summaryrefslogtreecommitdiff
path: root/utils
diff options
context:
space:
mode:
authorLaurent Pinchart <laurent.pinchart@ideasonboard.com>2022-07-25 02:37:28 +0300
committerJacopo Mondi <jacopo.mondi@ideasonboard.com>2023-03-08 13:18:10 +0100
commitda67ba6d35de73767bd8872c74f9c2ba91a34424 (patch)
tree7071b57801270eea3cfebae7fe4ff08eba990b31 /utils
parentebe733ed985eaea8f74dcfdd3cea860501b10763 (diff)
[RFC] ipa: rkisp1: dpcc: Generalize YAML parsing
The DefectPixelClusterCorrection::init() function contains a large manually written piece of tuning data parsing code, with duplication of similar but slightly different sections. This is error-prone as copy and paste errors easily creep in and can be hard to spot. As an attempt to address this issue, replace that code with two generic functions tbat operate over a data table which describes the structure of the tuning data and where it fits in the ISP configuration structure (which maps directly to registers). At the same time, restructure the tuning data to group parameter per method, to organize the data in a set-method-parameter structure. The line-threshold and line-mad-factor parameters are grouped in a "lc" (line check) method, and the rnd-threshold and rnd-offset parameters in a "rnd" (rank neighbour difference) method. The other parameters (pg-factor, rg-factor and ro-limits) are directly specified in a "pg", "rg" or "ro" method respectively without a parameter name for simplicity. Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com> --- A few questions (hence the RFC): - Should methods that have a single parameter (e.g. "pg") have an explicit parameter name in YAML, to match methods that have multiple parameters ? This would result in pg: green: factor: 10 red-blue: factor: 10 instead of pg: green: 10 red-blue: 10 - Should the method names be spelled out fully in YAML, instead of being abbreviated to 2 or 3 letters ? - Is this actually worth it ?
Diffstat (limited to 'utils')
0 files changed, 0 insertions, 0 deletions