]>
granicus.if.org Git - libvpx/log
summary |
shortlog | log |
commit |
commitdiff |
tree
first ⋅ prev ⋅ next
Yi Luo [Wed, 26 Apr 2017 21:19:49 +0000 (14:19 -0700)]
Update --enable-aom-highbitdepth to --enable-highbitdepth
Change-Id: I342d5644a46840cce6d3e493fe6af187ce943d01
Yi Luo [Tue, 18 Apr 2017 19:25:32 +0000 (12:25 -0700)]
Some changes on list_exp_speed.sh/runconfig.sh
- Use taskset.
- Decoder uses --noblit.
- Output bitstream to /dev/shm in encoder.
- Default project location is ~/Dev/aomedia.
- runconfig.sh does not disable daala-ec explicitly.
Change-Id: I45a564fd291b88c0cb1c691d4e7ce09ba6444c41
Yi Luo [Mon, 3 Apr 2017 22:22:05 +0000 (15:22 -0700)]
Update experiment list
- Add park_joy 360p sequence for fast testing.
Change-Id: I89378fb7959c1ca08bb8b09f4961744cc227e582
Yi Luo [Sat, 4 Mar 2017 22:54:52 +0000 (14:54 -0800)]
Adapt to the directory changes
Change-Id: I35c37ccafb7aeac2b19b5a4f860366596ae1a9dc
Yi Luo [Fri, 3 Mar 2017 22:27:25 +0000 (14:27 -0800)]
Encoding only, stop profiling the decoder
Change-Id: Ib8748470924a48130c18947ac2530c286dce5c33
Yi Luo [Thu, 23 Feb 2017 01:28:26 +0000 (17:28 -0800)]
Add profile_exp_speed_hbd.sh
Turn on/off debug in runconfig_hbd.sh
Change-Id: I22336febaf3adf57268a6d93c4a607ca1d38c736
Yi Luo [Wed, 22 Feb 2017 19:50:56 +0000 (11:50 -0800)]
All script commands are from scripts directory
Specify bit stream file for each tool
Change-Id: I7eb9e699bf3819c99f259460b3701ecd2e585433
Yi Luo [Wed, 22 Feb 2017 18:22:37 +0000 (10:22 -0800)]
Add configure scripts
Change-Id: If732b427e04e32f873d547606c0bcdd39ae2f1f3
Yi Luo [Tue, 21 Feb 2017 18:06:48 +0000 (10:06 -0800)]
Speed test script for standard bit depth
Change-Id: I9066362312994fc5250d713cea46e76aa145b873
Yi Luo [Tue, 21 Feb 2017 17:42:32 +0000 (09:42 -0800)]
Speed test script for high bit depth
Change-Id: Ie9ca0cd34e7072ffd008d8c11c0aacf84d7feadb
Yi Luo [Tue, 7 Jun 2016 22:14:57 +0000 (15:14 -0700)]
Use v4p and v8p in subpixel wrapper functions
Yi Luo [Tue, 7 Jun 2016 21:24:09 +0000 (14:24 -0700)]
Add subpixel 8 method (Horiz8)
Yi Luo [Tue, 7 Jun 2016 19:28:09 +0000 (12:28 -0700)]
Change filter_horiz_w4_ssse3() name
- w4 changes to v4p which is vertical 4-pixel parallel.
Then we can have vertial 8-pixel parallet (v8p).
Yi Luo [Tue, 7 Jun 2016 16:56:17 +0000 (09:56 -0700)]
Do a code clean-up
Yi Luo [Tue, 7 Jun 2016 01:25:25 +0000 (18:25 -0700)]
Move prototype wrapper function
- Found after setting height on command line change, HorizP turns
to be slower. The change include moving this wrapper function.
This change restores it. This is a mystery now.
Yi Luo [Tue, 7 Jun 2016 00:38:05 +0000 (17:38 -0700)]
Make horiz4 interface consistent with horizP
Yi Luo [Tue, 7 Jun 2016 00:23:33 +0000 (17:23 -0700)]
Add tap-10 filter for subpixel method
Yi Luo [Mon, 6 Jun 2016 22:35:33 +0000 (15:35 -0700)]
Set height at command line input
- Remove some unused code.
