simd bug, llvm output

does_not_vectorize.txt

;  @ /data/esdrd/tqff/tmp/issue.jl:20 within `does_not_vectorize!'
define nonnull %jl_value_t addrspace(10)* @"japi1_does_not_vectorize!_16049"(%jl_value_t addrspace(10)*, %jl_value_t addrspace(10)**, i32) #0 {
top:
  %3 = alloca %jl_value_t addrspace(10)**, align 8
  store volatile %jl_value_t addrspace(10)** %1, %jl_value_t addrspace(10)*** %3, align 8
  %4 = load %jl_value_t addrspace(10)*, %jl_value_t addrspace(10)** %1, align 8
  %5 = getelementptr inbounds %jl_value_t addrspace(10)*, %jl_value_t addrspace(10)** %1, i64 1
  %6 = load %jl_value_t addrspace(10)*, %jl_value_t addrspace(10)** %5, align 8
; ┌ @ array.jl:157 within `size'
; │┌ @ ntuple.jl:43 within `ntuple'
; ││┌ @ array.jl:157 within `#69'
; │││┌ @ array.jl:154 within `size'
      %7 = addrspacecast %jl_value_t addrspace(10)* %6 to %jl_value_t addrspace(11)*
      %8 = bitcast %jl_value_t addrspace(11)* %7 to %jl_value_t addrspace(10)* addrspace(11)*
      %9 = getelementptr inbounds %jl_value_t addrspace(10)*, %jl_value_t addrspace(10)* addrspace(11)* %8, i64 5
      %10 = bitcast %jl_value_t addrspace(10)* addrspace(11)* %9 to i64 addrspace(11)*
      %11 = load i64, i64 addrspace(11)* %10, align 8
; └└└└
;  @ /data/esdrd/tqff/tmp/issue.jl:21 within `does_not_vectorize!'
  %12 = call nonnull %jl_value_t addrspace(10)* @julia_ntuple_16050(i64 9)
;  @ /data/esdrd/tqff/tmp/issue.jl:23 within `does_not_vectorize!'
; ┌ @ int.jl:52 within `-'
   %13 = add i64 %11, -5
; └
; ┌ @ range.jl:5 within `Colon'
; │┌ @ range.jl:277 within `UnitRange'
; ││┌ @ range.jl:282 within `unitrange_last'
; │││┌ @ operators.jl:341 within `>='
; ││││┌ @ int.jl:424 within `<='
       %14 = icmp sgt i64 %13, 4
; └└└└└
  br i1 %14, label %L13.L18_crit_edge, label %L208

L13.L18_crit_edge:                                ; preds = %top
;  @ /data/esdrd/tqff/tmp/issue.jl:20 within `does_not_vectorize!'
; ┌ @ array.jl:157 within `size'
; │┌ @ ntuple.jl:43 within `ntuple'
; ││┌ @ array.jl:157 within `#69'
; │││┌ @ array.jl:154 within `size'
      %15 = getelementptr inbounds %jl_value_t addrspace(10)*, %jl_value_t addrspace(10)* addrspace(11)* %8, i64 3
      %16 = bitcast %jl_value_t addrspace(10)* addrspace(11)* %15 to i64 addrspace(11)*
      %17 = load i64, i64 addrspace(11)* %16, align 8
      %18 = getelementptr inbounds %jl_value_t addrspace(10)*, %jl_value_t addrspace(10)* addrspace(11)* %8, i64 4
      %19 = bitcast %jl_value_t addrspace(10)* addrspace(11)* %18 to i64 addrspace(11)*
      %20 = load i64, i64 addrspace(11)* %19, align 8
      %21 = add i64 %20, -5
      %22 = icmp sgt i64 %21, 4
      %23 = select i1 %22, i64 %21, i64 4
      %24 = add i64 %17, -5
      %25 = icmp sgt i64 %24, 4
      %26 = select i1 %25, i64 %24, i64 4
      %27 = add nsw i64 %26, -5
; └└└└
;  @ /data/esdrd/tqff/tmp/issue.jl:24 within `does_not_vectorize!'
; ┌ @ simdloop.jl:71 within `macro expansion'
; │┌ @ simdloop.jl:51 within `simd_inner_length'
; ││┌ @ range.jl:543 within `length'
; │││┌ @ checked.jl:165 within `checked_add'
; ││││┌ @ checked.jl:132 within `add_with_overflow'
       %28 = call { i64, i1 } @llvm.sadd.with.overflow.i64(i64 %27, i64 1)
       %29 = extractvalue { i64, i1 } %28, 0
       %30 = extractvalue { i64, i1 } %28, 1
       %31 = icmp slt i64 %29, 1
       %32 = addrspacecast %jl_value_t addrspace(10)* %12 to %jl_value_t addrspace(11)*
       %33 = bitcast %jl_value_t addrspace(11)* %32 to float addrspace(11)*
       %34 = bitcast %jl_value_t addrspace(11)* %7 to float addrspace(13)* addrspace(11)*
       %35 = load float addrspace(13)*, float addrspace(13)* addrspace(11)* %34, align 8
       %36 = bitcast %jl_value_t addrspace(11)* %32 to [9 x float] addrspace(11)*
       %37 = getelementptr [9 x float], [9 x float] addrspace(11)* %36, i64 0, i64 1
       %38 = getelementptr [9 x float], [9 x float] addrspace(11)* %36, i64 0, i64 2
       %39 = getelementptr [9 x float], [9 x float] addrspace(11)* %36, i64 0, i64 3
       %40 = getelementptr [9 x float], [9 x float] addrspace(11)* %36, i64 0, i64 4
       %41 = getelementptr [9 x float], [9 x float] addrspace(11)* %36, i64 0, i64 5
       %42 = getelementptr [9 x float], [9 x float] addrspace(11)* %36, i64 0, i64 6
       %43 = getelementptr [9 x float], [9 x float] addrspace(11)* %36, i64 0, i64 7
       %44 = getelementptr [9 x float], [9 x float] addrspace(11)* %36, i64 0, i64 8
       %45 = addrspacecast %jl_value_t addrspace(10)* %4 to %jl_value_t addrspace(11)*
       %46 = bitcast %jl_value_t addrspace(11)* %45 to %jl_value_t addrspace(10)* addrspace(11)*
       %47 = getelementptr inbounds %jl_value_t addrspace(10)*, %jl_value_t addrspace(10)* addrspace(11)* %46, i64 3
       %48 = bitcast %jl_value_t addrspace(10)* addrspace(11)* %47 to i64 addrspace(11)*
       %49 = load i64, i64 addrspace(11)* %48, align 8
       %50 = getelementptr inbounds %jl_value_t addrspace(10)*, %jl_value_t addrspace(10)* addrspace(11)* %46, i64 4
       %51 = bitcast %jl_value_t addrspace(10)* addrspace(11)* %50 to i64 addrspace(11)*
       %52 = load i64, i64 addrspace(11)* %51, align 8
       %53 = bitcast %jl_value_t addrspace(11)* %45 to float addrspace(13)* addrspace(11)*
       %54 = load float addrspace(13)*, float addrspace(13)* addrspace(11)* %53, align 8
; └└└└└
;  @ /data/esdrd/tqff/tmp/issue.jl:23 within `does_not_vectorize!'
  br label %L18

