2015-10-28 17:00 GMT+01:00 Emil Velikov :
> Hi Gwenole,
>
> On 28 October 2015 at 15:46, Gwenole Beauchesne wrote:
>> Statistics based on precise commits and actual contents (subsytems,
>> features, etc.). List of contributions could be derived into three
>> parts: pr
Statistics based on precise commits and actual contents (subsytems,
features, etc.). List of contributions could be derived into three
parts: primary/major (100+), regular (15+) and occasional.
Signed-off-by: Gwenole Beauchesne
---
FYI, this will be pushed by the end of this week.
AUTHORS
for instance.
BTW, why would you need to do that -- I mean, filling a surface with
black pixels?
Regards,
--
Gwenole Beauchesne
Intel Corporation SAS / 2 rue de Paris, 92196 Meudon Cedex, France
Registration Number (RCS): Nanterre B 302 456 199
___
Libva
0, 0x25003ae8, 0x3e8d0500, 0xc300 },
> { 0x0081, 0x24003ae0, 0x008d0480, 0x },
> @@ -285,6 +291,7 @@
> { 0x0061, 0x24e022e8, 0x00ae8451, 0x },
> { 0x0061, 0x250022e8, 0x00ae8041, 0x0000 },
> { 0x0061, 0x252022e8, 0x00ae8051,
rs/post_processing/gen8/YUV_to_RGB.g8a | 16
>>> src/shaders/post_processing/gen8/pl2_to_rgbx.g8b | 16
>>> src/shaders/post_processing/gen9/pl2_to_rgbx.g9b | 16
>>> 6 files changed, 96 insertions(+)
>>
>&
hments is strictly
> prohibited. If you have received this transmittal and/or attachments in
> error, please notify me immediately by reply email or telephone and
> immediately delete this message and all its attachments. Thank you.
>
>
> _
Allow creating VA surfaces with userptr allocated buffers. This requires
a recent enough version of libdrm (>= 2.4.57), but also a kernel (>= 3.16)
which contains appropriate fixes for userptr.
v2: only request synchronized mappings (Chris Wilson).
Signed-off-by: Gwenole Beau
Hi,
This patch series allows for importing userptr memory buffers into VA
surfaces. This requires a recent enough libdrm version (>= 2.4.57) and
also kernel (>= 3.16) with appropriate userptr fixes. The capabilities
are tentatively checked for at run-time.
Regards,
Gwenole Beauche
Fix possible buffer overflow when addressing the array of buffer handles
to be imported into VA surfaces.
Signed-off-by: Gwenole Beauchesne
---
src/i965_drv_video.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/src/i965_drv_video.c b/src/i965_drv_video.c
index e2abc1b
Shuffle the internal memory type ids by one, and turn them into clean
enum values. Also make sure to validate the VA surface memory type
early enough.
Signed-off-by: Gwenole Beauchesne
---
src/i965_drv_video.c | 5 +
src/i965_drv_video.h | 8 +---
2 files changed, 10 insertions(+), 3
Allow creating VA surfaces with userptr allocated buffers. This requires
a recent enough version of libdrm (>= 2.4.57), but also a kernel (>= 3.16)
which contains appropriate fixes for userptr.
Signed-off-by: Gwenole Beauchesne
---
configure.ac | 19
src/i965_drv_v
Hi,
This patch series allows for importing userptr memory buffers into VA
surfaces. This requires a recent enough libdrm version (>= 2.4.57) and
also kernel (>= 3.16) with appropriate userptr fixes. The capabilities
are tentatively checked for at run-time.
Regards,
Gwenole Beauche
Shuffle the internal memory type ids by one, and turn them into clean
enum values. Also make sure to validate the VA surface memory type
early enough.
Signed-off-by: Gwenole Beauchesne
---
src/i965_drv_video.c | 5 +
src/i965_drv_video.h | 8 +---
2 files changed, 10 insertions(+), 3
Fix possible buffer overflow when addressing the array of buffer handles
to be imported into VA surfaces.
