Rietveld Code Review Tool
Help | Bug tracker | Discussion group | Source code | Sign in
(108571)

Issue 1442: pvq: improve rounding bias for int max_theta

Can't Edit
Can't Publish+Mail
Start Review
Created:
1 year, 2 months ago by tmatth
Modified:
1 year, 1 month ago
Reviewers:
jm
Visibility:
Public.

Patch Set 1 #

Total comments: 1
Unified diffs Side-by-side diffs Delta from patch set Stats Patch
M src/pvq.c View 1 chunk +2 lines, -2 lines 1 comment Download

Messages

Total messages: 5
tmatth
1 year, 2 months ago #1
tmatth
subset1: LOW (%) MEDIUM (%) HIGH (%) PSNR -0.116496 -0.114880 -0.095201 PSNRHVS -0.133418 -0.071042 -0.010713 ...
1 year, 2 months ago #2
tmatth
Fixed-point equivalent to the -.25 version of https://review.xiph.org/1435/
1 year, 2 months ago #3
tmatth
Results from daala_2016-08-05 → max-theta-exp-fixed-neg-25-2016-08-04T07-06-46.127Z and daala_2016-08-05s → max-theta-fixed-bias-2016-08-08T12-26-26.483Z
1 year, 2 months ago #4
jm
1 year, 1 month ago #5
http://review.xiph.org/1442/diff/2443/src/pvq.c
File src/pvq.c (right):

http://review.xiph.org/1442/diff/2443/src/pvq.c#newcode738
src/pvq.c:738: - (1 << OD_CGAIN_SHIFT*2 >> 2)) >> OD_CGAIN_SHIFT*2;
This should probably be updated now that beta is fixed-point. Also, I'm not sure
you still need the floor()
Sign in to reply to this message.

RSS Feeds Recent Issues | This issue
This is Rietveld