fix typo in upstream code

bump version to 1.0.19 after sync to recent upstream git repo

add packaging files

Fix the size to malloc()

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

Check the pointer is NULL or not

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

Fix potential buffer overflow for JPEG decoding

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

Remove the dead code in gen6_vme.c

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

Check object for VA buffer

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

Check the object for VA buffer
Signed-off-by: Xiang, Haihao <haihao.xiang@intel.com>

Check the object instance of VAConfig

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

Check the object instance instead of the id for subpicture and image

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

VPP: check the backing store buffer

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

Render: directly use the backing store buffer

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

Decoder: use surface object for the workaround

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

Decoder: directly use surface object for decoding

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

Decoder: Verify picture parameter before set up pipeline for decoding

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

Decoder: check whether the surface for decoding output is valid

In addition, uses the corresponding surface object directly.

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

Encoder: directly use the objects for the reference pictures

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

Encoder: directly use the objects for the reconstructed picture and coded buffer

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

Encoder: check whether the coded buffer and reconstructed surface are valid

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

Encoder: unify the initialization of the context

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

Encoder: directly use the surface object of the input surface

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

Return the status when running the pipeline

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

Silence a bunch of warnings

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

Avoid potential buffer overflow issue

Warning if the slice type is wrong for encoding

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

Release resource if failed to initialize display attributes

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

Fix the initilization path and the termination path in reverse

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

Fix object_heap_init() &  object_heap_destroy()

Don't allocate resources if failed to initialize a heap

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

Issue memory fence message to assure memory ordering on Ivb/Hsw

Otherwise the data inconsistency between different GPU threads
is observed although the GPU threads are spawned by using hardware
scoreboard. Then it causes that avcenc encoding gets the different
results.

Reported-by: Xiang Haihao <haihao.xiang@intel.com>
Signed-off-by: Zhao Yakui <yakui.zhao@intel.com>

Fix H264 YUV400 surface render issue on IVB

All decoded frame are considered as NV12 format in driver,
for YUV400 stream format decoding senerios, we need set the
chroma component of NV12 to a constant value(0x80), otherwise
the converted ARGB from NV12 format is not correct and cause
blue screen when rending.

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

Update PCI IDs for Haswell CRW

Signed-off-by: Xiang, Haihao <haihao.xiang@intel.com>
(cherry picked from commit 82d6940694c7a650642ccb6d68bf01b70dba4dcc)

Update the size of DMV buffer for H.264 decoding on IVB

It is at least width_in_mbs * align(height_in_mbs, 2) * 64.
Use width_in_mbs * (height_in_mbs + 1) * 64 in the driver.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=59050
Signed-off-by: Xiang, Haihao <haihao.xiang@intel.com>
(cherry picked from commit 00f65b78e6de520a4820702207ce098c6b073724)

Don't have flag register f2, use f1 instead

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

gen6: MPEG2 is not supported by gen6 pipeline

Commit 65e10683 added MPEG2 encoding for IVB but also for gen6.
This causes a missmatch between the profile(MPEG) and the pipeline
used(AVC).

Signed-off-by: Adrian Marius Negreanu <adrian.m.negreanu@intel.com>

add replacement for log2f when not available

Signed-off-by: Adrian Marius Negreanu <adrian.m.negreanu@intel.com>

intel-driver, android: update src/Android.mk

Signed-off-by: Adrian Marius Negreanu <adrian.m.negreanu@intel.com>

Update the binding table index on IVB

The index used by the shader is 33 instead of 5.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=59693
Signed-off-by: Xiang, Haihao <haihao.xiang@intel.com>

Correct the mv offset for mpeg2 encoding on HSW

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

Fix the issue of GPU hang when encoding picture with one macroblock width

Signed-off-by: Zhao Yakui <yakui.zhao@intel.com>
Reported-by: Xiang Haihao <haihao.xiang@intel.com>

Enlarge deblocking filter row store for VC-1 decoding

Signed-off-by: Xiang, Haihao <haihao.xiang@intel.com>
(cherry picked from commit e8d3f90fc18ca3726e6b34156ba56ee92c2a2524)

Refine mpeg2 mfc pipeline code

move mpeg2 mfc code from gen6_mfc.c to gen7_mfc.c,
because mpeg2 paking is not supported on gen6.