Signed-off-by: Gwenole Beauchesne
---
src/i965_drv_video.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/src/i965_drv_video.c b/src/i965_drv_video.c
index e2abc1b
Hi,
2015-03-24 5:40 GMT+01:00 Xiang, Haihao :
> On Fri, 2015-03-20 at 16:09 +0100, Gwenole Beauchesne wrote:
>> Add support for "low-power" mode expressed with VA_PROC_PIPELINE_FAST
>> flag set to VAProcPipelineParameterBuffer.pipeline_flags. The purpose
>> is t
to
perform format conversion, scaling and basic bob-deinterlacing.
Signed-off-by: Gwenole Beauchesne
---
src/i965_post_processing.c | 239 +
1 file changed, 239 insertions(+)
diff --git a/src/i965_post_processing.c b/src/i965_post_processing.c
ind
From: "Xiang, Haihao"
Signed-off-by: Xiang, Haihao
---
src/i965_post_processing.c | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/src/i965_post_processing.c b/src/i965_post_processing.c
index 2c01e43..72f298f 100755
--- a/src/i965_post_processing.c
+++ b/src/i965_post_pr
The original current_out_surface was never released on exit. Main
reason for that is the legacy VPP framework that did not allow the
VADriverContextP handle to be passed down to the desired .finalize()
hook. Improved that to bring it on par with the VEBOX code path.
Signed-off-by: Gwenole
This patch series brings the Sandybridge, Ivybridge and Baytrail code
path for advanced deinterlacing on par with what we now have for VEBOX
(Haswell and newer).
Gwenole Beauchesne (3):
vpp: fix memory leak in DNDI code path.
vpp: fix advanced deinterlacing on Sandybridge and Ivybridge.
vpp
This fixes support for Motion Adaptive deinterlacing mode on both
Sandybridge and Ivybridge platforms. In particular, correct field
ordering is now supported, and STMM ping-pong buffering is added.
v2: changed STMM surface format to Y800 for a single plane.
Signed-off-by: Gwenole Beauchesne
Signed-off-by: Gwenole Beauchesne
---
src/i965_device_info.c | 1 +
src/i965_post_processing.c | 4 +++-
2 files changed, 4 insertions(+), 1 deletion(-)
diff --git a/src/i965_device_info.c b/src/i965_device_info.c
index 9227d0b..13e3daa 100644
--- a/src/i965_device_info.c
+++ b/src
//For VAAPI, the VAQMatrixBuffer needs to be in zigzag order.
> +//The App should send it in zigzag. Now, the driver has to extract
> the raster from it.
> for (j = 0; j < 64; j++)
> raster_qm[zigzag_direct[j]] =
> qmatrix->chroma_quantiser_mat
ntioned above, it's just a matter of using a VA/VPP pipeline
with filter_flags set accordingly.
Regards,
--
Gwenole Beauchesne
Intel Corporation SAS / 2 rue de Paris, 92196 Meudon Cedex, France
Registration Number (RCS): Nanterre B 302 456 199
___
Libva
solution requires a second patch to make
codec_info really per-codec. i.e. max resolution constraints can vary
on a per-codec basis, depending on the generation.
Regards,
--
Gwenole Beauchesne
Intel Corporation SAS / 2 rue de Paris, 92196 Meudon Cedex, France
Registration Number (RC
ng your canvas, you'd
stick to OGL, but I believe VA subpictures could suit your needs.
> If there's sample code on how to do this, that would be very helpful.
mplayer-vaapi, gstreamer-vaapi for VA subpictures composition.
<https://gitorious.org/vaapi/mplayer> ("hwaccel-vaa
M
> To: Varga, Michael
> Cc: Emil Velikov; Gwenole Beauchesne; libva@lists.freedesktop.org
> Subject: RE: [Libva] mpeg4 decoding
>
> On Mon, 2015-01-12 at 21:38 +, Varga, Michael wrote:
>> Haihao,
>>
>> I use modulo_time_base when reconstructing the head
0007fffdd97c02a in gen75_vpp_gpe (proc_ctx=0x7fffb0256c70,
> ctx=0x9c2ff0) at gen75_picture_process.c:105
> va_status = 0
> #8 gen75_proc_picture (ctx=0x9c2ff0, profile=VAProfileNone,
> codec_state=0x9c6a28, hw_context=0x7fffb0256c70) at
> gen75_picture_process.c:201
>
,
rather than going piecemal. :)
In particular, I believe you would also need the actual
vop_time_increment bits, right?
