llvm.jl

julia> @code_llvm diff(a)

; Function diff
; Location: REPL[7]:1
define void @julia_diff_35329([19 x i64]* noalias nocapture sret, [20 x i64] addrspace(11)* nocapture nonnull readonly dereferenceable(160)) {
top:
; Function getindex; {
; Location: tuple.jl:24
  %2 = getelementptr [20 x i64], [20 x i64] addrspace(11)* %1, i64 0, i64 1
  %3 = getelementptr [20 x i64], [20 x i64] addrspace(11)* %1, i64 0, i64 0
;}
; Function -; {
; Location: int.jl:52
  %4 = load i64, i64 addrspace(11)* %3, align 8
;}
; Function tail; {
; Location: essentials.jl:173
  %5 = getelementptr [20 x i64], [20 x i64] addrspace(11)* %1, i64 0, i64 5
  %6 = getelementptr [20 x i64], [20 x i64] addrspace(11)* %1, i64 0, i64 9
  %7 = getelementptr [20 x i64], [20 x i64] addrspace(11)* %1, i64 0, i64 13
  %8 = getelementptr [20 x i64], [20 x i64] addrspace(11)* %1, i64 0, i64 17
  %9 = getelementptr [20 x i64], [20 x i64] addrspace(11)* %1, i64 0, i64 18
;}
; Function -; {
; Location: int.jl:52
  %10 = bitcast i64 addrspace(11)* %2 to <4 x i64> addrspace(11)*
  %11 = load <4 x i64>, <4 x i64> addrspace(11)* %10, align 8
  %12 = insertelement <4 x i64> undef, i64 %4, i32 0
  %13 = shufflevector <4 x i64> %12, <4 x i64> %11, <4 x i32> <i32 0, i32 4, i32 5, i32 6>
  %14 = sub <4 x i64> %11, %13
;}
; Function diff; {
; Location: REPL[7]:1
; Function diff; {
; Location: REPL[7]:1
; Function diff; {
; Location: REPL[7]:1
; Function diff; {
; Location: REPL[7]:1
; Function -; {
; Location: int.jl:52
  %15 = bitcast i64 addrspace(11)* %5 to <4 x i64> addrspace(11)*
  %16 = load <4 x i64>, <4 x i64> addrspace(11)* %15, align 8
  %17 = shufflevector <4 x i64> %11, <4 x i64> %16, <4 x i32> <i32 3, i32 4, i32 5, i32 6>
  %18 = sub <4 x i64> %16, %17
;}
; Function diff; {
; Location: REPL[7]:1
; Function diff; {
; Location: REPL[7]:1
; Function diff; {
; Location: REPL[7]:1
; Function diff; {
; Location: REPL[7]:1
; Function -; {
; Location: int.jl:52
  %19 = bitcast i64 addrspace(11)* %6 to <4 x i64> addrspace(11)*
  %20 = load <4 x i64>, <4 x i64> addrspace(11)* %19, align 8
  %21 = shufflevector <4 x i64> %16, <4 x i64> %20, <4 x i32> <i32 3, i32 4, i32 5, i32 6>
  %22 = sub <4 x i64> %20, %21
;}
; Function diff; {
; Location: REPL[7]:1
; Function diff; {
; Location: REPL[7]:1
; Function diff; {
; Location: REPL[7]:1
; Function diff; {
; Location: REPL[7]:1
; Function -; {
; Location: int.jl:52
  %23 = bitcast i64 addrspace(11)* %7 to <4 x i64> addrspace(11)*
  %24 = load <4 x i64>, <4 x i64> addrspace(11)* %23, align 8
  %25 = shufflevector <4 x i64> %20, <4 x i64> %24, <4 x i32> <i32 3, i32 4, i32 5, i32 6>
  %26 = sub <4 x i64> %24, %25
;}
; Function diff; {
; Location: REPL[7]:1
; Function diff; {
; Location: REPL[7]:1
; Function diff; {
; Location: REPL[7]:1
