memory where
the codec data is stored.

Based on patch by Robert Swain.

Signed-off-by: Mark Nauwelaerts <mark.nauwelaerts@collabora.co.uk>
Signed-off-by: Felipe Contreras <felipe.contreras@nokia.com>
    Mark Nauwelaerts authored and Felipe Contreras committed Jan 31, 2012
    ea85e2b

  6. tidsp: mp4venc: fix codec_data leak 

    Signed-off-by: Mark Nauwelaerts <mark.nauwelaerts@collabora.co.uk>
Signed-off-by: Felipe Contreras <felipe.contreras@nokia.com>
    Mark Nauwelaerts authored and Felipe Contreras committed Jan 31, 2012
    8b72633

Commits on Nov 21, 2011

  1. tidsp: add H.264 parsing helper

    @felipec
    Marco Ballesio authored and felipec committed Nov 21, 2011
    cc3891f

  2. Move codec_data to base 

    Will be useful in the next commits.

Signed-off-by: Felipe Contreras <felipe.contreras@nokia.com>
    @felipec
    Felipe Contreras authored and felipec committed Nov 21, 2011
    b6a657a

  3. wmvdec: don't append startcode when it exists 

    Signed-off-by: Felipe Contreras <felipe.contreras@gmail.com>
    @felipec
    Elamparithi Shanmugam authored and felipec committed Nov 21, 2011
    d3c35f3

  4. tidsp: check for badly formatted extra-data stream 

    Signed-off-by: Mark Nauwelaerts <mark.nauwelaerts@collabora.co.uk>
Signed-off-by: Felipe Contreras <felipe.contreras@nokia.com>
    @felipec
    Mark Nauwelaerts authored and felipec committed Nov 21, 2011
    6f0c109

Commits on Jul 19, 2011

  1. jpegenc: fix regression for newer SN_API versions 

    Signed-off-by: Felipe Contreras <felipe.contreras@nokia.com>
    Felipe Contreras committed Jul 19, 2011
    d89a045

Commits on Apr 27, 2011

  1. Trivial coverity cleanups 

    Signed-off-by: Felipe Contreras <felipe.contreras@nokia.com>
    Felipe Contreras committed Apr 27, 2011
    38f9fd8

  2. tidsp: mp4venc: only try to extract extra-data for MPEG-4 

    Not H.263.

Signed-off-by: Felipe Contreras <felipe.contreras@nokia.com>
    Felipe Contreras committed Apr 27, 2011
    bd0ce20

  3. tidsp: h264enc: always strip/ignore sps/pps 

    We cannot rely on the 'keyframe' marks from the codec.

Signed-off-by: Felipe Contreras <felipe.contreras@nokia.com>
    Felipe Contreras committed Apr 27, 2011
    b14cab1

  4. tidsp: h264enc: re-add intra frame interval 

    It turns out it was working.

Signed-off-by: Felipe Contreras <felipe.contreras@nokia.com>
    Felipe Contreras committed Apr 27, 2011
    68f507a

  5. Remove hack to manually force keyframes 

    It turns out we don't want IDR frames, only I slices.

Signed-off-by: Felipe Contreras <felipe.contreras@nokia.com>
    Felipe Contreras committed Apr 27, 2011
    10121b5

  6. tidsp: h264enc: move I frame interval to right place 

    This is probably what we wanted although it doesn't make any difference
right now since it's always 0.

Signed-off-by: Felipe Contreras <felipe.contreras@nokia.com>
    Felipe Contreras committed Apr 27, 2011
    a986f29

Commits on Mar 5, 2011

  1. tidsp: h264dec: trivial cleanup 

    Signed-off-by: Felipe Contreras <felipe.contreras@nokia.com>
    Felipe Contreras committed Mar 5, 2011
    335cf2b

  2. td_h264dec: do not leak transformed codec data 

    Signed-off-by: Mark Nauwelaerts <mark.nauwelaerts@collabora.co.uk>
Signed-off-by: Felipe Contreras <felipe.contreras@nokia.com>
    Mark Nauwelaerts authored and Felipe Contreras committed Mar 5, 2011
    f731a3b

  3. Trivial cleanups 

    Signed-off-by: Felipe Contreras <felipe.contreras@nokia.com>
    Felipe Contreras committed Mar 5, 2011
    d037157

