General dimensional refinement

[hyrodium]

No description provided.

[codecov]

Codecov Report

Attention: 41 lines in your changes are missing coverage. Please review.

Comparison is base (b405724) 96.27% compared to head (1371c1c) 93.60%.

Files	Patch %	Lines
src/_Refinement.jl	56.38%	41 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #354      +/-   ##
==========================================
- Coverage   96.27%   93.60%   -2.67%     
==========================================
  Files          14       14              
  Lines        1583     1642      +59     
==========================================
+ Hits         1524     1537      +13     
- Misses         59      105      +46     

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

[hyrodium]

Performance improvements in this PR! 🚀

Before this PR

julia> using BasicBSpline, BenchmarkTools

julia> P1 = BSplineSpace{2}(knotvector"1112121113")
BSplineSpace{2, Int64, KnotVector{Int64}}(KnotVector([1, 2, 3, 4, 4, 5, 6, 6, 7, 8, 9, 10, 10, 10]))

julia> P2 = BSplineSpace{1}(knotvector"11111 12")
BSplineSpace{1, Int64, KnotVector{Int64}}(KnotVector([1, 2, 3, 4, 5, 7, 8, 8]))

julia> P1′ = expandspace_I(P1, Val(1))
BSplineSpace{3, Int64, KnotVector{Int64}}(KnotVector([1, 2, 3, 3, 4, 4, 4, 5, 5, 6, 6, 6, 7, 7, 8, 8, 9, 9, 10, 10, 10, 10]))

julia> P2′ = expandspace_I(P2, Val(1))
BSplineSpace{2, Int64, KnotVector{Int64}}(KnotVector([1, 2, 2, 3, 3, 4, 4, 5, 5, 7, 7, 8, 8, 8]))

julia> n1 = dim(P1)
11

julia> n2 = dim(P2)
6

julia> a = rand(n1, n2);

julia> M = BSplineManifold(a, P1, P2);

julia> @benchmark refinement(M, (P1, P2))
BenchmarkTools.Trial: 10000 samples with 1 evaluation.
 Range (min … max):  42.751 μs …  2.508 ms  ┊ GC (min … max): 0.00% … 95.87%
 Time  (median):     44.344 μs              ┊ GC (median):    0.00%
 Time  (mean ± σ):   45.398 μs ± 34.695 μs  ┊ GC (mean ± σ):  1.06% ±  1.36%

    ▁▂▄▄▇██▆▄▁                                                 
  ▂▅██████████▇▇▆▅▅▅▄▄▄▄▄▄▃▃▃▃▃▂▂▂▂▂▂▂▂▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁ ▃
  42.8 μs         Histogram: frequency by time        52.7 μs <

 Memory estimate: 8.47 KiB, allocs estimate: 87.

After this PR

julia> using BasicBSpline, BenchmarkTools

julia> P1 = BSplineSpace{2}(knotvector"1112121113")
BSplineSpace{2, Int64, KnotVector{Int64}}(KnotVector([1, 2, 3, 4, 4, 5, 6, 6, 7, 8, 9, 10, 10, 10]))

julia> P2 = BSplineSpace{1}(knotvector"11111 12")
BSplineSpace{1, Int64, KnotVector{Int64}}(KnotVector([1, 2, 3, 4, 5, 7, 8, 8]))

julia> P1′ = expandspace_I(P1, Val(1))
BSplineSpace{3, Int64, KnotVector{Int64}}(KnotVector([1, 2, 3, 3, 4, 4, 4, 5, 5, 6, 6, 6, 7, 7, 8, 8, 9, 9, 10, 10, 10, 10]))

julia> P2′ = expandspace_I(P2, Val(1))
BSplineSpace{2, Int64, KnotVector{Int64}}(KnotVector([1, 2, 2, 3, 3, 4, 4, 5, 5, 7, 7, 8, 8, 8]))

julia> n1 = dim(P1)
11

julia> n2 = dim(P2)
6

julia> a = rand(n1, n2);

julia> M = BSplineManifold(a, P1, P2);

julia> @benchmark refinement_I(M, (P1, P2))
BenchmarkTools.Trial: 10000 samples with 5 evaluations.
 Range (min … max):  6.582 μs … 536.418 μs  ┊ GC (min … max): 0.00% … 96.21%
 Time  (median):     7.013 μs               ┊ GC (median):    0.00%
 Time  (mean ± σ):   7.704 μs ±  15.731 μs  ┊ GC (mean ± σ):  5.97% ±  2.89%

   ▁▄▆▇███▇▆▆▄▃▂                ▁▁ ▁▁▁▁▁▁▁▁▁▁▁                ▂
  ▅████████████████▇██▇▇▇▇▇▇▇▇▇██████████████████▆▇▆▇▇▆▆▆▅▅▅▆ █
  6.58 μs      Histogram: log(frequency) by time      10.1 μs <

 Memory estimate: 8.86 KiB, allocs estimate: 89.

[hyrodium]

Okay, let's merge this for now. Here's a TODO list after fixing #353.

• Add more test cases
• Fix type inference
• Check performance
• Remove Base.promote_op use