Thanks,
--
Gwenole Beauchesne
Intel Corporation SAS / 2 rue de Paris, 92196 Meudon Cedex, France
Registration Number (RCS): Nanterre B 302 456 199
___
This fixes support for Motion Adaptive deinterlacing mode on both
Sandybridge and Ivybridge platforms. In particular, correct field
ordering is now supported, and STMM ping-pong buffering is added.
v2: changed STMM surface format to Y800 for a single plane.
Signed-off-by: Gwenole Beauchesne
Hi,
This patch series brings the Sandybridge, Ivybridge and Baytrail code
path for advanced deinterlacing on par with what we now have for VEBOX
(Haswell and newer).
Another patch series would come next to handle one extra future
reference frame.
Regards,
Gwenole Beauchesne (3):
vpp: fix
Signed-off-by: Gwenole Beauchesne
---
src/i965_device_info.c | 1 +
src/i965_post_processing.c | 4 +++-
2 files changed, 4 insertions(+), 1 deletion(-)
diff --git a/src/i965_device_info.c b/src/i965_device_info.c
index a415587..8d9db88 100644
--- a/src/i965_device_info.c
+++ b/src
The original current_out_surface was never released on exit. Main
reason for that is the legacy VPP framework that did not allow the
VADriverContextP handle to be passed down to the desired .finalize()
hook. Improved that to bring it on par with the VEBOX code path.
Signed-off-by: Gwenole
If scaling parameters don't change, i.e. if the same scaling algorithm
and factors are used, there is no point in calculating the filter
coefficients again. So, just cache them into the existing AVS context.
Signed-off-by: Gwenole Beauchesne
---
src/gen8_post_processing.c
was
never set. Dropped the associated dead code, which was never used. And,
anyway, advanced deinterlacing modes shall go through the explcit VPP
interfaces.
Signed-off-by: Gwenole Beauchesne
---
src/gen75_vpp_vebox.c | 6 +++---
src/gen8_render.c | 6 +++---
src/i965_output_dri.c
Fix AVS filter coefficients for Broadwell. At least, we now address
the expected entry in the table but the actual value is still not
viable enough.
Signed-off-by: Gwenole Beauchesne
---
src/gen8_post_processing.c | 83 ++
src/i965_structs.h
: Gwenole Beauchesne
---
src/i965_vpp_avs.c | 86 ++
1 file changed, 86 insertions(+)
diff --git a/src/i965_vpp_avs.c b/src/i965_vpp_avs.c
index 12f1dfe..2dfaecd 100644
--- a/src/i965_vpp_avs.c
+++ b/src/i965_vpp_avs.c
@@ -52,6 +52,72
ions ("hsw"), and to the legacy platforms
down to Ironlake ("ilk").
Signed-off-by: Gwenole Beauchesne
---
src/i965_post_processing.c | 21 +++--
src/i965_structs.h | 25 ++---
2 files changed, 33 insertions(+), 13
Honour advanced video scaling. i.e. propagate vaPutSurface() scaling
flags, but also VPP filter flags. Also enable the sharp 8x8 filter for
high-quality scaling options, while adaptive video scaling is disabled
(bypassed) for now.
Signed-off-by: Gwenole Beauchesne
---
src/gen75_vpp_vebox.c
: Gwenole Beauchesne
---
src/Makefile.am| 2 +
src/gen8_post_processing.c | 101 +
src/i965_post_processing.c | 218 ++---
src/i965_vpp_avs.c | 110 +++
src/i965_vpp_avs.h | 85
Make sure the newly calculated filter coefficients fall into the HW
accepted range of values. This normally should not be an issue with
the current configuration / scaling algorithms though.