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

apply new mv offset for mpeg2 encoding on HSW

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

Use the common scoreboard code on Ivy/Haswell to remove the duplicated code

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

Keep the old indent style in file of gen7_vme/gen75_vme

Signed-off-by: Zhao Yakui <yakui.zhao@intel.com>
Asked-by: Xiang Haihao <haihao.xiang@intel.com>

Fix one error of calculating VME parameter for inter-frame on Haswell

At the same time it fixes the typo in comments.

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

Update the binary code for MPEG2 VME on haswell

The same source code is used by AVC and MPEG2 VME. But the binary shader
for MPEG2 is not updated after adding the MVP for AVC.

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

Remove the intermedia file of mpeg2_inter_frame_haswell.gen75.asm

Add the bidirectional MVP to optimize the VME parameter on Ivb

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

Use the scoreboard for AVC encoding on Ivy

This is backported from the Haswell and it is required in order to
add the MVP prediction as it is based on the neighbour macroblocks.

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

Backport the mode/mv cost table on Ivy

Add the implement on Haswell for Ivy. Then it can select the different
predition mode based on the quality requirement.

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

Add the VME shader for Ivy that supports MVP

This is to optimize the VME parameter for Ivy, which is backported
from Haswell.

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

Add the separated encoding files for Ivy

This is to do the prepration of backing port the Haswell VME motion vector
prediction for Ivy.

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

Remove the intermediate file of mpeg2_inter_frame.gen7.asm as it is generated from compiling the shader.

Fix one compiling warning of floor implicit definition

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

Fix the error of reading neighbour macroblock pixels during VME prediction on Snb/Ivy

Otherwise the incorrect result of VME prediction is used.

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

VPP: Support multi filters for vebox pipeline

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

VEBOX: clean supported fourcc format in VPP pipeline

clean up the supported surface format at the begining
and the ending of post processing pipepile, these formats
will be convertd to/from vebox supported formats.

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

VEBOX: update filters info for every frame processing

filter parameters may alter when processing a video sequence,
so keep updating filters information before handing each frame.

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

VEBOX: support multi color balance params in one filter

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

Adjust the reference window based on MVP prediction to optimize VME param on Haswell

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

Add the bidirectional MV prediction for B-frame on Haswell

Now the B-frame only supports the forward prediction. In order to
get the better prediction result, the bidirection prediction is added
for B-frame. At the same time the MV prediction is also added.

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

Remove the preprocessor redefinitions for H264 and MPEG2 on Haswell

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

Packing the Intra macroblock in P/B slice for MPEG2

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

Enable MPEG2 encoding flag for gen7 codec info

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

Avoid intra VME for MPEG2

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

Add MPEG2 MFC pipeline for IVB

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

Add MPEG2 VME pipeline for IVB

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

Add MPEG2 VME shader files for IVB

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

Use the wavefront-scan instead of raster-scan during VME prediction on Haswell

When adding the MV prediction for AVC encoding, the GPU thread of current
macroblock needs to wait for the completion of neighbour macroblocks(left/top/up-right).
And the raster-scan will degrade the performance of AVC encoding. So the
wavefront-scan is used to spanw GPU thread, which can avoid the degrading of
the encoding performance.

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

Add the MV prediction to optimize VME parameter on Haswell

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

Discard the intermediate result during VME prediction on haswell

There is no functional change. It is only to avoid writing the
intermediate result of VME prediction.

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

MEDIA_OBJECT uses hardware scoreboard during VME prediction on Haswell

To get the precise VME prediction the current macroblock will depend on its
neighbour mb(left, top, up-right). So the hardware scoreboard is used
when submitting the MEDIA_OBJECT command. This is to do the preparation
of adding MV prediction in VME prediction.

