FaceSkeleton Vroom 1

[AbdAlazezAhmed]

Makes face skeleton construction faster by using topology information.
Benchmarks:

julia> using Ferrite

julia> using BenchmarkTools

julia> grid = generate_grid(Hexahedron,(100,100,100))
Grid{3, Hexahedron, Float64} with 1000000 Hexahedron cells and 1030301 nodes

julia> topology = ExclusiveTopology(grid);

julia> @benchmark Ferrite. _faceskeleton(topology, grid)
BenchmarkTools.Trial: 19 samples with 1 evaluation.
 Range (min … max):  221.792 ms … 410.902 ms  ┊ GC (min … max): 0.00% … 0.00%
 Time  (median):     254.305 ms               ┊ GC (median):    0.00%
 Time  (mean ± σ):   271.663 ms ±  46.478 ms  ┊ GC (mean ± σ):  0.00% ± 0.00%

  ▁▁   ██▁█ ▁  █      █▁    ▁   ▁        ▁                    ▁  
  ██▁▁▁████▁█▁▁█▁▁▁▁▁▁██▁▁▁▁█▁▁▁█▁▁▁▁▁▁▁▁█▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁█ ▁
  222 ms           Histogram: frequency by time          411 ms <

 Memory estimate: 46.23 MiB, allocs estimate: 2.

Current implementation:

julia> using Ferrite

julia> using BenchmarkTools

julia> grid = generate_grid(Hexahedron,(100,100,100))
Grid{3, Hexahedron, Float64} with 1000000 Hexahedron cells and 1030301 nodes

julia> topology = ExclusiveTopology(grid);

julia> @benchmark Ferrite. _faceskeleton(topology, grid)
BenchmarkTools.Trial: 3 samples with 1 evaluation.
Range (min … max):  1.578 s …    1.932 s  ┊ GC (min … max): 0.00% … 18.59%
Time  (median):     1.763 s               ┊ GC (median):    0.00%
Time  (mean ± σ):   1.758 s ± 176.799 ms  ┊ GC (mean ± σ):  6.81% ± 10.74%

 █                             █                          █  
 █▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁█▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁█ ▁
 1.58 s         Histogram: frequency by time         1.93 s <

Memory estimate: 652.91 MiB, allocs estimate: 17026973.

grid argument is kept in case it's needed in the future. The if-else thing can be avoided if we consistently define "faces" as vertices for 1D case.

[codecov-commenter]

Codecov Report

All modified and coverable lines are covered by tests ✅

Comparison is base (d0f0e07) 92.98% compared to head (49bc732) 93.12%.
Report is 6 commits behind head on master.

❗ Current head 49bc732 differs from pull request most recent head 439978b. Consider uploading reports for the commit 439978b to get more accurate results

❗ Your organization needs to install the Codecov GitHub app to enable full functionality.

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #759      +/-   ##
==========================================
+ Coverage   92.98%   93.12%   +0.14%     
==========================================
  Files          33       33              
  Lines        4950     4949       -1     
==========================================
+ Hits         4603     4609       +6     
+ Misses        347      340       -7     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[termi-official]

Can we add some additional tests to check if the construction works out fine? Especially for the 1D case.

[termi-official]

LGTM

[AbdAlazezAhmed]

Trying to improve performance a bit more

julia> @benchmark Ferrite. _faceskeleton(topology, grid)
BenchmarkTools.Trial: 39 samples with 1 evaluation.
 Range (min … max):  116.631 ms … 164.498 ms  ┊ GC (min … max): 0.00% … 28.98%
 Time  (median):     129.052 ms               ┊ GC (median):    0.00%
 Time  (mean ± σ):   128.925 ms ±   6.938 ms  ┊ GC (mean ± σ):  0.95% ±  4.64%

                 █▁                                              
  ▄▁▃▁▁▁▁▁▁▁▁▁▁▁▁██▆▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▃ ▁
  117 ms           Histogram: frequency by time          164 ms <

 Memory estimate: 46.23 MiB, allocs estimate: 2.

Maybe we can also get rid of the extra allocation (and GC?)

[termi-official]

Can you resolve the conflict? Should be good to go then. 10x improvement is really nice. :)