Added new surface attributes to enable VA surface creation from external buffers.

avcenc: Fix constraint flag

This fixes the issue that the output file can't be play correctly by mplayer.

Signed-off-by: Xiang, Haihao <haihao.xiang@intel.com>

Add RGB RT format definition for video process

RGB format surface is required for color
space conversion from YUV to RGB.

signed-off-by: Li,Xiaowei <xiaowei.a.li@intel.com>

Added baseline to the list of profiles for the documentation.

Add maximum type number define, and one addition VPP flag

Add maximum type number define for VAEntryPoint,
VAConfigAttribTypeMax, VABufferTypeMax, which can be used as
the current total number of entrypoint/attribute/buffer type.

Add END of stream flag for VPP pipeline which indicates it's
the last frames in the pipeline.

Signed-off-by: Binglin Chen <binglin.chen@intel.com>

va_trace: allow surface trace to dump part of surface content

To debug/isolate video record quality issue (blocky/corruption, etc),
we always need to dump the YUV image from camera. LibVA has such
functionality, but due to storage bandwidth issue (~10M Byte/second),
it is hard to write 1080P 30fps YUV data to disk (~88M Byte/second).
Now refine the functionality to only dump part of the image, and
won't need to modify the code in the default build.

The method is:
edit /etc/libva.conf to include bellow (all export them):
LIBVA_TRACE_SURFACE=/data/enc-XXX.yuv
LIBVA_TRACE_GEOMETRY=320x240+20+20

It will only save 320x240 from (20,20) of the source surface to file
/data/enc-XXX.yuv.

Signed-off-by: Austin Yuan <shengquan.yuan@intel.com>

va_trace_surface: dumpe surface content after vaEndPicture call

vaEndPicture will kick HW decode/encode, and at this time use CPU
to dump surface content can keep GPU/CPU in parallel.

Also for encode trace, don't need to wait current frame completions
but for decode, we must wait decode till it is completed

Signed-off-by: Austin Yuan <shengquan.yuan@intel.com>

va_trace.c: fix a wrong conditional check

Should check trace_fp_codedbuf instead of trace_fp_log because
we write to trace_fp_codedbuf

Signed-off-by: Austin Yuan <shengquan.yuan@intel.com>

va.c: need to check the symbol before calling it

This is to fix Klockwork critical issue

Signed-off-by: Austin Yuan <shengquan.yuan@intel.com>

android: use Android logging utilities.

Log functions available are now Android compatible and log statements
will show on user space android logging system.

Defined a LOG_TAG and make it available for all versions of Android.

Signed-off-by: Daniel Charles <daniel.charles@intel.com>
Signed-off-by: Gwenole Beauchesne <gwenole.beauchesne@intel.com>

wayland: don't export internal functions.

Signed-off-by: Gwenole Beauchesne <gwenole.beauchesne@intel.com>

wayland: allow de-interlaced picture buffers.

Signed-off-by: Gwenole Beauchesne <gwenole.beauchesne@intel.com>

wayland: ship with wayland-drm protocol.

Signed-off-by: Zhao Halley <halley.zhao@intel.com>
Signed-off-by: Gwenole Beauchesne <gwenole.beauchesne@intel.com>

tests: add PutSurface test for Wayland.

Signed-off-by: Gwenole Beauchesne <gwenole.beauchesne@intel.com>

tests: add support for Wayland.

Signed-off-by: Gwenole Beauchesne <gwenole.beauchesne@intel.com>

API: add support for Wayland.

Signed-off-by: Gwenole Beauchesne <gwenole.beauchesne@intel.com>

configure: fix EGL check.

This fixes the following error:
checking EGL/egl.h presence... no
configure: WARNING: EGL/egl.h: accepted by the compiler, rejected by the preprocessor!
configure: WARNING: EGL/egl.h: proceeding with the compiler's result

Signed-off-by: Gwenole Beauchesne <gwenole.beauchesne@intel.com>

drm: don't export internal functions.

Signed-off-by: Gwenole Beauchesne <gwenole.beauchesne@intel.com>

configure: check for visibility attribute.

Add a check for GCC "visibility" attribute. Also define DLL_HIDDEN helper
so that to declare functions as internal to a particular shared library.

