Re: [Libva] [Intel-gfx] [PATCH] drm/i915: Make sample_c messages go faster on Haswell.

2015-01-05 Thread Xiang, Haihao
On Mon, 2015-01-05 at 23:03 -0800, Kenneth Graunke wrote: > On Tuesday, January 06, 2015 02:39:36 PM Xiang, Haihao wrote: > > On Mon, 2015-01-05 at 21:54 -0800, Kenneth Graunke wrote: > > > On Tuesday, January 06, 2015 01:11:53 PM Xiang, Haihao wrote: > > > > > > > > Hi Kenneth, > > > > > > > > H

Re: [Libva] [Intel-gfx] [PATCH] drm/i915: Make sample_c messages go faster on Haswell.

2015-01-05 Thread Xiang, Haihao
On Tue, 2015-01-06 at 14:39 +0800, Xiang, Haihao wrote: > On Mon, 2015-01-05 at 21:54 -0800, Kenneth Graunke wrote: > > On Tuesday, January 06, 2015 01:11:53 PM Xiang, Haihao wrote: > > > > > > Hi Kenneth, > > > > > > How did you test OSD ? I can't reproduce the issue you mentioned, OSD > > > wor

Re: [Libva] [Intel-gfx] [PATCH] drm/i915: Make sample_c messages go faster on Haswell.

2015-01-05 Thread Xiang, Haihao
On Mon, 2015-01-05 at 21:54 -0800, Kenneth Graunke wrote: > On Tuesday, January 06, 2015 01:11:53 PM Xiang, Haihao wrote: > > > > Hi Kenneth, > > > > How did you test OSD ? I can't reproduce the issue you mentioned, OSD > > works well for me when using mplayer-vaapi with the latest > > libva/libv

Re: [Libva] [PATCH] Fix scaling of NV12 surfaces when no output_region is set

2015-01-05 Thread Xiang, Haihao
Applied. Thanks Haihao > The code block filling destination surface target rectangle is too early > and incorrectly fills input surface width & height as final target width > and height. This results in improper scaling of the source surface. > > Move the destination code block further below

Re: [Libva] [PATCH] Fix forward reference requirement for Bob deinterlacing in intel-driver

2015-01-05 Thread Xiang, Haihao
Applied, Thanks Haihao > Remove stray semicolon to fix condition check for forward reference surface > count in vaQueryVideoProcPipelineCaps. > > This fix allows the pipeline query to return 0 forward references for Bob > deinterlacer. > > Signed-off-by: Antti Seppälä > --- > src/i965_drv_vi

Re: [Libva] [PATCH 2/5] Added HEVC PAK context

2015-01-05 Thread Zhao, Yakui
On Mon, 2015-01-05 at 22:20 -0700, Qu,Pengfei wrote: > Signed-off-by: Qu,Pengfei > --- > src/Makefile.am | 2 + > src/gen9_mfc.h | 288 > > 2 files changed, 290 insertions(+) > create mode 100644 src/gen9_mfc.h > > diff --git a/src/Ma

Re: [Libva] [PATCH 1/5] VME pipeline for HEVC

2015-01-05 Thread Zhao, Yakui
On Mon, 2015-01-05 at 22:20 -0700, Qu,Pengfei wrote: > Signed-off-by: Qu,Pengfei The patch includes too much changes of coding style besides setting up VME pipeline for HEVC. (For example: the code change in intel_mfc_avc_prepare, intel_vme_update_mbmv_cost, gen7_vme_walker_fill_vme_batchbuffer)

[Libva] [PATCH 5/5] Added HEVC silce header generated internally

2015-01-05 Thread Qu,Pengfei
Signed-off-by: Qu,Pengfei --- src/i965_encoder_utils.c | 475 ++- src/i965_encoder_utils.h | 43 - 2 files changed, 502 insertions(+), 16 deletions(-) diff --git a/src/i965_encoder_utils.c b/src/i965_encoder_utils.c index e59e7fb..a21e65b 100644 -

[Libva] [PATCH 4/5] Added HEVC PAK pipeline

2015-01-05 Thread Qu,Pengfei
Signed-off-by: Qu,Pengfei --- src/gen9_mfc_hevc.c | 2636 +++ 1 file changed, 2636 insertions(+) create mode 100644 src/gen9_mfc_hevc.c diff --git a/src/gen9_mfc_hevc.c b/src/gen9_mfc_hevc.c new file mode 100644 index 000..76e13b3 --- /dev/nul

[Libva] [PATCH 1/5] VME pipeline for HEVC

2015-01-05 Thread Qu,Pengfei
Signed-off-by: Qu,Pengfei --- src/gen6_mfc_common.c | 688 - src/gen6_vme.h| 19 ++ src/gen9_vme.c| 928 ++ 3 files changed, 1180 insertions(+), 455 deletions(-) diff --git a/src/gen6_mfc_common.

