Example using Julia and LLVM trunk. SLPVectorizer appears to have generated superfluious extracts.

gistfile1.txt

$ cat vadd.jl
function vadd( a::NTuple{4,Float64}, b::NTuple{4,Float64} )
    (a[1]+b[1],a[2]+b[2],a[3]+b[3],a[4]+b[4])
end

function vadd_one!(arr::Array{Float64, 1})
    len  = length(arr) # assuming len multiple of 4
    one =  (1.0, 1.0, 1.0, 1.0)
    @inbounds for i = 1:4:len
        inp = (arr[i], arr[i+1], arr[i+2], arr[i+3])
        out = vadd(inp, one)
        for j = 1:4
             arr[i + j] = out[j]
       end
    end
end

code_llvm(vadd_one!,(Array{Float64, 1},))
$ julia vadd.jl

define %jl_value_t* @"julia_vadd_one!15119"(%jl_value_t*, %jl_value_t**, i32) {
top:
  %3 = load %jl_value_t** %1, align 8, !dbg !8
  %4 = getelementptr inbounds %jl_value_t* %3, i64 2, i32 0, !dbg !9
  %5 = load %jl_value_t** %4, align 8, !dbg !9, !tbaa %jtbaa_arraylen
  %6 = ptrtoint %jl_value_t* %5 to i64, !dbg !9
  %7 = call %Range @julia_colon820(i64 1, i64 4, i64 %6), !dbg !14, !julia_type !15
  %8 = extractvalue %Range %7, 2, !dbg !14
  %9 = icmp slt i64 %8, 1, !dbg !14
  br i1 %9, label %L9, label %pass2.preheader, !dbg !14

pass2.preheader:                                  ; preds = %top
  %10 = extractvalue %Range %7, 0, !dbg !14
  %11 = extractvalue %Range %7, 1, !dbg !14
  %12 = getelementptr inbounds %jl_value_t* %3, i64 1, i32 0, !dbg !16
  %13 = load %jl_value_t** %12, align 8, !dbg !16, !tbaa %jtbaa_arrayptr
  br label %pass2, !dbg !14

pass2:                                            ; preds = %pass2, %pass2.preheader
  %"#s2.0" = phi i64 [ %39, %pass2 ], [ 0, %pass2.preheader ]
  %14 = mul i64 %"#s2.0", %11, !dbg !14
  %15 = add i64 %14, %10, !dbg !14
  %16 = add i64 %15, -1, !dbg !16
  %17 = getelementptr %jl_value_t* %13, i64 %16, !dbg !16
  %18 = bitcast %jl_value_t* %17 to <2 x double>*, !dbg !16
  %19 = load <2 x double>* %18, align 8, !dbg !16, !tbaa %jtbaa_user
  %20 = add i64 %15, 1, !dbg !16
  %21 = getelementptr %jl_value_t* %13, i64 %20, !dbg !16
  %22 = bitcast %jl_value_t* %21 to <2 x double>*, !dbg !16
  %23 = load <2 x double>* %22, align 8, !dbg !16, !tbaa %jtbaa_user
  %24 = fadd <2 x double> %19, <double 1.000000e+00, double 1.000000e+00>, !dbg !19
  %25 = fadd <2 x double> %23, <double 1.000000e+00, double 1.000000e+00>, !dbg !19
  %26 = extractelement <2 x double> %24, i32 0, !dbg !20
  %27 = getelementptr %jl_value_t* %13, i64 %15, !dbg !20
  %28 = bitcast %jl_value_t* %27 to double*, !dbg !20
  store double %26, double* %28, align 8, !dbg !20, !tbaa %jtbaa_user
  %29 = extractelement <2 x double> %24, i32 1, !dbg !20
  %30 = bitcast %jl_value_t* %21 to double*, !dbg !20
  store double %29, double* %30, align 8, !dbg !20, !tbaa %jtbaa_user
  %31 = extractelement <2 x double> %25, i32 0, !dbg !20
  %32 = add i64 %15, 2, !dbg !20
  %33 = getelementptr %jl_value_t* %13, i64 %32, !dbg !20
  %34 = bitcast %jl_value_t* %33 to double*, !dbg !20
  store double %31, double* %34, align 8, !dbg !20, !tbaa %jtbaa_user
  %35 = extractelement <2 x double> %25, i32 1, !dbg !20
  %36 = add i64 %15, 3, !dbg !20
  %37 = getelementptr %jl_value_t* %13, i64 %36, !dbg !20
  %38 = bitcast %jl_value_t* %37 to double*, !dbg !20
  store double %35, double* %38, align 8, !dbg !20, !tbaa %jtbaa_user
  %39 = add i64 %"#s2.0", 1, !dbg !14
  %exitcond = icmp eq i64 %39, %8, !dbg !20
  br i1 %exitcond, label %L9, label %pass2, !dbg !20

L9:                                               ; preds = %pass2, %top
  ret %jl_value_t* inttoptr (i64 13122672 to %jl_value_t*), !dbg !20
}