Add support for minmax and bswap for Fixed

Although `minmax` has been available, but this specializes it.

Author: kimikage

src/FixedPointNumbers.jl

@@ -116,6 +116,13 @@ floattype(::Type{X}) where {T <: LongInts, X <: FixedPoint{T}} = BigFloat
 
 float(x::FixedPoint) = convert(floattype(x), x)
 
+function minmax(x::X, y::X) where {X <: FixedPoint}
+    a, b = minmax(reinterpret(x), reinterpret(y))
+    X(a,0), X(b,0)
+end
+
+bswap(x::X) where {X <: FixedPoint} = sizeof(X) == 1 ? x : X(bswap(x.i), 0)
+
 for f in (:zero, :oneunit, :one, :eps, :rawone, :rawtype, :floattype)
     @eval begin
         $f(x::FixedPoint) = $f(typeof(x))

src/normed.jl

@@ -269,14 +269,6 @@ isfinite(x::Normed) = true
 isnan(x::Normed) = false
 isinf(x::Normed) = false
 
-bswap(x::Normed{UInt8,f}) where {f} = x
-bswap(x::Normed)  = typeof(x)(bswap(reinterpret(x)),0)
-
-function minmax(x::T, y::T) where {T <: Normed}
-    a, b = minmax(reinterpret(x), reinterpret(y))
-    T(a,0), T(b,0)
-end
-
 # Iteration
 # The main subtlety here is that iterating over N0f8(0):N0f8(1) will wrap around
 # unless we iterate using a wider type

test/fixed.jl

@@ -180,6 +180,11 @@ end
     end
 end
 
+@testset "low-level arithmetic" begin
+    @test bswap(Q0f7(0.5)) === Q0f7(0.5)
+    @test bswap(Q0f15(0.5)) === reinterpret(Q0f15, signed(0x0040))
+end
+
 @testset "Promotion within Fixed" begin
     @test @inferred(promote(Q0f7(0.25), Q0f7(0.75))) ===
         (Q0f7(0.25), Q0f7(0.75))