L18:                                              ; preds = %L197, %L13.L18_crit_edge
  %value_phi3 = phi i64 [ 5, %L13.L18_crit_edge ], [ %174, %L197 ]
  br i1 %22, label %L27.L32_crit_edge, label %L197

L27.L32_crit_edge:                                ; preds = %L18
;  @ /data/esdrd/tqff/tmp/issue.jl:24 within `does_not_vectorize!'
; ┌ @ simdloop.jl:71 within `macro expansion'
; │┌ @ simdloop.jl:51 within `simd_inner_length'
; ││┌ @ range.jl:543 within `length'
; │││┌ @ checked.jl:166 within `checked_add'
      br i1 %30, label %L50, label %L55.lr.ph

L55.lr.ph:                                        ; preds = %L27.L32_crit_edge
      %55 = add i64 %value_phi3, -1
      %56 = mul i64 %20, %55
      %57 = mul i64 %52, %55
      br label %L55

L32:                                              ; preds = %L186
; └└└└
; ┌ @ range.jl:598 within `iterate'
; │┌ @ int.jl:53 within `+'
    %58 = add i64 %value_phi871, 1
; └└
; ┌ @ simdloop.jl:71 within `macro expansion'
; │┌ @ simdloop.jl:51 within `simd_inner_length'
; ││┌ @ range.jl:543 within `length'
; │││┌ @ checked.jl:166 within `checked_add'
      br label %L55

L50:                                              ; preds = %L27.L32_crit_edge
      call void @julia_throw_overflowerr_binaryop_13905(%jl_value_t addrspace(10)* addrspacecast (%jl_value_t* inttoptr (i64 47939032106784 to %jl_value_t*) to %jl_value_t addrspace(10)*), i64 %27, i64 1)
      call void @llvm.trap()
      unreachable

L55:                                              ; preds = %L32, %L55.lr.ph
      %value_phi871 = phi i64 [ 5, %L55.lr.ph ], [ %58, %L32 ]
; │└└└
; │ @ simdloop.jl:72 within `macro expansion'
   br i1 %31, label %L186, label %L61.lr.ph

L61.lr.ph:                                        ; preds = %L55
   %59 = load float, float addrspace(11)* %33, align 4
   %60 = add i64 %value_phi871, -1
   %reass.add = add i64 %60, %56
   %reass.mul = mul i64 %reass.add, %17
   %61 = load float, float addrspace(11)* %37, align 4
   %62 = load float, float addrspace(11)* %38, align 4
   %63 = load float, float addrspace(11)* %39, align 4
   %64 = load float, float addrspace(11)* %40, align 4
   %65 = load float, float addrspace(11)* %41, align 4
   %66 = load float, float addrspace(11)* %42, align 4
   %67 = load float, float addrspace(11)* %43, align 4
   %68 = load float, float addrspace(11)* %44, align 4
   %69 = add i64 %value_phi871, -5
   %reass.add49 = add i64 %69, %56
   %reass.mul50 = mul i64 %reass.add49, %17
   %70 = add i64 %value_phi871, -4
   %reass.add51 = add i64 %70, %56
   %reass.mul52 = mul i64 %reass.add51, %17
   %71 = add i64 %value_phi871, -3
   %reass.add53 = add i64 %71, %56
   %reass.mul54 = mul i64 %reass.add53, %17
   %72 = add i64 %value_phi871, -2
   %reass.add55 = add i64 %72, %56
   %reass.mul56 = mul i64 %reass.add55, %17
   %reass.add59 = add i64 %56, %value_phi871
   %reass.mul60 = mul i64 %reass.add59, %17
   %73 = add i64 %value_phi871, 1
   %reass.add61 = add i64 %73, %56
   %reass.mul62 = mul i64 %reass.add61, %17
   %74 = add i64 %value_phi871, 2
   %reass.add63 = add i64 %74, %56
   %reass.mul64 = mul i64 %reass.add63, %17
   %75 = add i64 %value_phi871, 3
   %reass.add65 = add i64 %75, %56
   %reass.mul66 = mul i64 %reass.add65, %17
   %reass.add67 = add i64 %60, %57
   %reass.mul68 = mul i64 %reass.add67, %49
; │ @ simdloop.jl:75 within `macro expansion'
   br label %L61

