Example of @simd and 32-bit floating point.

gistfile1.txt

$ cat v.jl
function vadd_one!(arr::Array{Float32, 1})
    @simd for i = 1:length(arr)
         @inbounds arr[i] += 1
    end
end

code_llvm(vadd_one!,(Array{Float32, 1},))
$ julia v.jl

define void @"julia_vadd_one!15588"(%jl_value_t*) {
top:
  %1 = getelementptr inbounds %jl_value_t* %0, i64 2, i32 0, !dbg !175
  %2 = load %jl_value_t** %1, align 8, !dbg !175, !tbaa %jtbaa_arraylen
  %3 = ptrtoint %jl_value_t* %2 to i64, !dbg !175
  %4 = call { i64, i1 } @llvm.ssub.with.overflow.i64(i64 %3, i64 1), !dbg !175
  %5 = extractvalue { i64, i1 } %4, 1, !dbg !175
  br i1 %5, label %fail, label %pass, !dbg !175

fail:                                             ; preds = %top
  %6 = load %jl_value_t** @jl_overflow_exception, align 8, !dbg !175
  call void @jl_throw_with_superfluous_argument(%jl_value_t* %6, i32 41), !dbg !175
  unreachable, !dbg !175

pass:                                             ; preds = %top
  %7 = extractvalue { i64, i1 } %4, 0, !dbg !175
  %8 = call { i64, i1 } @llvm.sadd.with.overflow.i64(i64 %7, i64 1), !dbg !175
  %9 = extractvalue { i64, i1 } %8, 1, !dbg !175
  br i1 %9, label %fail1, label %pass2, !dbg !175

fail1:                                            ; preds = %pass
  %10 = load %jl_value_t** @jl_overflow_exception, align 8, !dbg !175
  call void @jl_throw_with_superfluous_argument(%jl_value_t* %10, i32 41), !dbg !175
  unreachable, !dbg !175

pass2:                                            ; preds = %pass
  %11 = icmp sge %jl_value_t* %2, inttoptr (i64 1 to %jl_value_t*), !dbg !175
  %12 = extractvalue { i64, i1 } %8, 0, !dbg !175
  %13 = select i1 %11, i64 %12, i64 0, !dbg !175
  %14 = icmp slt i64 %13, 1, !dbg !180
  br i1 %14, label %L5, label %L.preheader, !dbg !180

L.preheader:                                      ; preds = %pass2
  %15 = getelementptr inbounds %jl_value_t* %0, i64 1, i32 0, !dbg !181
  %16 = load %jl_value_t** %15, align 8, !dbg !181, !tbaa %jtbaa_arrayptr, !llvm.mem.parallel_loop_access !183
  %17 = bitcast %jl_value_t* %16 to float*, !dbg !181
  %n.vec = and i64 %13, -8
  %end.idx.rnd.down12 = or i64 %n.vec, 1
  %cmp.zero = icmp eq i64 %end.idx.rnd.down12, 1
  br i1 %cmp.zero, label %middle.block, label %vector.body

vector.body:                                      ; preds = %L.preheader, %vector.body
  %index = phi i64 [ %index.next, %vector.body ], [ 1, %L.preheader ]
  %18 = add i64 %index, -1
  %19 = getelementptr float* %17, i64 %18, !dbg !181
  %20 = bitcast float* %19 to <4 x float>*
  %wide.load = load <4 x float>* %20, align 4
  %.sum = add i64 %index, 3
  %21 = getelementptr float* %17, i64 %.sum
  %22 = bitcast float* %21 to <4 x float>*
  %wide.load7 = load <4 x float>* %22, align 4
  %23 = fadd <4 x float> %wide.load, <float 1.000000e+00, float 1.000000e+00, float 1.000000e+00, float 1.000000e+00>
  %24 = fadd <4 x float> %wide.load7, <float 1.000000e+00, float 1.000000e+00, float 1.000000e+00, float 1.000000e+00>
  store <4 x float> %23, <4 x float>* %20, align 4
  store <4 x float> %24, <4 x float>* %22, align 4
  %index.next = add i64 %index, 8
  %25 = icmp eq i64 %index.next, %end.idx.rnd.down12
  br i1 %25, label %middle.block, label %vector.body

middle.block:                                     ; preds = %vector.body, %L.preheader
  %resume.val = phi i64 [ 1, %L.preheader ], [ %end.idx.rnd.down12, %vector.body ]
  %end.idx = add i64 %13, 1
  %cmp.n = icmp eq i64 %end.idx, %resume.val
  br i1 %cmp.n, label %L5, label %L

L:                                                ; preds = %middle.block, %L
  %i.0 = phi i64 [ %30, %L ], [ %resume.val, %middle.block ]
  %26 = add i64 %i.0, -1, !dbg !181
  %27 = getelementptr float* %17, i64 %26, !dbg !181
  %28 = load float* %27, align 4, !dbg !181, !tbaa %jtbaa_user, !llvm.mem.parallel_loop_access !183
  %29 = fadd float %28, 1.000000e+00, !dbg !181
  store float %29, float* %27, align 4, !dbg !181, !tbaa %jtbaa_user, !llvm.mem.parallel_loop_access !183
  %30 = add i64 %i.0, 1, !dbg !186
  %exitcond = icmp eq i64 %i.0, %13, !dbg !187
  br i1 %exitcond, label %L5, label %L, !dbg !187, !llvm.loop.parallel !184, !llvm.vectorizer.already_vectorized !184

L5:                                               ; preds = %middle.block, %L, %pass2
  ret void, !dbg !187
}