Signed-off-by: Gwenole Beauchesne
---
src/gen8_post_processing.c | 15 +++
src
: Gwenole Beauchesne
---
src/Makefile.am| 2 +
src/gen8_post_processing.c | 101 +
src/i965_post_processing.c | 218 ++---
src/i965_vpp_avs.c | 109 +++
src/i965_vpp_avs.h | 85
2014-10-13 19:27 GMT+02:00 Gwenole Beauchesne :
> Honour advanced video scaling. i.e. propagate vaPutSurface() scaling
> flags, but also VPP filter flags. Also enable the sharp 8x8 filter for
> high-quality scaling options, while adaptive video scaling is disabled
> (bypassed) for now
2014-10-13 19:27 GMT+02:00 Gwenole Beauchesne :
> If scaling parameters don't change, i.e. if the same scaling algorithm
> and factors are used, there is no point in calculating the filter
> coefficients again. So, just cache them into the existing AVS context.
>
> Si
2014-10-13 19:27 GMT+02:00 Gwenole Beauchesne :
> Add support for high-quality scaling during video processing. This is
> enabled with the VA_FILTER_SCALING_HQ filter flag. The algorithm used
> for that is based on a Lanczos convolution kernel: 3 lobes on either
> side for luma sa
confident enough experiment. More testing is needed, especially on
other platforms. Fully adaptive mode is not programmed yet.
Patched sources available for convenience here:
<https://github.com/gbeauchesne/libva-intel-driver/tree/18.vpp.avs>
Regards,
Gwenole Beauchesne (8):
vpp: fix adaptive filt
Make sure the newly calculated filter coefficients fall into the HW
accepted range of values. This normally should not be an issue with
the current configuration / scaling algorithms though.
Signed-off-by: Gwenole Beauchesne
---
src/gen8_post_processing.c | 15 +++
src
If scaling parameters don't change, i.e. if the same scaling algorithm
and factors are used, there is no point in calculating the filter
coefficients again. So, just cache them into the existing AVS context.
Signed-off-by: Gwenole Beauchesne
---
src/gen8_post_processing.c
Honour advanced video scaling. i.e. propagate vaPutSurface() scaling
flags, but also VPP filter flags. Also enable the sharp 8x8 filter for
high-quality scaling options, while adaptive video scaling is disabled
(bypassed) for now.
Signed-off-by: Gwenole Beauchesne
---
src/gen75_vpp_vebox.c
was
never set. Dropped the associated dead code, which was never used. And,
anyway, advanced deinterlacing modes shall go through the explcit VPP
interfaces.
Signed-off-by: Gwenole Beauchesne
---
src/gen75_vpp_vebox.c | 10 +-
src/gen8_render.c | 6 +++---
src
Signed-off-by: Gwenole Beauchesne
---
src/i965_post_processing.c | 21 +++--
src/i965_structs.h | 25 ++---
2 files changed, 33 insertions(+), 13 deletions(-)
diff --git a/src/i965_post_processing.c b/src/i965_post_processing.c
index 6d435a8..4d3c9b1 10
Fix AVS filter coefficients for Broadwell. At least, we now address
the expected entry in the table but the actual value is still not
viable enough.
Signed-off-by: Gwenole Beauchesne
---
src/gen8_post_processing.c | 83 ++
src/i965_structs.h
: Gwenole Beauchesne
---
src/i965_vpp_avs.c | 86 ++
1 file changed, 86 insertions(+)
diff --git a/src/i965_vpp_avs.c b/src/i965_vpp_avs.c
index 39407ac..33d6eba 100644
--- a/src/i965_vpp_avs.c
+++ b/src/i965_vpp_avs.c
@@ -52,6 +52,72
= 0
> write(2, "libva info: Trying to open /usr/"..., 75libva info: Trying to
> open /usr/lib/x86_64-linux-gnu/dri/i965_drv_video.so
> ) = 75
>
> Regards,
>
> Thomas
> This message is subject to the following terms and conditions: MA
> element's name and version is as follows: libgstreamer-vaapi1.0-0 and
> version: 0.5.7-0ubuntu4
>
>
> Could you please help to decode an rtsp stream on GPU by using gstreamer
> vaapi plugins?