Signed-off-by: Gwenole Beauchesne <gwenole.beauchesne@intel.com>

test/avcenc: link with -lpthread.

Otherwise it will complain during compiling:
usr/bin/ld: avcenc.o: undefined reference to symbol 'pthread_create@@GLIBC_2.2.5'
/usr/bin/ld: note: 'pthread_create@@GLIBC_2.2.5' is defined in DSO

Signed-off-by: Zhao Yakui <yakui.zhao@intel.com>

tests: fix JPEG decoding test to match the latest API.

Fix bug in parsing Huffman DC table and also sync the code to match
the latest API change. In particular, pad the Huffman table with
zero to suit 4-byte boundaries.

Signed-off-by: Gwenole Beauchesne <gwenole.beauchesne@intel.com>

jpeg: pad Huffman table to 4-byte boundaries (Haihao).

Some VA driver implementations require alignment to 4-byte boundaries
so that to allow for direct memory transfers.

Signed-off-by: Gwenole Beauchesne <gwenole.beauchesne@intel.com>

drm: move driver name detection to utilities.

Move VA driver name characterisation into a dedicated utilities file.
This is also meant to be useful to VA/Android that can re-use the same
file as is since it is self-contained.

Signed-off-by: Gwenole Beauchesne <gwenole.beauchesne@intel.com>

drm: fix build without X11 headers.

Signed-off-by: Dmitry Ermilov <dmitry.ermilov@intel.com>
Signed-off-by: Gwenole Beauchesne <gwenole.beauchesne@intel.com>

drm: allow authentication through X11 server.

Make it possible to authenticate a DRM connection through an X server
that may already be running.

Signed-off-by: Gwenole Beauchesne <gwenole.beauchesne@intel.com>

drm: ensure DRM connection is authenticated.

Make sure the DRM connection is authenticated prior to returning a
valid display from vaGetDisplayDRM(). If an X or Wayland server is
running then authentication will also fail.

Signed-off-by: Gwenole Beauchesne <gwenole.beauchesne@intel.com>

Allow build without VA/X11 API.

In this case, VA/GLX API is disabled and another backend shall be
built instead.

Signed-off-by: Dmitry Ermilov <dmitry.ermilov@intel.com>
Signed-off-by: Gwenole Beauchesne <gwenole.beauchesne@intel.com>

tests: add support for raw DRM.

Signed-off-by: Dmitry Ermilov <dmitry.ermilov@intel.com>
Signed-off-by: Gwenole Beauchesne <gwenole.beauchesne@intel.com>

API: add support for raw DRM.

This API makes it possible to use DRM-based VA drivers without an X server.
Since this is a renderless API, vaPutSurface() is not available.

Signed-off-by: Dmitry Ermilov <dmitry.ermilov@intel.com>
Signed-off-by: Gwenole Beauchesne <gwenole.beauchesne@intel.com>

test/basic: fix make dist.

Signed-off-by: Gwenole Beauchesne <gwenole.beauchesne@intel.com>

API: backend: add VA display types.

Make it possible to easily find how the VA display was created. This
helps implement multiple backends in VA drivers. e.g. X11, Wayland
while keeping a single generated binary.

Signed-off-by: Gwenole Beauchesne <gwenole.beauchesne@intel.com>

API: backend: rename VADriverContext.dri_state to drm_state.

VADriverContext.drm_state holds data structures derived from struct
drm_state, thus also including struct dri_state for VA/X11 drivers.

Signed-off-by: Gwenole Beauchesne <gwenole.beauchesne@intel.com>

API: introduce DRM common utilities and data structures.

Add <va/va_drmcommon.h> header for DRM-based drivers to use common
base utilities and data structures. So far, only the DRM state is
shared so that to maintain binary compatibility.

Signed-off-by: Gwenole Beauchesne <gwenole.beauchesne@intel.com>

tests: update JPEG decoding test case.

Signed-off-by: Xiang, Haihao <haihao.xiang@intel.com>

jpeg: drop sample_precision since P=8 for baseline profile.

Signed-off-by: Gwenole Beauchesne <gwenole.beauchesne@intel.com>

jpeg: drop JPEG-2000 fields from structures for Baseline profile.

