Integer exponentiation for immutables of 3 or 4 (maybe more) floats broken

[twadleigh]

Floating point exponentiation appears to work for all the types.

Integer exponentiation now appears to work for all tuples, although for a while it was also broken for 3-tuples. Integer exponentiation also seems to work for immutables of two floats, but not 3 or 4.

This regression first appeared with fa315c9. Discussion on that thread suggests that the problem may be in LLVM (which puts it beyond my capability to track down further).

Some test code:

function test(a,n,b)
    an = a.^n
    pf = an == b ? "pass" : "fail"
    println("$a.^$n == $an; $pf")
end

.^(v::(Float64,Float64),n::Number) = (v[1].^n,v[2].^n)
.^(v::(Float64,Float64,Float64),n::Number) = (v[1].^n,v[2].^n,v[3].^n)
.^(v::(Float64,Float64,Float64,Float64),n::Number) = (v[1].^n,v[2].^n,v[3].^n,v[4].^n)

immutable V2
    x :: Float64
    y :: Float64
end

.^(v::V2, n::Number) = V2(v.x.^n,v.y.^n)

immutable V3
    x :: Float64
    y :: Float64
    z :: Float64
end

.^(v::V3, n::Number) = V3(v.x.^n,v.y.^n,v.z.^n)

immutable V4
    x :: Float64
    y :: Float64
    z :: Float64
    w :: Float64
end

.^(v::V4, n::Number) = V4(v.x.^n,v.y.^n,v.z.^n,v.w.^n)

t2  = (1.0,2.0)
t22 = (1.0,4.0)
t3  = (1.0,2.0,3.0)
t32 = (1.0,4.0,9.0)
t4  = (1.0,2.0,3.0,4.0)
t42 = (1.0,4.0,9.0,16.0)
v2  = V2(1.0,2.0)
v22 = V2(1.0,4.0)
v3  = V3(1.0,2.0,3.0)
v32 = V3(1.0,4.0,9.0)
v4  = V4(1.0,2.0,3.0,4.0)
v42 = V4(1.0,4.0,9.0,16.0)

for case in [ (t2,t22), (t3,t32), (t4,t42), (v2,v22), (v3,v32), (v4,v42) ]
    test(case[1],2.0,case[2])
    test(case[1],2,case[2])
end

produces:

(1.0,2.0).^2.0 == (1.0,4.0); pass
(1.0,2.0).^2 == (1.0,4.0); pass
(1.0,2.0,3.0).^2.0 == (1.0,4.0,9.0); pass
(1.0,2.0,3.0).^2 == (1.0,4.0,9.0); pass
(1.0,2.0,3.0,4.0).^2.0 == (1.0,4.0,9.0,16.0); pass
(1.0,2.0,3.0,4.0).^2 == (1.0,4.0,9.0,16.0); pass
V2(1.0,2.0).^2.0 == V2(1.0,4.0); pass
V2(1.0,2.0).^2 == V2(1.0,4.0); pass
V3(1.0,2.0,3.0).^2.0 == V3(1.0,4.0,9.0); pass
V3(1.0,2.0,3.0).^2 == V3(9.0,9.0,NaN); fail
V4(1.0,2.0,3.0,4.0).^2.0 == V4(1.0,4.0,9.0,16.0); pass
V4(1.0,2.0,3.0,4.0).^2 == V4(16.0,16.0,NaN,NaN); fail

[JeffBezanson]

dup of #6506