[Libva] [PATCH][libva-intel-driver] vp9encoder: encoder to handle properly CSC on input surface

2016-08-12 Thread Daniel Charles
VP9 encoder was not checking for the yuv surface fourcc provided as the input. If the format is I420, the driver creates an underlying surface where the input is converted to NV12. The underlying temporary surface was not used properly by the vme_pipeline_vp9 as intel_encoder_check_yuv_surface wi

Re: [Libva] [PATCH] i965_device_info: add CBR and VBR rate control flags for vdenc

2016-08-12 Thread Sean V Kelley
On Fri, 2016-08-12 at 16:20 +0300, Sreerenj wrote: > > > On 12.08.2016 14:20, Sreerenj wrote: > > > > > > On 11.08.2016 19:47, Sean V Kelley wrote: > > > > > > > > > On Thu, 2016-08-11 at 14:55 +0300, sreerenj.balachand...@intel.co > > > m wrote: > > > > From: Sreerenj Balachandran > > > >

Re: [Libva] [PATCH] i965_device_info: add CBR and VBR rate control flags for vdenc

2016-08-12 Thread Sean V Kelley
On Fri, 2016-08-12 at 14:20 +0300, Sreerenj wrote: > > > On 11.08.2016 19:47, Sean V Kelley wrote: > > > > > > On Thu, 2016-08-11 at 14:55 +0300, sreerenj.balachand...@intel.com > > wrote: > > > From: Sreerenj Balachandran > > > > > > Low power encode through VDENC supports both CBR and VBR.

[Libva] Driver stuck on picture encode

2016-08-12 Thread Lionel Landwerlin
Hi, Playing around with the h264 encoder on Haswell, I've noticed that I can get the encoder stuck in gen75_mfc_avc_encode_picture() essentially by setting a bitrate above with can be generated by the encoder for a given picture. Should the driver just give up at some point? Do you think an

Re: [Libva] [PATCH] i965_device_info: add CBR and VBR rate control flags for vdenc

2016-08-12 Thread Sreerenj
On 12.08.2016 14:20, Sreerenj wrote: On 11.08.2016 19:47, Sean V Kelley wrote: Hi Sree, On Thu, 2016-08-11 at 14:55 +0300, sreerenj.balachand...@intel.com wrote: From: Sreerenj Balachandran mailto:sreerenj.balachand...@intel.com>> Low power encode through VDENC supports both CBR and VBR

Re: [Libva] [PATCH] i965_device_info: add CBR and VBR rate control flags for vdenc

2016-08-12 Thread Sreerenj
On 11.08.2016 19:47, Sean V Kelley wrote: Hi Sree, On Thu, 2016-08-11 at 14:55 +0300, sreerenj.balachand...@intel.com wrote: From: Sreerenj Balachandran mailto:sreerenj.balachand...@intel.com>> Low power encode through VDENC supports both CBR and VBR. Make sure vaGetConfigAttribute correctl

[Libva] [PATCH V4: 03/11] encoding: Send VME instruction uses one register as the desc parameter

2016-08-12 Thread Pengfei Qu
From: Zhao Yakui The desc parameter of current VME send instruction is hardcode. And it can't be updated based on the input parameter. Signed-off-by: Zhao Yakui Signed-off-by: pjl Signed-off-by: Pengfei Qu --- src/shaders/vme/inter_bframe_ivb.asm | 17 +++-- src/shaders/vme/inter

[Libva] [PATCH V4: 04/11] Encoding: Pass the qp parameter into VME shader and VME shader select the different cost table based on input Qp on Ivy

2016-08-12 Thread Pengfei Qu
From: Zhao Yakui v1: add assert after bo map In order to suppor that macroblocks have the different QP to do the motion prediction, different cost tables are provided so that the VME engine can select the different mode/motion-vector cost tables based on the input Qp. Signed-off-by: Zhao Yakui

[Libva] [PATCH V4: 08/11] Encoding: ROI support for CQP/CBR on Haswell/Ivy

2016-08-12 Thread Pengfei Qu
From: Zhao Yakui v2: remove unused variable set max ROI number to 3 according low power mode or 8 v1: merge 3 ROI patches together Encoding: Add the support of ROI under CQP on Haswell/Ivybridge Encoding: Add the support of ROI for CBR Currently it will allocate the different qp for th

[Libva] [PATCH V4: 10/11] Adjust the maximum number of motion vectors for B frame on HSW+

2016-08-12 Thread Pengfei Qu
From: Jia Meng Signed-off-by: Jia Meng Signed-off-by: Pengfei Qu --- src/gen75_vme.c | 13 +++-- src/gen8_vme.c | 15 --- src/gen9_vme.c | 17 + 3 files changed, 24 insertions(+), 21 deletions(-) diff --git a/src/gen75_vme.c b/src/gen75_vme.c index a0c355

[Libva] [PATCH V4: 00/11]Add ROI feature on Gen7/Gen7.5/Gen8/Gen9 and scaling list features.

