add Float16 conversions (#28)

Author: simonbyrne

src/Quadmath.jl

@@ -136,6 +136,10 @@ Float128(x::Float32) =
 Float32(x::Float128) =
     @ccall(quadoplib.__trunctfsf2(x::Cfloat128)::Cfloat)
 
+# Float16
+Float128(x::Float16) = Float128(Float32(x))
+Float16(x::Float128) = Float16(Float32(x)) # TODO: avoid double rounding
+
 # integer -> Float128
 Float128(x::Int32) =
     Float128(@ccall(quadoplib.__floatsitf(x::Int32)::Cfloat128))

test/runtests.jl

@@ -11,7 +11,7 @@ using Quadmath
 end
 
 @testset "conversions" begin
-@testset "conversion $T" for T in (Float32, Float64, Int32, Int64, Int128, UInt32, UInt64, UInt128, BigFloat, BigInt)
+@testset "conversion $T" for T in (Float16, Float32, Float64, Int32, Int64, Int128, UInt32, UInt64, UInt128, BigFloat, BigInt)
     @test Float128(T(1)) + Float128(T(2)) == Float128(T(3))
     @test Float128(T(1)) + Float128(T(2)) <= Float128(T(3))
     @test Float128(T(1)) + Float128(T(2)) != Float128(T(4))