>
> Regards.
>
>
> --
> *Engin FIRAT*
> Adoniss Yazılım Bilişim
> E
Poulsbo), and subsequently adopted by the major
multimedia frameworks.
Signed-off-by: Gwenole Beauchesne
---
va/va.h | 14 ++
va/va_dec_jpeg.h | 2 ++
2 files changed, 16 insertions(+)
diff --git a/va/va.h b/va/va.h
index 127ad49..01694a9 100644
--- a/va/va.h
+++ b/va/va.h
r 1.4.0 release. OK?
> On Wed, Sep 24, 2014 at 1:40 AM, Gwenole Beauchesne
> wrote:
>> Hi,
>>
>> This patch series adds support for DRM Render-Nodes through the
>> existing vaGetDisplayDRM() interface.
>>
>> Practical use-case: this allows for remo
2014-09-25 8:27 GMT+02:00 Zhao, Yakui :
> On Wed, 2014-09-24 at 23:35 -0600, Gwenole Beauchesne wrote:
>> Hi,
>>
>> 2014-09-25 3:00 GMT+02:00 Zhao, Yakui :
>> > On Wed, 2014-09-24 at 03:10 -0600, Gwenole Beauchesne wrote:
>> >> Hi,
>> >>
>
Hi,
2014-09-25 3:00 GMT+02:00 Zhao, Yakui :
> On Wed, 2014-09-24 at 03:10 -0600, Gwenole Beauchesne wrote:
>> Hi,
>>
>> Great to see that finally fixed.
>>
>> 2014-09-24 9:13 GMT+02:00 Zhao Yakui :
>> > On some systems there is no access to /proc/cp
2014-09-24 10:40 GMT+02:00 Gwenole Beauchesne :
> Signed-off-by: Gwenole Beauchesne
> ---
> test/common/va_display_drm.c | 26 +-
> 1 file changed, 21 insertions(+), 5 deletions(-)
>
> diff --git a/test/common/va_display_drm.c b/test/common/va_dis
&rdata[0], &rdata[1], &rdata[2], &rdata[3]);
> +cpuid(0x8003, &rdata[4], &rdata[5], &rdata[6], &rdata[7]);
> +cpuid(0x8004, &rdata[8], &rdata[9], &rdata[10], &rdata[11]);
> +
> +*(model_id + 48) = '\0';
That&
DRM Render-Node
devices to be supported, which is an important enough improvement.
Tested with gstreamer-vaapi and another proprietary software. This
should be landed to the libva 1.4.0 relese too.
Andrey Larionov (1):
drm: add support for render nodes.
Gwenole Beauchesne (1):
tests: try a
).
Signed-off-by: Andrey Larionov
[checked if fd is a render-node, fixed VA_DISPLAY_DRM_RENDERNODES]
Signed-off-by: Gwenole Beauchesne
---
va/drm/va_drm.c | 28 +---
va/drm/va_drm_utils.c | 21 +
va/drm/va_drm_utils.h | 10 ++
va
Signed-off-by: Gwenole Beauchesne
---
test/common/va_display_drm.c | 26 +-
1 file changed, 21 insertions(+), 5 deletions(-)
diff --git a/test/common/va_display_drm.c b/test/common/va_display_drm.c
index aa9f60a..d427984 100644
--- a/test/common/va_display_drm.c
+++ b
2014-09-18 18:45 GMT+02:00 Emil Velikov :
> The va backend does not use any of the libva functions.
> As such there is no need to link against it.
>
> Signed-off-by: Emil Velikov
> ---
> src/Makefile.am | 1 -
> 1 file changed, 1 deletion(-)
>
> diff --git a/src/Makefile.am b/src/Makefile.am
> in
ihao?
>
>> Thanks,
>>
>> Sean
>>
>>> Any input, suggestions and comments are appreciated.