  4. Move 'keyframe' field from dmm_buffer to td_buffer 

    Signed-off-by: Felipe Contreras <felipe.contreras@nokia.com>
    Felipe Contreras committed Mar 5, 2011
    0b50634

  5. Move 'user_data' field from dmm_buffer to td_buffer 

    Signed-off-by: Felipe Contreras <felipe.contreras@nokia.com>
    Felipe Contreras committed Mar 5, 2011
    5da3561

  6. Use td_buffer in {send,recv}_cb 

    Signed-off-by: Felipe Contreras <felipe.contreras@nokia.com>
    Felipe Contreras committed Mar 5, 2011
    ea070f3

Commits on Feb 20, 2011

  1. tidsp: fix JPEG encoder for old SN 

    Signed-off-by: Felipe Contreras <felipe.contreras@gmail.com>
    @felipec
    felipec committed Feb 20, 2011
    3146cbb

Commits on Jan 27, 2011

  1. Fix compilation warnings 

    And other trivial stuff.

Signed-off-by: Felipe Contreras <felipe.contreras@nokia.com>
    Felipe Contreras committed Jan 27, 2011
    74d4d35

  2. adec: aac decoder implementation 

    This is the first approach of the AAC decoder using the SN
mpeg4aacdec_sn.dll64P release 23.i3.3 and 23.i3.8

The performance of a simple pipeline goes among 10% and 0% of CPU
usage (montored with top and with gstreamer 0.10.31)

My testing pipeline is something like:

gst-launch filesrc location=test.aac ! aacparse ! dspadec ! alsasink

Issues: when a parametric stereo media is found, a downstream caps
renegotiation is done, but somehow (I don't understand why) the
renegotiation is not honored and the output is rendered wrong (because
of the reduced framerate).

Signed-off-by: Víctor Manuel Jáquez Leal <vjaquez@igalia.com>
Signed-off-by: Felipe Contreras <felipe.contreras@nokia.com>
    @ceyusa
    ceyusa authored and Felipe Contreras committed Jan 27, 2011
    719830c

Commits on Dec 28, 2010

  1. vpp: add codec info 

    Signed-off-by: Felipe Contreras <felipe.contreras@nokia.com>
    Felipe Contreras committed Dec 28, 2010
    3364bcf

  2. td_jpegdec: compute the correct structure size 

    Because of the memory alignment, the create_args structure in jpegdec reports
a size of 36, although its real size is 34 (1 uint32_t and 15 uint16_t).

As the DSP doesn't have that memory alignment, the given structure size is
wrong and returns an error when creating the node.

This patch calculates the correct size for the DSP.

Signed-off-by: Víctor Manuel Jáquez Leal <vjaquez@igalia.com>
Signed-off-by: Felipe Contreras <felipe.contreras@gmail.com>
    @ceyusa
    ceyusa authored and Felipe Contreras committed Dec 28, 2010
    fd5b8c2

Commits on Dec 1, 2010

  1. tidsp: h.264 dec: trivial fix for I420 

    Signed-off-by: Felipe Contreras <felipe.contreras@nokia.com>
    Felipe Contreras committed Dec 1, 2010
    988f406

Commits on Nov 24, 2010

  1. venc: decouple H.264 mode and intra-refresh 

    This seems to be more proper and makes mode=1 intra-refresh=0 work.

Signed-off-by: Felipe Contreras <felipe.contreras@nokia.com>
    Juha Alanen authored and Felipe Contreras committed Nov 24, 2010
    cf1a741

  2. tidsp: h264enc: remove dead code 

    There's a check just above for 'self->priv.h264.bytestream', and it's
true, the function returns.

Signed-off-by: Felipe Contreras <felipe.contreras@nokia.com>
    Felipe Contreras committed Nov 24, 2010
    f35cc3a

Commits on Nov 23, 2010

  1. tidsp: generic cleanups 

    Shuffle some code, remove prefixes, coding-style, etc.

No functional changes.

Signed-off-by: Felipe Contreras <felipe.contreras@nokia.com>
    Felipe Contreras committed Nov 23, 2010
    2779b42

  2. Create tidsp library, and split codecs 

    Now they are finally modular.

Signed-off-by: Felipe Contreras <felipe.contreras@nokia.com>
    Felipe Contreras committed Nov 23, 2010
    310728a