adjust base case size, add a few benchmarks to README

Author: chriselrod

README.md

@@ -46,3 +46,25 @@ searchsortedfirstcorrelated(v::AbstractVector, x, guess)
 
 An accelerated `findfirst` on sorted vectors using a bracketed search. Requires a `guess`
 to start the search from.
+
+
+Some benchmarks:
+```julia
+julia> x = rand(Int, 2048); s = sort(x);
+
+julia> @btime findfirst(==($x[1011]), $x)
+  266.427 ns (0 allocations: 0 bytes)
+1011
+
+julia> @btime FindFirstFunctions.findfirstequal($x[1011], $x)
+  67.502 ns (0 allocations: 0 bytes)
+1011
+
+julia> @btime searchsortedfirst($s, $s[1011])
+  8.897 ns (0 allocations: 0 bytes)
+1011
+
+julia> @btime FindFirstFunctions.findfirstsortedequal($s[1011], $s)
+  10.896 ns (0 allocations: 0 bytes)
+1011
+```

src/FindFirstFunctions.jl

@@ -79,7 +79,7 @@ end
 Note that this differs from `searchsortedfirst` by returning `nothing` when absent.
 """
 function findfirstsortedequal(var::Int64, vars::DenseVector{Int64},
-  ::Val{basecase}=Val(64)) where {basecase}
+  ::Val{basecase}=Val(16)) where {basecase}
   len = length(vars)
   offset = 0
   @inbounds while len > basecase