>>>
>>> Cheers,
>>> Emil
>>>
>>> ___
>>> Libva mailing list
>>> Libva@lists.freedesktop.org
>>> http://lists.fr
version of libva < 1.0.x where x is to be determined, or
(ii) the pkg-config check failed somehow. (i) cannot occur because you
cannot even build the current VA intel-driver git master branch
against libva 1.3 I'd guess. So you are left with (ii) and another fix
might be needed then. i.e. w
@ stamp-h1
> /test/egl/va_egl
> /test/encode/avcenc
> /test/encode/h264encode
> +/test/encode/mpeg2vaenc
> /test/putsurface/putsurface
> /test/putsurface/putsurface_wayland
> /test/transcode/mpeg2transcode
OK for all branches.
Thanks,
--
Gwenole Beauchesne
Intel Corporati
2014-09-15 2:24 GMT+02:00 Zhao, Yakui :
> On Thu, 2014-09-11 at 20:20 -0600, Zhao, Yakui wrote:
>> V1->V2: Refine the call back function name from hw_codec_hook to
>> preinit_hw_codec
>> And it is called after VADriverContext is fully initialized. This is based
>>
Hi,
2014-09-09 7:27 GMT+02:00 Antti Seppälä :
> On 9 September 2014 06:36, Gwenole Beauchesne wrote:
>> Hi,
>>
>>
>> 2014-09-08 15:18 GMT+02:00 Antti Seppälä :
>>> On 27 August 2014 14:50, Gwenole Beauchesne wrote:
>>>> Hi,
>>>>
>&g
.
This is *not* strictly needed for 1.4-branch since we no longer print
the git version string in release builds.
Regards,
Gwenole.
2014-09-02 10:13 GMT+02:00 Gwenole Beauchesne :
> Signed-off-by: Gwenole Beauchesne
> ---
> configure.ac |5 +
> src/Makefi
Pushed to git master and staging branches.
This needs to go to 1.4-branch too.
2014-09-09 15:21 GMT+02:00 Gwenole Beauchesne :
> If the intel-gen4asm tool is not available, ship with the pre-built
> EU kernels instead of trying to regenerate them. In particular, just
> don't ex
If the intel-gen4asm tool is not available, ship with the pre-built
EU kernels instead of trying to regenerate them. In particular, just
don't expose the build rules if intel-gen4asm is not installed.
Signed-off-by: Gwenole Beauchesne
---
src/shaders/post_processing/gen7/Makefile.am |
Pushed to git master and staging branches.
This needs to go to 1.4-branch too.
Thanks,
Gwenole
2014-09-08 19:10 GMT+02:00 Gwenole Beauchesne :
> Allow for vaQuerySurfaceAttributes() to return BGRA and BGRX formats
> for VPP on Ivybridge and Haswell. This is supported as both source
> a
Hi,
2014-09-08 15:18 GMT+02:00 Antti Seppälä :
> On 27 August 2014 14:50, Gwenole Beauchesne wrote:
>> Hi,
>>
>> This patch series improves support for advanced deinterlacing on
>> Haswell and newer generation processors that support the VEBOX
>> block. The ne
truct buffer_store));
>> assert(buffer_store);
>
>
> ___
> Libva mailing list
> Libva@lists.freedesktop.org
> http://lists.freedesktop.org/mailman/listinfo/libva
--
Gwenole Beauchesne
Intel Corporation SAS / 2 rue
Allow for vaQuerySurfaceAttributes() to return BGRA and BGRX formats
for VPP on Ivybridge and Haswell. This is supported as both source
and target surface formats.
This fixes VA/EGL interop on Gen7 processors when a BGR[AX] surface
is exported into an EGLImage.
Signed-off-by: Gwenole Beauchesne
Move out generation of the vendor string to its specific helper function,
while also making it more robust and aware of possible overflows.
