Specialize iterate

[jishnub]

On master:

julia> @btime fill(2, 300, 300);
  61.167 μs (2 allocations: 703.17 KiB)

julia> x = Fill(2, 300, 300);

julia> @btime [i for i in $x];
  77.769 μs (2 allocations: 703.17 KiB)

This PR makes them both equally fast:

julia> @btime [i for i in $x];
  60.280 μs (2 allocations: 703.17 KiB)

I've also removed some specializations of sum, as the specialized Base._mapreduce_dim covers these. On this PR:

julia> @btime sum(x->x^2, $x);
  3.355 ns (0 allocations: 0 bytes)

julia> @btime sum($x);
  3.358 ns (0 allocations: 0 bytes)

[codecov]

Codecov Report

Merging #181 (f6f0fd0) into master (d04c596) will increase coverage by 0.00%.
The diff coverage is 100.00%.

@@           Coverage Diff           @@
##           master     #181   +/-   ##
=======================================
  Coverage   96.94%   96.95%           
=======================================
  Files           4        4           
  Lines         655      657    +2     
=======================================
+ Hits          635      637    +2     
  Misses         20       20           

Impacted Files	Coverage Δ
src/FillArrays.jl	95.57% <100.00%> (+0.02%)	⬆️

📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more

[jishnub]

This change helps in cases where types aren't inferred:

julia> x = (1:3, 1:3); y = Any[x];

julia> @btime (x -> map(length, x[1]))($y);
  23.936 ns (1 allocation: 32 bytes)

julia> @btime (x -> length.(x[1]))($y);
  367.806 ns (3 allocations: 128 bytes)

[jishnub]

Gentle bump @dlfivefifty

[jishnub]

Gentle bump @dlfivefifty

[jishnub]

For the record, this performance match is obtained on v1.9. The performance is still somewhat inferior on v1.8.
On v1.9

julia> x = Fill(2, 300, 300);

julia> @benchmark [i for i in $x]
BenchmarkTools.Trial: 10000 samples with 1 evaluation.
 Range (min … max):  60.455 μs …  1.435 ms  ┊ GC (min … max): 0.00% …  0.00%
 Time  (median):     68.263 μs              ┊ GC (median):    0.00%
 Time  (mean ± σ):   75.443 μs ± 41.992 μs  ┊ GC (mean ± σ):  6.53% ± 10.20%

  ▅█▅▃▂▁▁                                                     ▁
  ██████████▇▆▅▅▆▆▄▃▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▄▄▅▇▇ █
  60.5 μs      Histogram: log(frequency) by time       375 μs <

 Memory estimate: 703.17 KiB, allocs estimate: 2.

julia> @benchmark fill(2, 300, 300)
BenchmarkTools.Trial: 10000 samples with 1 evaluation.
 Range (min … max):  59.974 μs … 715.494 μs  ┊ GC (min … max): 0.00% … 90.87%
 Time  (median):     69.090 μs               ┊ GC (median):    0.00%
 Time  (mean ± σ):   79.698 μs ±  62.497 μs  ┊ GC (mean ± σ):  9.81% ± 10.92%

  ▇█▄▃▂▂▁                                                      ▂
  █████████▇▄▄▃▃▄▄▁▁▁▁▁▁▁▃▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▄▆▇▇ █
  60 μs         Histogram: log(frequency) by time       552 μs <

 Memory estimate: 703.17 KiB, allocs estimate: 2.

On Julia v1.8:

julia> @benchmark fill(2, 300, 300)
BenchmarkTools.Trial: 10000 samples with 1 evaluation.
 Range (min … max):  58.523 μs … 576.305 μs  ┊ GC (min … max): 0.00% … 55.65%
 Time  (median):     68.670 μs               ┊ GC (median):    0.00%
 Time  (mean ± σ):   76.727 μs ±  40.618 μs  ┊ GC (mean ± σ):  6.07% ± 10.07%

  ▂██▅▄▃▂▁▁                                                    ▂
  ███████████▇▇▆▅▄▃▄▄▁▃▁▁▁▁▁▁▁▁▁▁▁▁▁▄▄▄▅▆▆▅▆▅▃▄▃▄▁▁▁▁▁▁▁▁▁▄▅▇█ █
  58.5 μs       Histogram: log(frequency) by time       358 μs <

 Memory estimate: 703.17 KiB, allocs estimate: 2.

julia> @benchmark [i for i in $x]
BenchmarkTools.Trial: 10000 samples with 1 evaluation.
 Range (min … max):  67.364 μs …  2.151 ms  ┊ GC (min … max): 0.00% …  0.00%
 Time  (median):     75.189 μs              ┊ GC (median):    0.00%
 Time  (mean ± σ):   86.099 μs ± 65.544 μs  ┊ GC (mean ± σ):  8.51% ± 10.57%

  █▇▄▂▁                                                       ▂
  ██████▇▇▆▅▄▄▅▄▁▁▃▃▁▁▁▁▁▁▁▁▃▁▅▆▇▆▆▄▁▁▁▃▁▁▁▁▃▁▁▁▁▁▁▁▁▁▁▁▁▃▆▇▇ █
  67.4 μs      Histogram: log(frequency) by time       537 μs <

 Memory estimate: 703.17 KiB, allocs estimate: 2.