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

Set Fix_Prev_Mb_skipped flag in AVC_BSD_OBJECT command instead of VC1_BSD_OBJECT command

A stupid error by me in the commit f750d1b

Signed-off-by: Xiang, Haihao <haihao.xiang@intel.com>
(cherry picked from commit 88b21b9aba9be13e08109fe5d213973447f38558)

Use the internal CondOver flag

Signed-off-by: Xiang, Haihao <haihao.xiang@intel.com>
(cherry picked from commit 2cb319923b6ae5c1b2c643534d7a97911b9bd54e)

Fix ttmbf/ttfrm when vstransform is 0 on HSW

This is the same fix of c7d23b1

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

No overlap smoothing for B frame in Main and Simple profiles

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=58448
Signed-off-by: Xiang, Haihao <haihao.xiang@intel.com>
(cherry picked from commit 660cb88ec00a896eefc9580e691d0a07fb2870ec)

Always set Fix_Prev_Mb_skipped in AVC_BSD_OBJECT command

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=57720
Signed-off-by: Xiang, Haihao <haihao.xiang@intel.com>
(cherry picked from commit f750d1b69e1aae65463eab224013b506481d8459)

render: fix rotation vertices for Ironlake.

Ironlake requires the vertex buffer to be ordered in a particular way.
More specifically, the correct order is bottom-right, bottom-left and
top-left vertices in "output" view, i.e. transformed.

Signed-off-by: Gwenole Beauchesne <gwenole.beauchesne@intel.com>
(cherry picked from commit dc4d4005d0f3de59b7218506cf5b20bcef61323f)

subpicture: don't overallocate palette on 64-bit systems.

Allocate the exact amount of memory for VA image palettes on 64-bit
systems. No more.

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

subpicture: fix creation of IA88/AI88 subpicture images.

IA88 format is 16bpp, with one byte for alpha and one byte for the color
index. Besides, a palette with 256 entries is also needed.

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

subpicture: expose "global-alpha" is supported.

Make sure vaQuerySubpictureFormats() reports that "global-alpha" is
supported, along with "screen-coords".

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

Render: Update the maximum number of WM threads

The number is stolen from Mesa.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=57323
Signe-off-by: Xiang, Haihao <haihao.xiang@intel.com>
Tested-by: zaverel <zaverel@free.fr>
(cherry picked from commit a140c632046e50a41bf75da097834fd9954b9561)

Add IS_SNB_GT1/IS_SNB_GT2/IS_IVB_GT1/IS_IVB_GT2 and remove IS_HSW_ULT and IS_HSW_GT2_PLUS

Signed-off-by: Xiang, Haihao <haihao.xiang@intel.com>
(cherry picked from commit bb946b7d2c97c94ac1888195ab1d5b9c59750d23)

Adopt the same fix used in the commit 29d73dc to HSW

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

h264: fix first macroblock bit offset calculation (ILK, SNB, IVB).

Fix and simplify the scan for emulation_prevention_bytes, thus avoiding
a read beyond the end of the slice data buffer. Besides, if slice_header()
bytes are needed, use dri_bo_get_subdata() instead.

HW specific changes:
- SNB: make the HW skip the emulation prevention bytes itself.
- IVB: fix MFD_AVC_BSD_OBJECT to report the actual slice data buffer size.

Signed-off-by: Gwenole Beauchesne <gwenole.beauchesne@intel.com>
(cherry picked from commit 9b7863bf49dcf8bf1de9b45ce4e986dfd1cca418)

Fix the compile error in the previous commit

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

VPP: fix the AVS steping and frame origin issue

On HSW/IVB, when the resolution of source rectangle is
not the same as src surface, or the source rectangle does
not locate at (0,0), the scaled picture is not correct,
adjust the steping size and frame origin according to the
src rectangle's location and resolution.

Signed-off-by: Li Xiaowei <xiaowei.a.li@intel.com>
Signed-off-by: Xiang, Haihao <haihao.xiang@intel.com>

workaround to set VC1 overlap filter flag

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

Fill the bitplane for VC1 skip picture decoding

This is a workaround for VC1 skip picture, the corresponding
bit value in bitplane should be 1 for skip picture, but sometimes
application pass down wrong value.

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

MPEG-2 encoding on Haswell

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

MPEG-2 encoding path

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

Fix Motion Vector

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

MPEG-2 encoding: Use pre deblocking output for reconstructed picture.

This avoids OLDB enabling for MPEG-2

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

Fixed an assertion failure

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

Follow the vme output to configure PAK command

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

Shader for MPEG-2 encoding

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

Fix the number of MVs for MPEG-2 encoding

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

PAK command for inter macroblock

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

vme for MPEG-2 on Haswell

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

New field profile in struct intel_encoder_context

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