2016-08-12 Thread Pengfei Qu
this add ROI feature on Gen7/Gen7.5/Gen8/Gen9 and scaling list features. v3: do style change accordingly; remove the frame qp delta related patch as interface will be redefined. remove the warning when compiling v2: fix the quality regression issue Jia Meng (2): Adjust the maximum number of m

[Libva] [PATCH V4: 06/11] Encoding: Add one ROI flag and ROI buffer

2016-08-12 Thread Pengfei Qu
From: Zhao Yakui v3: free the qp_per_mb for vme context v2: remove unused variable to avoid warning when compiling. v1: Add one flag to indicate whether ROI is supported in one encode context Allocate one ROI buffer to hold qp per mb dynamically Signed-off-by: Zhao Yakui Signed-off-by: pjl S

[Libva] [PATCH V4: 07/11] encoding:use the qp per every macroblock on Ivy and haswell

2016-08-12 Thread Pengfei Qu
From: Zhao Yakui v1: combine the patch together for Ivy and haswell use-CPU-to-construct-the-MFC-pak-command Signed-off-by: Zhao Yakui Signed-off-by: ceciliapeng Signed-off-by: Pengfei Qu --- src/gen6_mfc.c| 15 --- src/gen6_mfc_common.c | 23 ++- src/

[Libva] [PATCH V4: 05/11] Encoding: Dynamically select one mechanism to construct encoding command buffer for each frame on Haswell and Gen7/Gen6

2016-08-12 Thread Pengfei Qu
From: Zhao Yakui v2:remove the warning when compiling v1: combine the Haswell and Gen7/6 patch together Currently it uses the fixed policy to construct encoding command buffer. (Use CPU or GPU). And it is statically compiled. But sometimes it needs to choose the different mechanism on the fly i

[Libva] [PATCH V4: 09/11] ROI:enable on gen8 and gen9

2016-08-12 Thread Pengfei Qu
v2: use ASSERT_RET to check the ROI flag setted by user. instead of assert. v1: ROI enable on gen8 and gen9 Enable GPU to construct GPU command under ROI scenario fix roi attrib config incorrectly Signed-off-by: Zhao Yakui Signed-off-by: Pengfei Qu --- src/gen6_mfc_common.c | 61 ++

[Libva] [PATCH V4: 11/11] scaling matrix of h264 encoder on gen7/gen7.5/gen8/gen9

2016-08-12 Thread Pengfei Qu
From: Jia Meng v1: change the title according to yakui's comments. qm is in raster scan order per va api, and fqm is in column wise raster scan order per hardware requirement. Signed-off-by: Jia Meng Signed-off-by: Pengfei Qu --- src/gen6_mfc.c | 12 -- src/gen6_mfc.h | 2 + src/gen

[Libva] [PATCH V4: 01/11] Encoding: mbmv cost table related changes for ROI

2016-08-12 Thread Pengfei Qu
From: Zhao Yakui v3:remove the warning according to haihao's comments v2: merge three mbmv cost table related patches together. Encoding:Abstract the calculation of mbmv cost for qp as one function. Encoding:Add one function that initialize mbmv cost table for supported Qp range. Encoding:Setup

[Libva] [PATCH V4: 02/11] Encoding: VME shader reads mbmv_cost from cost_table surface instead of constant buffer on Haswell

2016-08-12 Thread Pengfei Qu
From: Zhao Yakui This is to do the prepartion of enhanced features. Signed-off-by: Zhao Yakui Signed-off-by: pjl Signed-off-by: Pengfei Qu --- src/shaders/vme/inter_bframe_haswell.asm | 24 +++- src/shaders/vme/inter_bframe_haswell.g75b | 23 ++- src/