improve performance for string range indexing

[KristofferC]

using BenchmarkTools
using Random
using BenchmarkTools
s = randstring(200)
f_ib(s) = @inbounds s[1:100]
f(s) = s[1:100]

Master

julia> @btime f($s);
  103.603 ns (1 allocation: 128 bytes)

julia> @btime f_ib($s);
  103.692 ns (1 allocation: 128 bytes)

PR

julia> @btime f($s);
  37.007 ns (1 allocation: 128 bytes)

julia> @btime f_ib($s);
  25.304 ns (1 allocation: 128 bytes)

Not sure we have any Nanosoldier for this but might as well

Also added some @inline where functions expected themselves to get inlined (since they used @boundscheck).

@nanosoldier runbenchmarks(ALL, vs = ":master")

[KristofferC]

Made some changes, fire off another one

@nanosoldier runbenchmarks(ALL, vs = ":master")

[timholy]

Ref #29798

[nanosoldier]

Your benchmark job has completed - possible performance regressions were detected. A full report can be found here. cc @ararslan

[nanosoldier]

Your benchmark job has completed - possible performance regressions were detected. A full report can be found here. cc @ararslan

[KristofferC]

I checked the small memory regression in spellcheck and it seems to occur when getindex with a range is inlined but only in the very specific case when the range indexing of a string is inside a double comprehension. Rewriting the benchmark to avoid this apparently pathological case the performance is better and since we now can use @inbounds it is significantly better if that is used.

I am therefore of the opinion that this regression is acceptable and that in other code the benefit of inlining getindex is worth it. Will merge in a few days unless someone objects to the analysis above.