Yi Luo [Mon, 6 Jun 2016 22:18:11 +0000 (15:18 -0700)]
Support block size 128 (EXT_PARTITION)
Yi Luo [Mon, 6 Jun 2016 22:04:35 +0000 (15:04 -0700)]
Fix a bug on subpixel method on src/dst updating
Yi Luo [Mon, 6 Jun 2016 21:16:00 +0000 (14:16 -0700)]
Avoid shuffle in subpixel method
Yi Luo [Mon, 6 Jun 2016 19:36:46 +0000 (12:36 -0700)]
Subpix method supports 12-tap file
Yi Luo [Mon, 6 Jun 2016 18:12:44 +0000 (11:12 -0700)]
Start the subpixel method for 12-tap filter
Yi Luo [Mon, 6 Jun 2016 17:08:11 +0000 (10:08 -0700)]
Use stride instead of width
- This simulates the real situation in encoder.
Yi Luo [Fri, 3 Jun 2016 23:26:45 +0000 (16:26 -0700)]
Simulate large i/o buffer with width and height
- Change sr/sc to sumPairRow/sumPairCol.
Yi Luo [Fri, 3 Jun 2016 19:36:49 +0000 (12:36 -0700)]
Ignore the 6th and 7th column vector in transpose
Yi Luo [Fri, 3 Jun 2016 19:22:36 +0000 (12:22 -0700)]
Add 10 tap filtering
Yi Luo [Fri, 3 Jun 2016 19:01:35 +0000 (12:01 -0700)]
Add test wrapper functions
- Clean-up the code.
Yi Luo [Fri, 3 Jun 2016 18:41:32 +0000 (11:41 -0700)]
Fix the random number input
Yi Luo [Fri, 3 Jun 2016 18:15:50 +0000 (11:15 -0700)]
Set random seed number at command line
Yi Luo [Fri, 3 Jun 2016 18:08:55 +0000 (11:08 -0700)]
The 6th and 7th column sum vector are zero in theory
- Finally we're correct.
- 10-tap filter cofficients should be the same structure as 12-tap.
Yi Luo [Fri, 3 Jun 2016 17:44:43 +0000 (10:44 -0700)]
The 7th column is zero
Yi Luo [Fri, 3 Jun 2016 17:30:22 +0000 (10:30 -0700)]
Align the horizontal sum vector
- For tap 12 filter, make sure the 2nd and 3rd pair sum are the max.
- We add [2] and [3] at last to tackle the saturation issue.
Yi Luo [Fri, 3 Jun 2016 17:11:04 +0000 (10:11 -0700)]
Fix horiz_filter.c indents
Yi Luo [Fri, 3 Jun 2016 16:50:04 +0000 (09:50 -0700)]
filtering.c uses random number generator for input
Yi Luo [Fri, 3 Jun 2016 16:38:03 +0000 (09:38 -0700)]
Print subpel_q4 and step_q4 behaviour
Yi Luo [Fri, 3 Jun 2016 16:19:57 +0000 (09:19 -0700)]
Random number seed is defined as a macro constant
Yi Luo [Fri, 3 Jun 2016 01:39:41 +0000 (18:39 -0700)]
Add horiz_filter.c
- Saturation correct solution 3
Yi Luo [Wed, 1 Jun 2016 19:38:07 +0000 (12:38 -0700)]
Build as ssse3 instead of sse4.1
Yi Luo [Wed, 1 Jun 2016 19:11:54 +0000 (12:11 -0700)]
Move control mask assignment in controling function
Yi Luo [Wed, 1 Jun 2016 18:28:44 +0000 (11:28 -0700)]
Second solution also bit-exact
Yi Luo [Tue, 31 May 2016 16:41:47 +0000 (09:41 -0700)]
Add tap number in struct Filter
Yi Luo [Sat, 28 May 2016 01:07:06 +0000 (18:07 -0700)]
Repeat 64 times and clean-up unused functions
Yi Luo [Sat, 28 May 2016 00:54:50 +0000 (17:54 -0700)]
Add a cycle counting for performance
Yi Luo [Sat, 28 May 2016 00:44:40 +0000 (17:44 -0700)]
12-tap filtering works