codegen: ensure Bool arithmetic is done in mod-2

[vtjnash]

As @stevengj found in #31486, we're doing boolean arithmetic wrong in codegen, and that's causing us to potentially give up lots of performance since we can't vectorize it, and in addition we get wrong answers.

setup:

julia> xor_(a, b) = Core.Intrinsics.xor_int(a, b);
julia> b = rand(Bool, 150); i = UInt8.(b);
julia> f(a) = reduce(xor, a); g(a) = reduce(!=, a); h(a) = reduce(xor_, a);
julia> using BenchmarkTools

before:

julia> @btime f(b)
  115.304 ns (0 allocations: 0 bytes)
true

julia> @btime f(i)
  34.677 ns (0 allocations: 0 bytes)
0x01

julia> @btime g(b)
  114.314 ns (0 allocations: 0 bytes)
true

julia> @btime g(i)
  119.157 ns (0 allocations: 0 bytes)
true

julia> @btime h(b)
  33.699 ns (0 allocations: 0 bytes)
true

julia> @btime h(i)
  35.412 ns (0 allocations: 0 bytes)
0x01

after:

julia> @btime f(b)
  38.064 ns (0 allocations: 0 bytes)
true

julia> @btime f(i)
  34.773 ns (0 allocations: 0 bytes)
0x01

julia> @btime g(b)
  35.565 ns (0 allocations: 0 bytes)
true

julia> @btime g(i)
  118.535 ns (0 allocations: 0 bytes)
true

julia> @btime h(b)
  39.737 ns (0 allocations: 0 bytes)
true

julia> @btime h(i)
  34.269 ns (0 allocations: 0 bytes)
0x01

julia> @btime h(i)
  34.299 ns (0 allocations: 0 bytes)
0x01

PS: ignore the first commit, it's just deleting dead code

[vtjnash]

@nanosoldier runbenchmarks(ALL, ":master")

[vtjnash]

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

[nanosoldier]

Something went wrong when running your job:

NanosoldierError: failed to run benchmarks against primary commit: failed process: Process(`make -j3 USECCACHE=1`, ProcessExited(2)) [2]

Logs and partial data can be found here
cc @ararslan

[ararslan]

Nanosoldier is currently broken due to the BinaryBuilder changes.

[ararslan]

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

[nanosoldier]

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

[vtjnash]

Looks like it's probably a wash, so no major issues created by fixing this, but lets see if it is repeatable: @nanosoldier runbenchmarks(ALL, vs=":master")

[nanosoldier]

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

[rfourquet]

"mod 2" instead of "mod 1" ?