; Function diff; {
; Location: REPL[7]:1
; Function -; {
; Location: int.jl:52
  %27 = load i64, i64 addrspace(11)* %8, align 8
  %28 = extractelement <4 x i64> %24, i32 3
  %29 = sub i64 %27, %28
;}
; Function diff; {
; Location: REPL[7]:1
; Function -; {
; Location: int.jl:52
  %30 = bitcast i64 addrspace(11)* %9 to <2 x i64> addrspace(11)*
  %31 = load <2 x i64>, <2 x i64> addrspace(11)* %30, align 8
  %32 = insertelement <2 x i64> undef, i64 %27, i32 0
  %33 = shufflevector <2 x i64> %32, <2 x i64> %31, <2 x i32> <i32 0, i32 2>
  %34 = sub <2 x i64> %31, %33
;}}}}}}}}}}}}}}}}}}
  %35 = bitcast [19 x i64]* %0 to <4 x i64>*
  store <4 x i64> %14, <4 x i64>* %35, align 8
  %.sroa.5.0..sroa_idx4 = getelementptr inbounds [19 x i64], [19 x i64]* %0, i64 0, i64 4
  %36 = bitcast i64* %.sroa.5.0..sroa_idx4 to <4 x i64>*
  store <4 x i64> %18, <4 x i64>* %36, align 8
  %.sroa.9.0..sroa_idx8 = getelementptr inbounds [19 x i64], [19 x i64]* %0, i64 0, i64 8
  %37 = bitcast i64* %.sroa.9.0..sroa_idx8 to <4 x i64>*
  store <4 x i64> %22, <4 x i64>* %37, align 8
  %.sroa.13.0..sroa_idx12 = getelementptr inbounds [19 x i64], [19 x i64]* %0, i64 0, i64 12
  %38 = bitcast i64* %.sroa.13.0..sroa_idx12 to <4 x i64>*
  store <4 x i64> %26, <4 x i64>* %38, align 8
  %.sroa.17.0..sroa_idx16 = getelementptr inbounds [19 x i64], [19 x i64]* %0, i64 0, i64 16
  store i64 %29, i64* %.sroa.17.0..sroa_idx16, align 8
  %.sroa.18.0..sroa_idx17 = getelementptr inbounds [19 x i64], [19 x i64]* %0, i64 0, i64 17
  %39 = bitcast i64* %.sroa.18.0..sroa_idx17 to <2 x i64>*
  store <2 x i64> %34, <2 x i64>* %39, align 8
  ret void
}

julia> @code_llvm gen_diff(a)

; Function gen_diff
; Location: REPL[10]:2
define void @julia_gen_diff_35326([19 x i64]* noalias nocapture sret, [20 x i64] addrspace(11)* nocapture nonnull readonly dereferenceable(160)) {
top:
; Function macro expansion; {
; Location: REPL[10]
; Function getindex; {
; Location: tuple.jl:24
  %2 = getelementptr [20 x i64], [20 x i64] addrspace(11)* %1, i64 0, i64 1
  %3 = getelementptr [20 x i64], [20 x i64] addrspace(11)* %1, i64 0, i64 0
;}
; Function -; {
; Location: int.jl:52
  %4 = load i64, i64 addrspace(11)* %3, align 8
  %5 = bitcast i64 addrspace(11)* %2 to <4 x i64> addrspace(11)*
  %6 = load <4 x i64>, <4 x i64> addrspace(11)* %5, align 8
  %7 = insertelement <4 x i64> undef, i64 %4, i32 0
  %8 = shufflevector <4 x i64> %7, <4 x i64> %6, <4 x i32> <i32 0, i32 4, i32 5, i32 6>
  %9 = sub <4 x i64> %6, %8
;}
; Function getindex; {
; Location: tuple.jl:24
  %10 = getelementptr [20 x i64], [20 x i64] addrspace(11)* %1, i64 0, i64 5
;}
; Function -; {
; Location: int.jl:52
  %11 = bitcast i64 addrspace(11)* %10 to <4 x i64> addrspace(11)*
  %12 = load <4 x i64>, <4 x i64> addrspace(11)* %11, align 8