L61:                                              ; preds = %L61.lr.ph, %L61
   %value_phi1170 = phi i64 [ 0, %L61.lr.ph ], [ %80, %L61 ]
; │ @ simdloop.jl:77 within `macro expansion' @ /data/esdrd/tqff/tmp/issue.jl:25
; │┌ @ array.jl:743 within `getindex'
    %76 = add i64 %reass.mul, %value_phi1170
    %77 = getelementptr inbounds float, float addrspace(13)* %35, i64 %76
    %78 = load float, float addrspace(13)* %77, align 4
; │└
; │┌ @ float.jl:404 within `*'
    %79 = fmul float %59, %78
; │└
; │┌ @ array.jl:743 within `getindex'
    %80 = add nuw nsw i64 %value_phi1170, 1
    %81 = add i64 %80, %reass.mul
    %82 = getelementptr inbounds float, float addrspace(13)* %35, i64 %81
    %83 = load float, float addrspace(13)* %82, align 4
; │└
; │┌ @ float.jl:404 within `*'
    %84 = fmul float %61, %83
; │└
; │┌ @ operators.jl:529 within `+' @ float.jl:400
    %85 = fadd float %79, %84
; │└
; │┌ @ array.jl:743 within `getindex'
    %86 = add nuw nsw i64 %value_phi1170, 2
    %87 = add i64 %86, %reass.mul
    %88 = getelementptr inbounds float, float addrspace(13)* %35, i64 %87
    %89 = load float, float addrspace(13)* %88, align 4
; │└
; │┌ @ float.jl:404 within `*'
    %90 = fmul float %62, %89
; │└
; │┌ @ operators.jl:529 within `+' @ float.jl:400
    %91 = fadd float %85, %90
; │└
; │┌ @ array.jl:743 within `getindex'
    %92 = add nuw nsw i64 %value_phi1170, 3
    %93 = add i64 %92, %reass.mul
    %94 = getelementptr inbounds float, float addrspace(13)* %35, i64 %93
    %95 = load float, float addrspace(13)* %94, align 4
; │└
; │┌ @ float.jl:404 within `*'
    %96 = fmul float %63, %95
; │└
; │┌ @ operators.jl:529 within `+'
; ││┌ @ operators.jl:519 within `afoldl'
; │││┌ @ float.jl:400 within `+'
      %97 = fadd float %91, %96
; │└└└
; │┌ @ array.jl:743 within `getindex'
    %98 = add nuw nsw i64 %value_phi1170, 4
    %99 = add i64 %98, %reass.mul
    %100 = getelementptr inbounds float, float addrspace(13)* %35, i64 %99
    %101 = load float, float addrspace(13)* %100, align 4
; │└
; │┌ @ float.jl:404 within `*'
    %102 = fmul float %64, %101
; │└
; │┌ @ operators.jl:529 within `+'
; ││┌ @ operators.jl:519 within `afoldl'
; │││┌ @ float.jl:400 within `+'
      %103 = fadd float %97, %102
; │└└└
; │┌ @ array.jl:743 within `getindex'
    %104 = add nuw nsw i64 %value_phi1170, 5
    %105 = add i64 %104, %reass.mul
    %106 = getelementptr inbounds float, float addrspace(13)* %35, i64 %105
    %107 = load float, float addrspace(13)* %106, align 4
; │└
; │┌ @ float.jl:404 within `*'
    %108 = fmul float %65, %107
; │└
; │┌ @ operators.jl:529 within `+'
; ││┌ @ operators.jl:519 within `afoldl'
; │││┌ @ float.jl:400 within `+'
      %109 = fadd float %103, %108
; │└└└
; │┌ @ array.jl:743 within `getindex'
    %110 = add nuw i64 %value_phi1170, 6
    %111 = add i64 %110, %reass.mul
    %112 = getelementptr inbounds float, float addrspace(13)* %35, i64 %111
    %113 = load float, float addrspace(13)* %112, align 4
; │└
; │┌ @ float.jl:404 within `*'
    %114 = fmul float %66, %113
; │└
; │┌ @ operators.jl:529 within `+'
; ││┌ @ operators.jl:519 within `afoldl'
; │││┌ @ float.jl:400 within `+'
      %115 = fadd float %109, %114
; │└└└
; │┌ @ array.jl:743 within `getindex'
    %116 = add nuw i64 %value_phi1170, 7
    %117 = add i64 %116, %reass.mul
    %118 = getelementptr inbounds float, float addrspace(13)* %35, i64 %117
    %119 = load float, float addrspace(13)* %118, align 4
; │└
; │┌ @ float.jl:404 within `*'
    %120 = fmul float %67, %119
; │└
; │┌ @ operators.jl:529 within `+'
; ││┌ @ operators.jl:519 within `afoldl'
; │││┌ @ float.jl:400 within `+'
      %121 = fadd float %115, %120
; │└└└
; │┌ @ array.jl:743 within `getindex'
    %122 = add nuw i64 %value_phi1170, 8
    %123 = add i64 %122, %reass.mul
    %124 = getelementptr inbounds float, float addrspace(13)* %35, i64 %123
    %125 = load float, float addrspace(13)* %124, align 4
; │└
; │┌ @ float.jl:404 within `*'
    %126 = fmul float %68, %125
; │└
; │┌ @ operators.jl:529 within `+'
; ││┌ @ operators.jl:519 within `afoldl'
; │││┌ @ float.jl:400 within `+'
      %127 = fadd float %121, %126
; │└└└
; │┌ @ array.jl:743 within `getindex'
    %128 = add i64 %98, %reass.mul50
    %129 = getelementptr inbounds float, float addrspace(13)* %35, i64 %128
    %130 = load float, float addrspace(13)* %129, align 4
; │└
; │┌ @ float.jl:404 within `*'
    %131 = fmul float %59, %130
; │└
; │┌ @ operators.jl:529 within `+'
; ││┌ @ operators.jl:519 within `afoldl'
; │││┌ @ float.jl:400 within `+'
      %132 = fadd float %127, %131
; │└└└
; │┌ @ array.jl:743 within `getindex'
    %133 = add i64 %98, %reass.mul52
    %134 = getelementptr inbounds float, float addrspace(13)* %35, i64 %133
    %135 = load float, float addrspace(13)* %134, align 4
; │└
; │┌ @ float.jl:404 within `*'
    %136 = fmul float %61, %135
; │└
; │┌ @ operators.jl:529 within `+'
; ││┌ @ operators.jl:519 within `afoldl'
; │││┌ @ float.jl:400 within `+'
      %137 = fadd float %132, %136
; │└└└
; │┌ @ array.jl:743 within `getindex'
    %138 = add i64 %98, %reass.mul54
    %139 = getelementptr inbounds float, float addrspace(13)* %35, i64 %138
    %140 = load float, float addrspace(13)* %139, align 4
; │└
; │┌ @ float.jl:404 within `*'
    %141 = fmul float %62, %140
; │└
; │┌ @ operators.jl:529 within `+'
; ││┌ @ operators.jl:519 within `afoldl'
; │││┌ @ float.jl:400 within `+'
      %142 = fadd float %137, %141
; │└└└
; │┌ @ array.jl:743 within `getindex'
    %143 = add i64 %98, %reass.mul56
    %144 = getelementptr inbounds float, float addrspace(13)* %35, i64 %143
    %145 = load float, float addrspace(13)* %144, align 4
; │└
; │┌ @ float.jl:404 within `*'
    %146 = fmul float %63, %145
; │└
; │┌ @ operators.jl:529 within `+'
; ││┌ @ operators.jl:519 within `afoldl'
; │││┌ @ float.jl:400 within `+'
      %147 = fadd float %142, %146
      %148 = fadd float %147, %102
; │└└└
; │┌ @ array.jl:743 within `getindex'
    %149 = add i64 %98, %reass.mul60
    %150 = getelementptr inbounds float, float addrspace(13)* %35, i64 %149
    %151 = load float, float addrspace(13)* %150, align 4
; │└
; │┌ @ float.jl:404 within `*'
    %152 = fmul float %65, %151
; │└
; │┌ @ operators.jl:529 within `+'
; ││┌ @ operators.jl:519 within `afoldl'
; │││┌ @ float.jl:400 within `+'
      %153 = fadd float %148, %152
; │└└└
; │┌ @ array.jl:743 within `getindex'
    %154 = add i64 %98, %reass.mul62
    %155 = getelementptr inbounds float, float addrspace(13)* %35, i64 %154
    %156 = load float, float addrspace(13)* %155, align 4
; │└
; │┌ @ float.jl:404 within `*'
    %157 = fmul float %66, %156
; │└
; │┌ @ operators.jl:529 within `+'
; ││┌ @ operators.jl:519 within `afoldl'
; │││┌ @ float.jl:400 within `+'
      %158 = fadd float %153, %157
; │└└└
; │┌ @ array.jl:743 within `getindex'
    %159 = add i64 %98, %reass.mul64
    %160 = getelementptr inbounds float, float addrspace(13)* %35, i64 %159
    %161 = load float, float addrspace(13)* %160, align 4
; │└
; │┌ @ float.jl:404 within `*'
    %162 = fmul float %67, %161
; │└
; │┌ @ operators.jl:529 within `+'
; ││┌ @ operators.jl:519 within `afoldl'
; │││┌ @ float.jl:400 within `+'
      %163 = fadd float %158, %162
; │└└└
; │┌ @ array.jl:743 within `getindex'
    %164 = add i64 %98, %reass.mul66
    %165 = getelementptr inbounds float, float addrspace(13)* %35, i64 %164
    %166 = load float, float addrspace(13)* %165, align 4
; │└
; │┌ @ float.jl:404 within `*'
    %167 = fmul float %68, %166
; │└
; │┌ @ operators.jl:529 within `+'
; ││┌ @ operators.jl:519 within `afoldl'
; │││┌ @ float.jl:400 within `+'
      %168 = fadd float %163, %167
; │└└└
; │┌ @ array.jl:782 within `setindex!'
    %169 = add i64 %98, %reass.mul68
    %170 = getelementptr inbounds float, float addrspace(13)* %54, i64 %169
    store float %168, float addrspace(13)* %170, align 4
; │└
; │ @ simdloop.jl:75 within `macro expansion'
; │┌ @ int.jl:49 within `<'
    %171 = icmp ult i64 %80, %29
; │└
   br i1 %171, label %L61, label %L186