Signed-off-by: Gwenole Beauchesne
---
src/i965_drv_video.c | 47 +++
1 file changed, 35 insertions(+), 12
Signed-off-by: Gwenole Beauchesne
---
configure.ac |5 +
src/Makefile.am | 39 +--
src/i965_drv_version.h.in | 36
src/i965_drv_video.c |7 +++
4 files changed, 85 insertions
ounted for MVC/SVC */
> skip_cnt += 3;
> }
> --
> 1.9.1
>
> ___
> Libva mailing list
> Libva@lists.freedesktop.org
> http://lists.freedesktop.org/mailman/listinfo/libva
Thanks,
--
Gwenole Beauchesne
Intel Corporation SA
Hi,
2014-08-29 2:44 GMT+02:00 Zhao, Yakui :
> On Thu, 2014-08-28 at 02:35 -0600, Gwenole Beauchesne wrote:
>> Hi Yakui,
>>
>> 2014-08-28 3:16 GMT+02:00 Zhao, Yakui :
>> > On Wed, 2014-08-27 at 05:50 -0600, Gwenole Beauchesne wrote:
>> >> Silence the
Hi Yakui,
2014-08-28 3:16 GMT+02:00 Zhao, Yakui :
> On Wed, 2014-08-27 at 05:50 -0600, Gwenole Beauchesne wrote:
>> Silence the following compilation warning:
>> CC i965_drv_video_la-gen75_vpp_vebox.lo
>> gen75_vpp_vebox.c: In function 'bdw_veb_dndi_iecp_command
: Gwenole Beauchesne
---
src/gen75_vpp_vebox.c | 101 ++---
src/gen75_vpp_vebox.h |4 +-
2 files changed, 91 insertions(+), 14 deletions(-)
diff --git a/src/gen75_vpp_vebox.c b/src/gen75_vpp_vebox.c
index b060ae2..7e37d9c 100644
--- a/src
If IECP is enabled, for instance when color conversion is performed
or ProcAmp adjustments are applied, the ultimate denoised output with
additional processing operations applied is the Current Output frame,
not the plain Current Denoised Output frame.
Signed-off-by: Gwenole Beauchesne
---
src
Rename FRAME_STORE_SUM to FRAME_STORE_COUNT, use existing macros to
determine the number of elements in the frame store array, avoid
duplicate zero initializations.
Signed-off-by: Gwenole Beauchesne
---
src/gen75_vpp_vebox.c | 10 +++---
src/gen75_vpp_vebox.h |4 ++--
2 files changed
Clean-up frame store surface storage allocation, rename the helper
function to gen75_vebox_ensure_surfaces_storage() and make it more
robust to allocation errors. i.e. propagate them up right away.
Signed-off-by: Gwenole Beauchesne
---
src/gen75_vpp_vebox.c | 206
For each frame, the Intel HD Graphics hardware specification mandates
that the following commands should be submitted, in-order: VEBOX_STATE,
VEBOX_SURFACE_STATE(input), VEBOX_SURFACE_STATE(output), VEB_DI_IECP.
Signed-off-by: Gwenole Beauchesne
---
src/gen75_vpp_vebox.c | 10 --
1
Factor out initialization and validation of pipeline parameters.
In particular, introduce a new gen75_vebox_init_pipe_params() helper
function that initializes the filters mask and ensures the supplied
filters are going to be supported.
Signed-off-by: Gwenole Beauchesne
---
src
detection
code, thus allowing more multiple VEBOX calls with an additional format
conversion away.
Signed-off-by: Gwenole Beauchesne
---
src/gen75_vpp_vebox.c | 380 -
src/gen75_vpp_vebox.h | 13 +-
2 files changed, 165 insertions(+), 228
Signed-off-by: Gwenole Beauchesne
---
src/gen75_vpp_vebox.c | 14 +-
1 file changed, 5 insertions(+), 9 deletions(-)
diff --git a/src/gen75_vpp_vebox.c b/src/gen75_vpp_vebox.c
index 9d400d9..f2d64ac 100644
--- a/src/gen75_vpp_vebox.c
+++ b/src/gen75_vpp_vebox.c
@@ -964,7 +964,6
Always use Y-tiling in native surface format for performance reasons.
Signed-off-by: Gwenole Beauchesne
---
src/gen75_picture_process.c |2 +-
src/gen75_vpp_vebox.c |4 ++--
2 files changed, 3 insertions(+), 3 deletions(-)
diff --git a/src/gen75_picture_process.c b/src
e calculation of the VEB_DI_IECP::endingX variable
with existing helper macros.