Signed-off-by: Gwenole Beauchesne <gwenole.beauchesne@intel.com>

jpeg: include <va/va_dec_jpeg.h> into core VA header.

Signed-off-by: Gwenole Beauchesne <gwenole.beauchesne@intel.com>

jpeg: document decoding API is only suitable to baseline profile.

Signed-off-by: Gwenole Beauchesne <gwenole.beauchesne@intel.com>

jpeg: add "Baseline" suffix the existing data structures.

These structures are for Baseline profile only.

Signed-off-by: Xiang, Haihao <haihao.xiang@intel.com>

tests: avcenc: allow selection of display.

Signed-off-by: Gwenole Beauchesne <gwenole.beauchesne@intel.com>

rename NativeDisplay to VANativeDisplay

Signed-off-by: Xiang, Haihao <haihao.xiang@intel.com>
[cherry-pick'ed from commit a65acb298894fa90e3cb7692795027143d6b1dbe]

tests: allow selection of display.

Make it possible to select display for test applications that need
to render the decoded surfaces. Usage: --display <name> args.

Signed-off-by: Gwenole Beauchesne <gwenole.beauchesne@intel.com>

tests: avcenc: use common VA display creation routines.

Signed-off-by: Dmitry Ermilov <dmitry.ermilov@intel.com>

tests: simplify VA display construction.

Signed-off-by: Gwenole Beauchesne <gwenole.beauchesne@intel.com>

test: fix avcenc build on Android.

Signed-off-by: Gwenole Beauchesne <gwenole.beauchesne@intel.com>

test/basic: changed build target to noinst so they are built by default

Signed-off-by: Haitao Huang <haitao.huang@intel.com>
(cherry picked from commit cc86031259117b3cd5607865858c9054cd44db0e)

test/encode: add avcenc target for Android.mk.

Signed-off-by: Haitao Huang <haitao.huang@intel.com>

test/encode/avcenc: rename swap32() to va_swap32().

Signed-off-by: Haitao Huang <haitao.huang@intel.com>

test: remove trailing whitespaces in Android.mk files.

Signed-off-by: Haitao Huang <haitao.huang@intel.com>

test/basic: fix build of basic tests.

Signed-off-by: Haitao Huang <haitao.huang@intel.com>

test/basic: update to new API.

Signed-off-by: Haitao Huang <haitao.huang@intel.com>

.gitignore: update.

- Add new tests and generated files.
- Sort the lines by alphabetical order.

Signed-off-by: Gwenole Beauchesne <gwenole.beauchesne@intel.com>
[cherry-pick'ed from master 09533757aca8b8764a925dfa92204a8bb3c396f4]

build: fix make dist.

Signed-off-by: Gwenole Beauchesne <gwenole.beauchesne@intel.com>

Fix build with alternate X11, GLX and EGL prefixes.

Signed-off-by: Gwenole Beauchesne <gwenole.beauchesne@intel.com>

Fix regression caused by 027a1298c22aeecc2fee2d270cd86790530892f6

VA/EGL depends on EGL/egl.h, so checking for eglGetDisplay in -lEGL
doesn't mean it could be compiled or not.

Signed-off-by: Xiang, Haihao <haihao.xiang@intel.com>

configure: fix GLX and EGL tests.

Migrate checks for GLX and EGL to pkg-config checks or use default
include and libraries for testing.

Signed-off-by: Gwenole Beauchesne <gwenole.beauchesne@intel.com>

fix the check for EGL

It is possible that EGL/egl.h doesn't exist however EGL library is
available. We don't want to build EGL related stuff in this case

Signed-off-by: Xiang, Haihao <haihao.xiang@intel.com>
[Gwenole: removed the non-existing check for the testcase]

API for MPEG-2 encoding

Signed-off-by: Xiang, Haihao <haihao.xiang@intel.com>

jpeg: mark default Huffman table as valid.

When no Huffman table is defined in the JPEG segments, then the default
Huffman tables are used, but values were not marked as valid. Fixed this
to notify HW decoder that it needs to load the default tables.

Note: this is not specifically needed to be handled at the application
level. The driver also ought to generate default tables, should the HW
decoder require so and no table was provided from the application.