L186:                                             ; preds = %L61, %L55
; └
; ┌ @ range.jl:597 within `iterate'
; │┌ @ promotion.jl:403 within `=='
    %172 = icmp eq i64 %value_phi871, %23
; └└
  br i1 %172, label %L197, label %L32

L197:                                             ; preds = %L186, %L18
; ┌ @ range.jl:597 within `iterate'
; │┌ @ promotion.jl:403 within `=='
    %173 = icmp eq i64 %value_phi3, %13
; │└
; │ @ range.jl:598 within `iterate'
; │┌ @ int.jl:53 within `+'
    %174 = add i64 %value_phi3, 1
; └└
  br i1 %173, label %L208, label %L18

L208:                                             ; preds = %L197, %top
;  @ /data/esdrd/tqff/tmp/issue.jl:31 within `does_not_vectorize!'
  ret %jl_value_t addrspace(10)* addrspacecast (%jl_value_t* inttoptr (i64 47938844788480 to %jl_value_t*) to %jl_value_t addrspace(10)*)
}

does_vectorize.txt

;  @ /data/esdrd/tqff/tmp/issue.jl:2 within `does_vectorize!'
define nonnull %jl_value_t addrspace(10)* @"japi1_does_vectorize!_16046"(%jl_value_t addrspace(10)*, %jl_value_t addrspace(10)**, i32) #0 {
top:
  %3 = alloca %jl_value_t addrspace(10)**, align 8
  store volatile %jl_value_t addrspace(10)** %1, %jl_value_t addrspace(10)*** %3, align 8
  %4 = load %jl_value_t addrspace(10)*, %jl_value_t addrspace(10)** %1, align 8
  %5 = getelementptr inbounds %jl_value_t addrspace(10)*, %jl_value_t addrspace(10)** %1, i64 1
  %6 = load %jl_value_t addrspace(10)*, %jl_value_t addrspace(10)** %5, align 8
; ┌ @ array.jl:157 within `size'
; │┌ @ ntuple.jl:43 within `ntuple'
; ││┌ @ array.jl:157 within `#69'
; │││┌ @ array.jl:154 within `size'
      %7 = addrspacecast %jl_value_t addrspace(10)* %6 to %jl_value_t addrspace(11)*
      %8 = bitcast %jl_value_t addrspace(11)* %7 to %jl_value_t addrspace(10)* addrspace(11)*
      %9 = getelementptr inbounds %jl_value_t addrspace(10)*, %jl_value_t addrspace(10)* addrspace(11)* %8, i64 5
      %10 = bitcast %jl_value_t addrspace(10)* addrspace(11)* %9 to i64 addrspace(11)*
      %11 = load i64, i64 addrspace(11)* %10, align 8
; └└└└
;  @ /data/esdrd/tqff/tmp/issue.jl:3 within `does_vectorize!'
  %12 = call nonnull %jl_value_t addrspace(10)* @julia_ntuple_16047(i64 9)
;  @ /data/esdrd/tqff/tmp/issue.jl:5 within `does_vectorize!'
; ┌ @ int.jl:52 within `-'
   %13 = add i64 %11, -5
; └
; ┌ @ range.jl:5 within `Colon'
; │┌ @ range.jl:277 within `UnitRange'
; ││┌ @ range.jl:282 within `unitrange_last'
; │││┌ @ operators.jl:341 within `>='
; ││││┌ @ int.jl:424 within `<='
       %14 = icmp sgt i64 %13, 4
; └└└└└
  br i1 %14, label %L13.L18_crit_edge, label %L189