[Libva] [PATCH] Added HEVC encoder. summary

2015-01-05 Thread Qu,Pengfei
Signed-off-by: Qu,Pengfei --- feature: 1. LCU size is 16x16 only 2. Min CU size is 16x16 only 3. Transform size 16x16~8x8 4. I/P/B frame enabled. 5. Constant QP only --- src/Makefile.am |2 + src/gen6_mfc_common.c| 688 src/gen6_vme.h | 19 +

[Libva] [PATCH 3/5] Added to support HEVC related to intel driver API

2015-01-05 Thread Qu,Pengfei
Signed-off-by: Qu,Pengfei --- src/i965_device_info.c | 17 +++--- src/i965_drv_video.c | 153 ++--- src/i965_drv_video.h | 15 - src/i965_encoder.c | 74 +++- 4 files changed, 213 insertions(+), 46 deletions(-) diff -

[Libva] [PATCH 2/5] Added HEVC PAK context

2015-01-05 Thread Qu,Pengfei
Signed-off-by: Qu,Pengfei --- src/Makefile.am | 2 + src/gen9_mfc.h | 288 2 files changed, 290 insertions(+) create mode 100644 src/gen9_mfc.h diff --git a/src/Makefile.am b/src/Makefile.am index 566a7ae..cb88cca 100755 --- a/src/Make

Re: [Libva] [Intel-gfx] [PATCH] drm/i915: Make sample_c messages go faster on Haswell.

2015-01-05 Thread Xiang, Haihao
Hi Kenneth, How did you test OSD ? I can't reproduce the issue you mentioned, OSD works well for me when using mplayer-vaapi with the latest libva/libva-intel-driver master branch. I tried your patch, what surprised me is OSD still works well after applying your patch. It seems your patch didn't

[Libva] [PATCH] Added HEVC encoder. summary

2015-01-05 Thread Qu,Pengfei
Signed-off-by: Qu,Pengfei --- feature: 1. LCU size is 16x16 only 2. Min CU size is 16x16 only 3. Transform size 16x16~8x8 4. I/P/B frame enabled. 5. Constant QP only --- src/Makefile.am |2 + src/gen6_mfc_common.c| 688 src/gen6_vme.h | 19 +

[Libva] [PATCH 1/5] VME pipeline for HEVC

2015-01-05 Thread Qu,Pengfei
Signed-off-by: Qu,Pengfei --- src/gen6_mfc_common.c | 688 - src/gen6_vme.h| 19 ++ src/gen9_vme.c| 928 ++ 3 files changed, 1180 insertions(+), 455 deletions(-) diff --git a/src/gen6_mfc_common.

[Libva] [PATCH 4/5] Added HEVC PAK pipeline

2015-01-05 Thread Qu,Pengfei
Signed-off-by: Qu,Pengfei --- src/gen9_mfc_hevc.c | 2636 +++ 1 file changed, 2636 insertions(+) create mode 100644 src/gen9_mfc_hevc.c diff --git a/src/gen9_mfc_hevc.c b/src/gen9_mfc_hevc.c new file mode 100644 index 000..76e13b3 --- /dev/nul

[Libva] [PATCH 2/5] Added HEVC PAK context

2015-01-05 Thread Qu,Pengfei
Signed-off-by: Qu,Pengfei --- src/Makefile.am | 2 + src/gen9_mfc.h | 288 2 files changed, 290 insertions(+) create mode 100644 src/gen9_mfc.h diff --git a/src/Makefile.am b/src/Makefile.am index 566a7ae..cb88cca 100755 --- a/src/Make

[Libva] [PATCH 5/5] Added HEVC silce header generated internally

2015-01-05 Thread Qu,Pengfei
Signed-off-by: Qu,Pengfei --- src/i965_encoder_utils.c | 475 ++- src/i965_encoder_utils.h | 43 - 2 files changed, 502 insertions(+), 16 deletions(-) diff --git a/src/i965_encoder_utils.c b/src/i965_encoder_utils.c index e59e7fb..a21e65b 100644 -

[Libva] [PATCH 3/5] Added to support HEVC related to intel driver API

2015-01-05 Thread Qu,Pengfei
Signed-off-by: Qu,Pengfei --- src/i965_device_info.c | 17 +++--- src/i965_drv_video.c | 153 ++--- src/i965_drv_video.h | 15 - src/i965_encoder.c | 74 +++- 4 files changed, 213 insertions(+), 46 deletions(-) diff -

[Libva] [PATCH] Fix scaling of NV12 surfaces when no output_region is set

2015-01-05 Thread Antti Seppälä
The code block filling destination surface target rectangle is too early and incorrectly fills input surface width & height as final target width and height. This results in improper scaling of the source surface. Move the destination code block further below the point when actual target surfac

Re: [Libva] [Intel-gfx] [PATCH] drm/i915: Make sample_c messages go faster on Haswell.

2015-01-05 Thread Daniel Vetter
On Wed, Dec 31, 2014 at 04:23:00PM -0800, Kenneth Graunke wrote: > Haswell significantly improved the performance of sampler_c messages, > but the optimization appears to be off by default. Later platforms > remove this bit, and apparently always enable the optimization. > > Improves performance