  %13 = shufflevector <4 x i64> %6, <4 x i64> %12, <4 x i32> <i32 3, i32 4, i32 5, i32 6>
  %14 = sub <4 x i64> %12, %13
;}
; Function getindex; {
; Location: tuple.jl:24
  %15 = getelementptr [20 x i64], [20 x i64] addrspace(11)* %1, i64 0, i64 9
;}
; Function -; {
; Location: int.jl:52
  %16 = bitcast i64 addrspace(11)* %15 to <4 x i64> addrspace(11)*
  %17 = load <4 x i64>, <4 x i64> addrspace(11)* %16, align 8
  %18 = shufflevector <4 x i64> %12, <4 x i64> %17, <4 x i32> <i32 3, i32 4, i32 5, i32 6>
  %19 = sub <4 x i64> %17, %18
;}
; Function getindex; {
; Location: tuple.jl:24
  %20 = getelementptr [20 x i64], [20 x i64] addrspace(11)* %1, i64 0, i64 13
;}
; Function -; {
; Location: int.jl:52
  %21 = bitcast i64 addrspace(11)* %20 to <4 x i64> addrspace(11)*
  %22 = load <4 x i64>, <4 x i64> addrspace(11)* %21, align 8
  %23 = shufflevector <4 x i64> %17, <4 x i64> %22, <4 x i32> <i32 3, i32 4, i32 5, i32 6>
  %24 = sub <4 x i64> %22, %23
;}
; Function getindex; {
; Location: tuple.jl:24
  %25 = getelementptr [20 x i64], [20 x i64] addrspace(11)* %1, i64 0, i64 17
;}
; Function -; {
; Location: int.jl:52
  %26 = load i64, i64 addrspace(11)* %25, align 8
  %27 = extractelement <4 x i64> %22, i32 3
  %28 = sub i64 %26, %27
;}
; Function getindex; {
; Location: tuple.jl:24
  %29 = getelementptr [20 x i64], [20 x i64] addrspace(11)* %1, i64 0, i64 18
;}
; Function -; {
; Location: int.jl:52
  %30 = bitcast i64 addrspace(11)* %29 to <2 x i64> addrspace(11)*
  %31 = load <2 x i64>, <2 x i64> addrspace(11)* %30, align 8
  %32 = insertelement <2 x i64> undef, i64 %26, i32 0
  %33 = shufflevector <2 x i64> %32, <2 x i64> %31, <2 x i32> <i32 0, i32 2>
  %34 = sub <2 x i64> %31, %33
;}
  %35 = bitcast [19 x i64]* %0 to <4 x i64>*
  store <4 x i64> %9, <4 x i64>* %35, align 8
  %.sroa.5.0..sroa_idx4 = getelementptr inbounds [19 x i64], [19 x i64]* %0, i64 0, i64 4
  %36 = bitcast i64* %.sroa.5.0..sroa_idx4 to <4 x i64>*
  store <4 x i64> %14, <4 x i64>* %36, align 8
  %.sroa.9.0..sroa_idx8 = getelementptr inbounds [19 x i64], [19 x i64]* %0, i64 0, i64 8
  %37 = bitcast i64* %.sroa.9.0..sroa_idx8 to <4 x i64>*
  store <4 x i64> %19, <4 x i64>* %37, align 8
  %.sroa.13.0..sroa_idx12 = getelementptr inbounds [19 x i64], [19 x i64]* %0, i64 0, i64 12
  %38 = bitcast i64* %.sroa.13.0..sroa_idx12 to <4 x i64>*
  store <4 x i64> %24, <4 x i64>* %38, align 8
  %.sroa.17.0..sroa_idx16 = getelementptr inbounds [19 x i64], [19 x i64]* %0, i64 0, i64 16
  store i64 %28, i64* %.sroa.17.0..sroa_idx16, align 8
  %.sroa.18.0..sroa_idx17 = getelementptr inbounds [19 x i64], [19 x i64]* %0, i64 0, i64 17
  %39 = bitcast i64* %.sroa.18.0..sroa_idx17 to <2 x i64>*
  store <2 x i64> %34, <2 x i64>* %39, align 8
  ret void
;}
}