L13.L18_crit_edge:                                ; preds = %top
;  @ /data/esdrd/tqff/tmp/issue.jl:2 within `does_vectorize!'
; ┌ @ array.jl:157 within `size'
; │┌ @ ntuple.jl:43 within `ntuple'
; ││┌ @ array.jl:157 within `#69'
; │││┌ @ array.jl:154 within `size'
      %15 = getelementptr inbounds %jl_value_t addrspace(10)*, %jl_value_t addrspace(10)* addrspace(11)* %8, i64 3
      %16 = bitcast %jl_value_t addrspace(10)* addrspace(11)* %15 to i64 addrspace(11)*
      %17 = load i64, i64 addrspace(11)* %16, align 8
      %18 = getelementptr inbounds %jl_value_t addrspace(10)*, %jl_value_t addrspace(10)* addrspace(11)* %8, i64 4
      %19 = bitcast %jl_value_t addrspace(10)* addrspace(11)* %18 to i64 addrspace(11)*
      %20 = load i64, i64 addrspace(11)* %19, align 8
      %21 = add i64 %20, -5
      %22 = icmp sgt i64 %21, 4
      %23 = select i1 %22, i64 %21, i64 4
      %24 = add i64 %17, -5
      %25 = icmp sgt i64 %24, 4
      %26 = select i1 %25, i64 %24, i64 4
      %27 = add nsw i64 %26, -5
; └└└└
;  @ /data/esdrd/tqff/tmp/issue.jl:6 within `does_vectorize!'
; ┌ @ simdloop.jl:71 within `macro expansion'
; │┌ @ simdloop.jl:51 within `simd_inner_length'
; ││┌ @ range.jl:543 within `length'
; │││┌ @ checked.jl:165 within `checked_add'
; ││││┌ @ checked.jl:132 within `add_with_overflow'
       %28 = call { i64, i1 } @llvm.sadd.with.overflow.i64(i64 %27, i64 1)
       %29 = extractvalue { i64, i1 } %28, 0
       %30 = extractvalue { i64, i1 } %28, 1
       %31 = icmp slt i64 %29, 1
       %32 = addrspacecast %jl_value_t addrspace(10)* %12 to %jl_value_t addrspace(11)*
       %33 = bitcast %jl_value_t addrspace(11)* %32 to float addrspace(11)*
       %34 = bitcast %jl_value_t addrspace(11)* %7 to float addrspace(13)* addrspace(11)*
       %35 = load float addrspace(13)*, float addrspace(13)* addrspace(11)* %34, align 8
       %36 = bitcast %jl_value_t addrspace(11)* %32 to [9 x float] addrspace(11)*
       %37 = getelementptr [9 x float], [9 x float] addrspace(11)* %36, i64 0, i64 1
       %38 = getelementptr [9 x float], [9 x float] addrspace(11)* %36, i64 0, i64 2
       %39 = getelementptr [9 x float], [9 x float] addrspace(11)* %36, i64 0, i64 3
       %40 = getelementptr [9 x float], [9 x float] addrspace(11)* %36, i64 0, i64 4
       %41 = getelementptr [9 x float], [9 x float] addrspace(11)* %36, i64 0, i64 5
       %42 = getelementptr [9 x float], [9 x float] addrspace(11)* %36, i64 0, i64 6
       %43 = getelementptr [9 x float], [9 x float] addrspace(11)* %36, i64 0, i64 7
       %44 = getelementptr [9 x float], [9 x float] addrspace(11)* %36, i64 0, i64 8
       %45 = addrspacecast %jl_value_t addrspace(10)* %4 to %jl_value_t addrspace(11)*
       %46 = bitcast %jl_value_t addrspace(11)* %45 to %jl_value_t addrspace(10)* addrspace(11)*
       %47 = getelementptr inbounds %jl_value_t addrspace(10)*, %jl_value_t addrspace(10)* addrspace(11)* %46, i64 3
       %48 = bitcast %jl_value_t addrspace(10)* addrspace(11)* %47 to i64 addrspace(11)*
       %49 = load i64, i64 addrspace(11)* %48, align 8
       %50 = getelementptr inbounds %jl_value_t addrspace(10)*, %jl_value_t addrspace(10)* addrspace(11)* %46, i64 4
       %51 = bitcast %jl_value_t addrspace(10)* addrspace(11)* %50 to i64 addrspace(11)*
       %52 = load i64, i64 addrspace(11)* %51, align 8
       %53 = bitcast %jl_value_t addrspace(11)* %45 to float addrspace(13)* addrspace(11)*
       %54 = load float addrspace(13)*, float addrspace(13)* addrspace(11)* %53, align 8
; └└└└└
;  @ /data/esdrd/tqff/tmp/issue.jl:5 within `does_vectorize!'
  %55 = shl i64 %52, 2
  %56 = add i64 %55, 4
  %57 = mul i64 %49, %56
  %58 = add i64 %57, 4
  %59 = mul i64 %52, %49
  %60 = add i64 %26, %57
  %61 = shl i64 %20, 2
  %62 = add i64 %61, 7
  %63 = mul i64 %17, %62
  %64 = add i64 %63, 4
  %65 = mul i64 %20, %17
  %66 = add i64 %26, %63
  %67 = add i64 %61, 6
  %68 = mul i64 %17, %67
  %69 = add i64 %68, 4
  %70 = add i64 %26, %68
  %71 = add i64 %61, 5
  %72 = mul i64 %17, %71
  %73 = add i64 %72, 4
  %74 = add i64 %26, %72
  %75 = or i64 %61, 3
  %76 = mul i64 %17, %75
  %77 = add i64 %76, 4
  %78 = add i64 %26, %76
  %79 = or i64 %61, 2
  %80 = mul i64 %17, %79
  %81 = add i64 %80, 4
  %82 = add i64 %26, %80
  %83 = or i64 %61, 1
  %84 = mul i64 %17, %83
  %85 = add i64 %84, 4
  %86 = add i64 %26, %84
  %87 = mul i64 %20, %17
  %88 = shl i64 %87, 2
  %89 = add i64 %88, 4
  %90 = add i64 %26, %88
  %91 = add i64 %61, 4
  %92 = mul i64 %17, %91
  %93 = add i64 %26, %92
  %94 = add i64 %93, 4
  br label %L18

L18:                                              ; preds = %L178, %L13.L18_crit_edge
  %indvar = phi i64 [ %indvar.next, %L178 ], [ 0, %L13.L18_crit_edge ]
  %value_phi3 = phi i64 [ %340, %L178 ], [ 5, %L13.L18_crit_edge ]
  %95 = mul i64 %59, %indvar
  %96 = add i64 %58, %95
  %97 = add i64 %60, %95
  %98 = mul i64 %65, %indvar
  %99 = add i64 %64, %98
  %100 = add i64 %66, %98
  %101 = add i64 %69, %98
  %102 = add i64 %70, %98
  %103 = add i64 %73, %98
  %104 = add i64 %74, %98
  %105 = add i64 %77, %98
  %106 = add i64 %78, %98
  %107 = add i64 %81, %98
  %108 = add i64 %82, %98
  %109 = add i64 %85, %98
  %110 = add i64 %86, %98
  %111 = add i64 %89, %98
  %112 = add i64 %90, %98
  %113 = add i64 %92, %98
  %114 = add i64 %94, %98
  br i1 %22, label %L27.L32_crit_edge, label %L178

L27.L32_crit_edge:                                ; preds = %L18
;  @ /data/esdrd/tqff/tmp/issue.jl:6 within `does_vectorize!'
; ┌ @ simdloop.jl:71 within `macro expansion'
; │┌ @ simdloop.jl:51 within `simd_inner_length'
; ││┌ @ range.jl:543 within `length'
; │││┌ @ checked.jl:166 within `checked_add'
      br i1 %30, label %L50, label %L55.lr.ph

L55.lr.ph:                                        ; preds = %L27.L32_crit_edge
      %115 = add i64 %value_phi3, -1
      %116 = mul i64 %20, %115
      %117 = mul i64 %52, %115
      br label %L55

L32:                                              ; preds = %L167
; └└└└
; ┌ @ range.jl:598 within `iterate'
; │┌ @ int.jl:53 within `+'
    %118 = add i64 %value_phi867, 1
; └└
; ┌ @ simdloop.jl:71 within `macro expansion'
; │┌ @ simdloop.jl:51 within `simd_inner_length'
; ││┌ @ range.jl:543 within `length'
; │││┌ @ checked.jl:166 within `checked_add'
      %indvar.next72 = add i64 %indvar71, 1
      br label %L55

L50:                                              ; preds = %L27.L32_crit_edge
      call void @julia_throw_overflowerr_binaryop_13905(%jl_value_t addrspace(10)* addrspacecast (%jl_value_t* inttoptr (i64 47939032106784 to %jl_value_t*) to %jl_value_t addrspace(10)*), i64 %27, i64 1)
      call void @llvm.trap()
      unreachable