Signed-off-by: Gwenole Beauchesne
---
src/gen75_vpp_vebox.c | 16 ++--
1 file changed, 6 insertions(+), 10 deletions(-)
diff --git a/src/gen75_vpp_vebox.c b/src/gen75_vpp_vebox.c
index 1113c90..f452
erspective, is VA_DEINTERLACING_BOTTOM_FIELD_FIRST.
https://bugs.freedesktop.org/show_bug.cgi?id=72518
https://bugs.freedesktop.org/show_bug.cgi?id=72522
Signed-off-by: Gwenole Beauchesne
---
src/gen75_vpp_vebox.c |2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/src/gen75_v
The arguments to vpp_surface_convert() were mixed up. i.e. both input
and output surfaces were reversed. Changed the vpp_surface_scaling()
arguments order as well to have more consistent helper functions.
Signed-off-by: Gwenole Beauchesne
---
src/gen75_vpp_vebox.c | 22
Use the existing VPP_{DNDI,IECP}_xxx flags instead of magic numbers.
Signed-off-by: Gwenole Beauchesne
---
src/gen75_vpp_vebox.c | 16
src/gen75_vpp_vebox.h |2 ++
2 files changed, 10 insertions(+), 8 deletions(-)
diff --git a/src/gen75_vpp_vebox.c b/src
eries applies to the "staging" branch. Applying to "master"
is straightforward too. A convenience repository for the latter is
available here:
<https://github.com/gbeauchesne/libva-intel-driver/tree/17.vpp.vebox>
Regards,
Gwenole Beauchesne (13):
vebox: silence compilati
.freedesktop.org/mailman/listinfo/libva
This looks OK as a last resort solution, but it should be completed
with an earlier check in vaQueryVideoProcPipelineCaps(). This is the
function used to validate the filter chain somehow.
Besides, I believe we should probably return
VA_STATUS_ERROR_INVALID_IM
of free frame store entries
> */
> static int
> diff --git a/src/i965_decoder_utils.h b/src/i965_decoder_utils.h
> index acd21f2..48ef814 100644
> --- a/src/i965_decoder_utils.h
> +++ b/src/i965_decoder_utils.h
> @@ -89,6 +89,13 @@ gen6_send_avc_ref_idx_state(
> const GenFrameStore frame_store[MAX_GEN_REFERENCE_FRAMES]
> );
>
> +void
> +gen6_mfd_avc_phantom_slice(VADriverContextP ctx,
> + VAPictureParameterBufferH264 *pic_param,
> + VASliceParameterBufferH264 *next_slice_param,
> + struct intel_batchbuffer *batch
> +);
> +
> VAStatus
> intel_decoder_sanity_check_input(VADriverContextP ctx,
> VAProfile profile,
The rest looks OK, assuming this is just factored code moved around. :)
Thanks,
--
Gwenole Beauchesne
Intel Corporation SAS / 2 rue de Paris, 92196 Meudon Cedex, France
Registration Number (RCS): Nanterre B 302 456 199
___
Libva mailing list
Libva@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/libva
next_slice_group_param = (VASliceParameterBufferH264
> *)decode_state->slice_params[j + 1]->buffer;
>
> +if (j == 0 && slice_param->first_mb_in_slice)
> +gen8_mfd_avc_phantom_slice_first(ctx, pic_param, slice_param,
> gen7_
: Gwenole Beauchesne
---
src/i965_drv_video.c | 48
src/i965_drv_video.h |1 +
2 files changed, 49 insertions(+)
diff --git a/src/i965_drv_video.c b/src/i965_drv_video.c
index e45cfb3..a628e05 100755
--- a/src/i965_drv_video.c
+++ b/src
Use VA internal objects wherever possible for internal functions,
thus leaving VA object ids to the base interface implementation.
Robustify params validation earlier in there too.
Signed-off-by: Gwenole Beauchesne
---
src/i965_drv_video.c | 283
1 - 100 of 671 matches
Mail list logo