1
73%
master: 73%

Ran 13 Sep 2020 09:03PM UTC

Files 111

Run time 7s

Badge

Embed ▾

Committed 28 Aug 2020 04:32AM UTC coverage: 72.519%. Remained the same

Job # BUILD_TYPE=Debug COVERALLS=ON

Build Type

cron

travis-ci

Committed by

web-flow

Commit Message

Performance optimization for supportConvex (#488)

* Performance optimization for supportConvex

The original implementation of supportConvex did a linear search across
all vertices. This makes the Convex mesh smarter to make the search for
the support vertex cheaper.

 - It builds an adjacency graph on all vertices upon construction.
 - It introduces the method findExtremeVertex Method which does a walk
   along the edges of the mesh to find the extreme vertex.
   - This is faster for large meshes but slower for small meshes.
   - Empirical data suggests around 32 vertices to be a good cut off.
   - This uses vertex count to determine which algorithm to use.
 - It also introduces initial validation of the mesh -- that validation
   absolutely necessary for the edge walking to provide a correct answer.
 - Several aspects of the implementation have been tested and tweaked
   for performance. See:
   - use of vector<int8_t> in Convex::findExtremeVertex
   - cache-friendly implementation of adjacency graph Convex::neighbors_.

Run Details

285 of 393 relevant lines covered (72.52%)

664185.94 hits per line

flexible-collision-library / fcl / 1868 / 1
73%
master: 73%

README BADGES
x

Markdown

Textile

RDoc

HTML

Rst

Source Files on job 1868.1 (BUILD_TYPE=Debug COVERALLS=ON)

flexible-collision-library / fcl / 1868 / 1 73% master: 73%

README BADGES x

Markdown

Textile

RDoc

HTML

Rst

Source Files on job 1868.1 (BUILD_TYPE=Debug COVERALLS=ON)

flexible-collision-library / fcl / 1868 / 1
73%
master: 73%

README BADGES
x