L55:                                              ; preds = %L32, %L55.lr.ph
      %indvar71 = phi i64 [ %indvar.next72, %L32 ], [ 0, %L55.lr.ph ]
      %value_phi867 = phi i64 [ %118, %L32 ], [ 5, %L55.lr.ph ]
; │└└└
; │ @ simdloop.jl:72 within `macro expansion'
   %119 = mul i64 %49, %indvar71
   %120 = add i64 %96, %119
   %scevgep = getelementptr float, float addrspace(13)* %54, i64 %120
   %121 = add i64 %97, %119
   %scevgep74 = getelementptr float, float addrspace(13)* %54, i64 %121
   %122 = mul i64 %17, %indvar71
   %123 = add i64 %99, %122
   %scevgep76 = getelementptr float, float addrspace(13)* %35, i64 %123
   %124 = add i64 %100, %122
   %scevgep78 = getelementptr float, float addrspace(13)* %35, i64 %124
   %125 = add i64 %101, %122
   %scevgep80 = getelementptr float, float addrspace(13)* %35, i64 %125
   %126 = add i64 %102, %122
   %scevgep82 = getelementptr float, float addrspace(13)* %35, i64 %126
   %127 = add i64 %103, %122
   %scevgep84 = getelementptr float, float addrspace(13)* %35, i64 %127
   %128 = add i64 %104, %122
   %scevgep86 = getelementptr float, float addrspace(13)* %35, i64 %128
   %129 = add i64 %105, %122
   %scevgep88 = getelementptr float, float addrspace(13)* %35, i64 %129
   %130 = add i64 %106, %122
   %scevgep90 = getelementptr float, float addrspace(13)* %35, i64 %130
   %131 = add i64 %107, %122
   %scevgep92 = getelementptr float, float addrspace(13)* %35, i64 %131
   %132 = add i64 %108, %122
   %scevgep94 = getelementptr float, float addrspace(13)* %35, i64 %132
   %133 = add i64 %109, %122
   %scevgep96 = getelementptr float, float addrspace(13)* %35, i64 %133
   %134 = add i64 %110, %122
   %scevgep98 = getelementptr float, float addrspace(13)* %35, i64 %134
   %135 = add i64 %111, %122
   %scevgep100 = getelementptr float, float addrspace(13)* %35, i64 %135
   %136 = add i64 %112, %122
   %scevgep102 = getelementptr float, float addrspace(13)* %35, i64 %136
   %137 = add i64 %113, %122
   %scevgep104 = getelementptr float, float addrspace(13)* %35, i64 %137
   %138 = add i64 %114, %122
   %scevgep106 = getelementptr float, float addrspace(13)* %35, i64 %138
   br i1 %31, label %L167, label %L61.lr.ph

L61.lr.ph:                                        ; preds = %L55
   %139 = add i64 %value_phi867, -1
   %reass.add = add i64 %139, %116
   %reass.mul = mul i64 %reass.add, %17
   %140 = load float, float addrspace(11)* %33, align 4
   %141 = load float, float addrspace(11)* %37, align 4
   %142 = load float, float addrspace(11)* %38, align 4
   %143 = load float, float addrspace(11)* %39, align 4
   %144 = load float, float addrspace(11)* %40, align 4
   %145 = load float, float addrspace(11)* %41, align 4
   %146 = load float, float addrspace(11)* %42, align 4
   %147 = load float, float addrspace(11)* %43, align 4
   %148 = load float, float addrspace(11)* %44, align 4
   %149 = add i64 %value_phi867, -5
   %reass.add47 = add i64 %149, %116
   %reass.mul48 = mul i64 %reass.add47, %17
   %150 = add i64 %value_phi867, -4
   %reass.add49 = add i64 %150, %116
   %reass.mul50 = mul i64 %reass.add49, %17
   %151 = add i64 %value_phi867, -3
   %reass.add51 = add i64 %151, %116
   %reass.mul52 = mul i64 %reass.add51, %17
   %152 = add i64 %value_phi867, -2
   %reass.add53 = add i64 %152, %116
   %reass.mul54 = mul i64 %reass.add53, %17
   %reass.add57 = add i64 %116, %value_phi867
   %reass.mul58 = mul i64 %reass.add57, %17
   %153 = add i64 %value_phi867, 1
   %reass.add59 = add i64 %153, %116
   %reass.mul60 = mul i64 %reass.add59, %17
   %154 = add i64 %value_phi867, 2
   %reass.add61 = add i64 %154, %116
   %reass.mul62 = mul i64 %reass.add61, %17
   %reass.add63 = add i64 %139, %117
   %reass.mul64 = mul i64 %reass.add63, %49
; │ @ simdloop.jl:75 within `macro expansion'
   %min.iters.check = icmp ult i64 %29, 8
   br i1 %min.iters.check, label %scalar.ph, label %vector.memcheck

vector.memcheck:                                  ; preds = %L61.lr.ph
   %bound0 = icmp ult float addrspace(13)* %scevgep, %scevgep78
   %bound1 = icmp ult float addrspace(13)* %scevgep76, %scevgep74
   %found.conflict = and i1 %bound0, %bound1
   %bound0108 = icmp ult float addrspace(13)* %scevgep, %scevgep82
   %bound1109 = icmp ult float addrspace(13)* %scevgep80, %scevgep74
   %found.conflict110 = and i1 %bound0108, %bound1109
   %conflict.rdx = or i1 %found.conflict, %found.conflict110
   %bound0111 = icmp ult float addrspace(13)* %scevgep, %scevgep86
   %bound1112 = icmp ult float addrspace(13)* %scevgep84, %scevgep74
   %found.conflict113 = and i1 %bound0111, %bound1112
   %conflict.rdx114 = or i1 %conflict.rdx, %found.conflict113
   %bound0115 = icmp ult float addrspace(13)* %scevgep, %scevgep90
   %bound1116 = icmp ult float addrspace(13)* %scevgep88, %scevgep74
   %found.conflict117 = and i1 %bound0115, %bound1116
   %conflict.rdx118 = or i1 %conflict.rdx114, %found.conflict117
   %bound0119 = icmp ult float addrspace(13)* %scevgep, %scevgep94
   %bound1120 = icmp ult float addrspace(13)* %scevgep92, %scevgep74
   %found.conflict121 = and i1 %bound0119, %bound1120
   %conflict.rdx122 = or i1 %conflict.rdx118, %found.conflict121
   %bound0123 = icmp ult float addrspace(13)* %scevgep, %scevgep98
   %bound1124 = icmp ult float addrspace(13)* %scevgep96, %scevgep74
   %found.conflict125 = and i1 %bound0123, %bound1124
   %conflict.rdx126 = or i1 %conflict.rdx122, %found.conflict125
   %bound0127 = icmp ult float addrspace(13)* %scevgep, %scevgep102
   %bound1128 = icmp ult float addrspace(13)* %scevgep100, %scevgep74
   %found.conflict129 = and i1 %bound0127, %bound1128
   %conflict.rdx130 = or i1 %conflict.rdx126, %found.conflict129
   %bound0131 = icmp ult float addrspace(13)* %scevgep, %scevgep106
   %bound1132 = icmp ult float addrspace(13)* %scevgep104, %scevgep74
   %found.conflict133 = and i1 %bound0131, %bound1132
   %conflict.rdx134 = or i1 %conflict.rdx130, %found.conflict133
   br i1 %conflict.rdx134, label %scalar.ph, label %vector.ph