Signed-off-by: Gwenole Beauchesne <gwenole.beauchesne@intel.com>

jpeg: fix display of dimensions.

Use human-readable form for image dimensions. i.e. use decimal representation
instead of hexadecimal one.

Signed-off-by: Gwenole Beauchesne <gwenole.beauchesne@intel.com>

Add VP8 decoding API

Added new profile, buffer types and data structures for VP8 decoding

Signed-off-by: Jonathan Bian <jonathan.bian@intel.com>

Add Deblocking and FRC filter to VPP API

Added Deblocking and FRC filter types and data structures

Signed-off-by: Jonathan Bian <jonathan.bian@intel.com>

license header: add the missing license header to files

Signed-off-by: Austin Yuan <shengquan.yuan@intel.com>

jpeg: fix IQ matrix and Huffman table buffer name.

Align quantiser matrix and Huffman table buffer names for JPEG to existing
practice for other codecs. i.e. drop the "Parameter" suffix in the name.

Signed-off-by: Gwenole Beauchesne <gwenole.beauchesne@intel.com>

Update JPEG decode demo to match current API.

Signed-off-by: Gwenole Beauchesne <gwenole.beauchesne@intel.com>

A simple JPEG example to use VAAPI

It is based on Luc Saillard <luc@saillard.org>'s JPEG program

Usage: loadjpeg <input_filename.jpeg>

Signed-off-by: Zhao Halley <halley.zhao@intel.com>
Singed-off-by: Wind Yuan <feng.yuan@intel.com>
Signed-off-by: Xiang, Haihao <haihao.xiang@intel.com>

jpeg: cosmetics (vertical align).

Signed-off-by: Gwenole Beauchesne <gwenole.beauchesne@intel.com>

Android.mk: install VPP header files

Signed-off-by: Austin Yuan <shengquan.yuan@intel.com>

VAEncXXXH264Baseline: start to drop VAEncXXXXBaseline

Signed-off-by: Austin Yuan <shengquan.yuan@intel.com>

tpi: clean the old interface, and use new one

Signed-off-by: Austin Yuan <shengquan.yuan@intel.com>

android backend: add the new device ID into it

Signed-off-by: Austin Yuan <shengquan.yuan@intel.com>

Add new fourcc codes

Added fourcc codes for YV24 and YV32 (4:4:4 planar formats)

Signed-off-by: Jonathan Bian <jonathan.bian@intel.com>

Add initial JPEG decoding API.

This is work-in-progress API subject to changes. In particular, this
API is only suitable to Baseline profile. Huffman and quantization
tables also need to be refined.

Signed-off-by: Gwenole Beauchesne <gwenole.beauchesne@intel.com>

vatrace: trace H.264 encode parameters

Signed-off-by: Ermilov, Dmitry <dmitry.ermilov@intel.com>

API: h264 decode: fix the definition of slice_data_bit_offset.

Align with the existing practice in various media stacks available
in Linux today. i.e. codec layer submits slice data that represents
the original bitstream, thus including any emulation prevent bytes.
However, the slice_data_bit_offset represents an offset relative
to the buffer that contains the NAL unit byte and the rest of the
slice with any emulation prevention bytes removed.

Signed-off-by: Gwenole Beauchesne <gwenole.beauchesne@intel.com>

Revert "API: h264 decode: introduce num_emulation_prevention_bytes."

This change turns out to be useless based on a survey of all existing
drivers and codec layers implemented so far. It's better to stick to
the behaviour whereby slice_data_bit_offset excludes emulation prevention
bytes count. Let's schedule this change is this is really necessary.

This reverts commit a870f078984ce829f797e5610b30d39a3921af4f.

Signed-off-by: Gwenole Beauchesne <gwenole.beauchesne@intel.com>

API: h264 decode: introduce num_emulation_prevention_bytes.

num_emulation_prevention_bytes represent the number of emulation prevention
bytes available in the slice_header(). This helps VA drivers determine how
many emulation prevention bytes there are so that to program the hardware
appropriately, without having to parse the slice_header() again.

Rationale: it is not always possible to parse the slice_header() again.
For instance, if "protected slices" are used, the slice data is uploaded
into a VAProtectedSliceDataBuffer and it is no longer possible to read
bytes from it afterwards. e.g. it could be in encrypted memory region
or protected memory region only accessible from the video decoder.

