Broadcasting doesn't seem to be vectorised

[giordano]

On c14fbe8:

julia> using BenchmarkTools, FixedSizeArrays

julia> @benchmark B .= a .* A setup=(A = FixedSizeVector(randn(16)); B = similar(A); a = randn())
BenchmarkTools.Trial: 10000 samples with 980 evaluations.
 Range (min … max):  63.871 ns … 93.812 ns  ┊ GC (min … max): 0.00% … 0.00%
 Time  (median):     64.097 ns              ┊ GC (median):    0.00%
 Time  (mean ± σ):   65.218 ns ±  3.136 ns  ┊ GC (mean ± σ):  0.00% ± 0.00%

  █▅▄▄ ▃▂▁▁                                                   ▁
  ██████████▅▇▇▄▄▅▄▅▄▄▅▆▅▆▇▇▇▆▅▆▅▅▄▅▅▆▇▅▆▅▄▅▅▄▃▃▅▅▇██▇▇▇▇█▇▆▇ █
  63.9 ns      Histogram: log(frequency) by time      78.1 ns <

 Memory estimate: 0 bytes, allocs estimate: 0.

julia> @benchmark B .= a .* A setup=(A = randn(16); B = similar(A); a = randn())
BenchmarkTools.Trial: 10000 samples with 999 evaluations.
 Range (min … max):  9.765 ns … 34.962 ns  ┊ GC (min … max): 0.00% … 0.00%
 Time  (median):     9.789 ns              ┊ GC (median):    0.00%
 Time  (mean ± σ):   9.895 ns ±  0.860 ns  ┊ GC (mean ± σ):  0.00% ± 0.00%

  █▂  ▁▁                                                     ▁
  ██▇▇██▆▆▆▄▅▄▅▅▄▁▄▄▅▃▅▁▄▄▃▁▃▃▃▁▃▃▃▄▁▅▄▃▃▁▃▃▅▃▁▄▅▄▄▃▁▃▄▄▄▄▁▃ █
  9.76 ns      Histogram: log(frequency) by time     13.9 ns <

 Memory estimate: 0 bytes, allocs estimate: 0.

julia> versioninfo()
Julia Version 1.12.0-DEV.1170
Commit 95c643a6892 (2024-09-08 12:42 UTC)
Build Info:
  Official https://julialang.org release
Platform Info:
  OS: Linux (x86_64-linux-gnu)
  CPU: 8 × Intel(R) Core(TM) i7-4870HQ CPU @ 2.50GHz
  WORD_SIZE: 64
  LLVM: libLLVM-18.1.7 (ORCJIT, haswell)
Threads: 1 default, 0 interactive, 1 GC (on 8 virtual cores)

[nsajko]

julia> using BenchmarkTools, FixedSizeArrays

julia> (@benchmark B .= a .* A setup=(A = randn(16); B = similar(A); a = randn()));

julia> median(ans).time
6.713

julia> (@benchmark B .= a .* A setup=(A = FixedSizeVector(randn(16)); B = similar(A); a = randn()));

julia> median(ans).time
51.17933130699088

julia> Base.dataids(a::FixedSizeArray) = Base.dataids(a.mem)

julia> (@benchmark B .= a .* A setup=(A = FixedSizeVector(randn(16)); B = similar(A); a = randn()));

julia> median(ans).time
7.661661661661662

xref #53