vector.ph:                                        ; preds = %vector.memcheck
   %n.vec = and i64 %29, -8
   %broadcast.splatinsert135 = insertelement <8 x float> undef, float %140, i32 0
   %broadcast.splat136 = shufflevector <8 x float> %broadcast.splatinsert135, <8 x float> undef, <8 x i32> zeroinitializer
   %broadcast.splatinsert138 = insertelement <8 x float> undef, float %141, i32 0
   %broadcast.splat139 = shufflevector <8 x float> %broadcast.splatinsert138, <8 x float> undef, <8 x i32> zeroinitializer
   %broadcast.splatinsert141 = insertelement <8 x float> undef, float %142, i32 0
   %broadcast.splat142 = shufflevector <8 x float> %broadcast.splatinsert141, <8 x float> undef, <8 x i32> zeroinitializer
   %broadcast.splatinsert144 = insertelement <8 x float> undef, float %143, i32 0
   %broadcast.splat145 = shufflevector <8 x float> %broadcast.splatinsert144, <8 x float> undef, <8 x i32> zeroinitializer
   %broadcast.splatinsert147 = insertelement <8 x float> undef, float %144, i32 0
   %broadcast.splat148 = shufflevector <8 x float> %broadcast.splatinsert147, <8 x float> undef, <8 x i32> zeroinitializer
   %broadcast.splatinsert150 = insertelement <8 x float> undef, float %145, i32 0
   %broadcast.splat151 = shufflevector <8 x float> %broadcast.splatinsert150, <8 x float> undef, <8 x i32> zeroinitializer
   %broadcast.splatinsert153 = insertelement <8 x float> undef, float %146, i32 0
   %broadcast.splat154 = shufflevector <8 x float> %broadcast.splatinsert153, <8 x float> undef, <8 x i32> zeroinitializer
   %broadcast.splatinsert156 = insertelement <8 x float> undef, float %147, i32 0
   %broadcast.splat157 = shufflevector <8 x float> %broadcast.splatinsert156, <8 x float> undef, <8 x i32> zeroinitializer
   %broadcast.splatinsert159 = insertelement <8 x float> undef, float %148, i32 0
   %broadcast.splat160 = shufflevector <8 x float> %broadcast.splatinsert159, <8 x float> undef, <8 x i32> zeroinitializer
   br label %vector.body