In newly written code, this member shall be filled in appropriately along
with emulation_prevention_bytes_flag set to 1.

Signed-off-by: Gwenole Beauchesne <gwenole.beauchesne@intel.com>

API: h264 decode: clarify slice_data_bit_offset field.

slice_data_bit_offset represents a bit offset from the NAL Header Unit
byte to the begining of the slice_data(), thus including any emulation
prevention bytes. Baiscally, this represents the number of bits parsed
for slice_header().

Signed-off-by: Gwenole Beauchesne <gwenole.beauchesne@intel.com>

API: vpp: remove extraneous flags member.

VAProcPipelineCaps contained an extra flags member which existed before
its split into pipeline_flags and filter_flags.

Signed-off-by: Gwenole Beauchesne <gwenole.beauchesne@intel.com>

API: bump version to 0.34.0 (libva 1.2.0).

Signed-off-by: Gwenole Beauchesne <gwenole.beauchesne@intel.com>

test/transcode/mpeg2transcode: set right picture width/height in macroblocks

Signed-off-by: Xiang, Haihao <haihao.xiang@intel.com>
(cherry picked from commit f8be4a5477e58c4a27170f0c87eeedac8de60aef)

Added high profile support in SPS for avcenc.

Added aspect ratio related fields to VAEncSequenceParameterBufferH264

Signed-off-by: Jonathan Bian <jonathan.bian@intel.com>

VACodedBufferSegment: correct the description

Signed-off-by: Austin Yuan <shengquan.yuan@intel.com>

va_TraceCreateSurface->va_TraceCreateSurfaces: correct name change

Signed-off-by: Austin Yuan <shengquan.yuan@intel.com>

Android.mk: change Android Makefile accordingly

Signed-off-by: Austin Yuan <shengquan.yuan@intel.com>

test: use new vaCreateSurfaces interface

Signed-off-by: Austin Yuan <shengquan.yuan@intel.com>

VAProcPipelineParameterBufferType: start from 41 in case there are more encode buffer types to be added in future

Signed-off-by: Austin Yuan <shengquan.yuan@intel.com>

VA_STATUS_ERROR_HW_BUSY: error code to indicate HW is busy

Example: the harward only supports one way encode, but we
start two encode apps. This error code can return to app

Signed-off-by: Austin Yuan <shengquan.yuan@intel.com>

VAEncMiscParameterRateControl: add back some fields and refine the comments

Signed-off-by: Austin Yuan <shengquan.yuan@intel.com>

VADisplayAttribDirectSurface: drop it

vaPutSurface may use overlay or GPU alternately. App should
use vaQuerySurfaceStatus to check if the surface can be used
for decoding. If VASurfaceDisplaying is returned, it means
the surface is being rendered by overlay and can't be over-writeen.

Signed-off-by: Austin Yuan <shengquan.yuan@intel.com>

VADisplayAttribBLEMode: drop it since no implementation for it

Signed-off-by: Austin Yuan <shengquan.yuan@intel.com>

VACodedBufferSegment: modify the semantics/description

VACodedBufferSegment may contain more than one NALs, and flag
VA_CODED_BUF_STATUS_SINGLE_NALU will be used to singal the
segment only contain one NAL

Signed-off-by: Austin Yuan <shengquan.yuan@intel.com>

test/transcode: Fix DSO issue

Signed-off-by: Xiang, Haihao <haihao.xiang@intel.com>

test/transcode/mpeg2transcode: use the method of constant qp

Signed-off-by: Xiang, Haihao <haihao.xiang@intel.com>

Fix sample code for VPP

Signed-off-by: Xiang, Haihao <haihao.xiang@intel.com>

install va_enc.h and va_enc_h264.h

Signed-off-by: Xiang, Haihao <haihao.xiang@intel.com>

test: use the new H264 encoding and VPP interfaces

Signed-off-by: Xiang, Haihao <haihao.xiang@intel.com>

Add test/transcode/Makefile.am

Signed-off-by: Hai Lan <hai.lan@intel.com>

test: use new interface for all test cases

Signed-off-by: Xiang, Haihao <haihao.xiang@intel.com>