vector.body:                                      ; preds = %vector.body, %vector.ph
; │ @ simdloop.jl:77 within `macro expansion' @ /data/esdrd/tqff/tmp/issue.jl:7
; │┌ @ array.jl:743 within `getindex'
    %index = phi i64 [ 0, %vector.ph ], [ %index.next, %vector.body ]
    %155 = add i64 %reass.mul, %index
    %156 = getelementptr inbounds float, float addrspace(13)* %35, i64 %155
    %157 = bitcast float addrspace(13)* %156 to <8 x float> addrspace(13)*
    %wide.load = load <8 x float>, <8 x float> addrspace(13)* %157, align 4
; │└
; │┌ @ float.jl:404 within `*'
    %158 = fmul <8 x float> %wide.load, %broadcast.splat136
; │└
; │┌ @ array.jl:743 within `getindex'
    %159 = or i64 %index, 1
    %160 = add i64 %159, %reass.mul
    %161 = getelementptr inbounds float, float addrspace(13)* %35, i64 %160
    %162 = bitcast float addrspace(13)* %161 to <8 x float> addrspace(13)*
    %wide.load137 = load <8 x float>, <8 x float> addrspace(13)* %162, align 4
; │└
; │┌ @ float.jl:404 within `*'
    %163 = fmul <8 x float> %wide.load137, %broadcast.splat139
; │└
; │┌ @ array.jl:743 within `getindex'
    %164 = or i64 %index, 2
    %165 = add i64 %164, %reass.mul
    %166 = getelementptr inbounds float, float addrspace(13)* %35, i64 %165
    %167 = bitcast float addrspace(13)* %166 to <8 x float> addrspace(13)*
    %wide.load140 = load <8 x float>, <8 x float> addrspace(13)* %167, align 4
; │└
; │┌ @ float.jl:404 within `*'
    %168 = fmul <8 x float> %wide.load140, %broadcast.splat142
; │└
; │┌ @ array.jl:743 within `getindex'
    %169 = or i64 %index, 3
    %170 = add i64 %169, %reass.mul
    %171 = getelementptr inbounds float, float addrspace(13)* %35, i64 %170
    %172 = bitcast float addrspace(13)* %171 to <8 x float> addrspace(13)*
    %wide.load143 = load <8 x float>, <8 x float> addrspace(13)* %172, align 4
; │└
; │┌ @ float.jl:404 within `*'
    %173 = fmul <8 x float> %wide.load143, %broadcast.splat145
; │└
; │┌ @ array.jl:743 within `getindex'
    %174 = or i64 %index, 4
    %175 = add i64 %174, %reass.mul
    %176 = getelementptr inbounds float, float addrspace(13)* %35, i64 %175
    %177 = bitcast float addrspace(13)* %176 to <8 x float> addrspace(13)*
    %wide.load146 = load <8 x float>, <8 x float> addrspace(13)* %177, align 4
; │└
; │┌ @ float.jl:404 within `*'
    %178 = fmul <8 x float> %wide.load146, %broadcast.splat148
; │└
; │┌ @ array.jl:743 within `getindex'
    %179 = or i64 %index, 5
    %180 = add i64 %179, %reass.mul
    %181 = getelementptr inbounds float, float addrspace(13)* %35, i64 %180
    %182 = bitcast float addrspace(13)* %181 to <8 x float> addrspace(13)*
    %wide.load149 = load <8 x float>, <8 x float> addrspace(13)* %182, align 4
; │└
; │┌ @ float.jl:404 within `*'
    %183 = fmul <8 x float> %wide.load149, %broadcast.splat151
; │└
; │┌ @ array.jl:743 within `getindex'
    %184 = or i64 %index, 6
    %185 = add i64 %184, %reass.mul
    %186 = getelementptr inbounds float, float addrspace(13)* %35, i64 %185
    %187 = bitcast float addrspace(13)* %186 to <8 x float> addrspace(13)*
    %wide.load152 = load <8 x float>, <8 x float> addrspace(13)* %187, align 4
; │└
; │┌ @ float.jl:404 within `*'
    %188 = fmul <8 x float> %wide.load152, %broadcast.splat154
; │└
; │┌ @ array.jl:743 within `getindex'
    %189 = or i64 %index, 7
    %190 = add i64 %189, %reass.mul
    %191 = getelementptr inbounds float, float addrspace(13)* %35, i64 %190
    %192 = bitcast float addrspace(13)* %191 to <8 x float> addrspace(13)*
    %wide.load155 = load <8 x float>, <8 x float> addrspace(13)* %192, align 4
; │└
; │┌ @ float.jl:404 within `*'
    %193 = fmul <8 x float> %wide.load155, %broadcast.splat157
; │└
; │┌ @ array.jl:743 within `getindex'
    %194 = add nuw i64 %index, 8
    %195 = add i64 %194, %reass.mul
    %196 = getelementptr inbounds float, float addrspace(13)* %35, i64 %195
    %197 = bitcast float addrspace(13)* %196 to <8 x float> addrspace(13)*
    %wide.load158 = load <8 x float>, <8 x float> addrspace(13)* %197, align 4
; │└
; │┌ @ float.jl:404 within `*'
    %198 = fmul <8 x float> %wide.load158, %broadcast.splat160
; │└
; │┌ @ array.jl:743 within `getindex'
    %199 = add i64 %174, %reass.mul48
    %200 = getelementptr inbounds float, float addrspace(13)* %35, i64 %199
    %201 = bitcast float addrspace(13)* %200 to <8 x float> addrspace(13)*
    %wide.load161 = load <8 x float>, <8 x float> addrspace(13)* %201, align 4
; │└
; │┌ @ float.jl:404 within `*'
    %202 = fmul <8 x float> %wide.load161, %broadcast.splat136
; │└
; │┌ @ array.jl:743 within `getindex'
    %203 = add i64 %174, %reass.mul50
    %204 = getelementptr inbounds float, float addrspace(13)* %35, i64 %203
    %205 = bitcast float addrspace(13)* %204 to <8 x float> addrspace(13)*
    %wide.load162 = load <8 x float>, <8 x float> addrspace(13)* %205, align 4
; │└
; │┌ @ float.jl:404 within `*'
    %206 = fmul <8 x float> %wide.load162, %broadcast.splat139
; │└
; │┌ @ array.jl:743 within `getindex'
    %207 = add i64 %174, %reass.mul52
    %208 = getelementptr inbounds float, float addrspace(13)* %35, i64 %207
    %209 = bitcast float addrspace(13)* %208 to <8 x float> addrspace(13)*
    %wide.load163 = load <8 x float>, <8 x float> addrspace(13)* %209, align 4
; │└
; │┌ @ float.jl:404 within `*'
    %210 = fmul <8 x float> %wide.load163, %broadcast.splat142
; │└
; │┌ @ array.jl:743 within `getindex'
    %211 = add i64 %174, %reass.mul54
    %212 = getelementptr inbounds float, float addrspace(13)* %35, i64 %211
    %213 = bitcast float addrspace(13)* %212 to <8 x float> addrspace(13)*
    %wide.load164 = load <8 x float>, <8 x float> addrspace(13)* %213, align 4
; │└
; │┌ @ float.jl:404 within `*'
    %214 = fmul <8 x float> %wide.load164, %broadcast.splat145
; │└
; │┌ @ array.jl:743 within `getindex'
    %215 = add i64 %174, %reass.mul58
    %216 = getelementptr inbounds float, float addrspace(13)* %35, i64 %215
    %217 = bitcast float addrspace(13)* %216 to <8 x float> addrspace(13)*
    %wide.load165 = load <8 x float>, <8 x float> addrspace(13)* %217, align 4
; │└
; │┌ @ float.jl:404 within `*'
    %218 = fmul <8 x float> %wide.load165, %broadcast.splat151
; │└
; │┌ @ array.jl:743 within `getindex'
    %219 = add i64 %174, %reass.mul60
    %220 = getelementptr inbounds float, float addrspace(13)* %35, i64 %219
    %221 = bitcast float addrspace(13)* %220 to <8 x float> addrspace(13)*
    %wide.load166 = load <8 x float>, <8 x float> addrspace(13)* %221, align 4
; │└
; │┌ @ float.jl:404 within `*'
    %222 = fmul <8 x float> %wide.load166, %broadcast.splat154
; │└
; │┌ @ array.jl:743 within `getindex'
    %223 = add i64 %174, %reass.mul62
    %224 = getelementptr inbounds float, float addrspace(13)* %35, i64 %223
    %225 = bitcast float addrspace(13)* %224 to <8 x float> addrspace(13)*
    %wide.load167 = load <8 x float>, <8 x float> addrspace(13)* %225, align 4
; │└
; │┌ @ float.jl:404 within `*'
    %226 = fmul <8 x float> %wide.load167, %broadcast.splat157
; │└
; │┌ @ operators.jl:529 within `+' @ float.jl:400
    %227 = fadd <8 x float> %158, %163
    %228 = fadd <8 x float> %227, %168
; ││ @ operators.jl:529 within `+'
; ││┌ @ operators.jl:517 within `afoldl'
; │││┌ @ float.jl:400 within `+'
      %229 = fadd <8 x float> %228, %173
; │││└
; │││ @ operators.jl:517 within `afoldl' @ operators.jl:517
; │││┌ @ float.jl:400 within `+'
      %230 = fadd <8 x float> %229, %178
; │││└
; │││ @ operators.jl:517 within `afoldl' @ operators.jl:517 @ operators.jl:517
; │││┌ @ float.jl:400 within `+'
      %231 = fadd <8 x float> %230, %183
; │││└
; │││ @ operators.jl:517 within `afoldl' @ operators.jl:517 @ operators.jl:517 @ operators.jl:517
; │││┌ @ float.jl:400 within `+'
      %232 = fadd <8 x float> %231, %188
; │││└
; │││ @ operators.jl:517 within `afoldl' @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517
; │││┌ @ float.jl:400 within `+'
      %233 = fadd <8 x float> %232, %193
; │││└
; │││ @ operators.jl:517 within `afoldl' @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517
; │││┌ @ float.jl:400 within `+'
      %234 = fadd <8 x float> %233, %198
; │││└
; │││ @ operators.jl:517 within `afoldl' @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517
; │││┌ @ float.jl:400 within `+'
      %235 = fadd <8 x float> %234, %202
; │││└
; │││ @ operators.jl:517 within `afoldl' @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517
; │││┌ @ float.jl:400 within `+'
      %236 = fadd <8 x float> %235, %206
; │││└
; │││ @ operators.jl:517 within `afoldl' @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517
; │││┌ @ float.jl:400 within `+'
      %237 = fadd <8 x float> %236, %210
; │││└
; │││ @ operators.jl:517 within `afoldl' @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517
; │││┌ @ float.jl:400 within `+'
      %238 = fadd <8 x float> %237, %214
; │││└
; │││ @ operators.jl:517 within `afoldl' @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517
; │││┌ @ float.jl:400 within `+'
      %239 = fadd <8 x float> %238, %178
; │││└
; │││ @ operators.jl:517 within `afoldl' @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517
; │││┌ @ float.jl:400 within `+'
      %240 = fadd <8 x float> %239, %218
; │││└
; │││ @ operators.jl:517 within `afoldl' @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517
; │││┌ @ float.jl:400 within `+'
      %241 = fadd <8 x float> %240, %222
; │││└
; │││ @ operators.jl:517 within `afoldl' @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:516
; │││┌ @ float.jl:400 within `+'
      %242 = fadd <8 x float> %241, %226
; │└└└
; │┌ @ array.jl:782 within `setindex!'
    %243 = add i64 %174, %reass.mul64
    %244 = getelementptr inbounds float, float addrspace(13)* %54, i64 %243
    %245 = bitcast float addrspace(13)* %244 to <8 x float> addrspace(13)*
    store <8 x float> %242, <8 x float> addrspace(13)* %245, align 4
; │└
; │┌ @ array.jl:743 within `getindex'
    %index.next = add i64 %index, 8
    %246 = icmp eq i64 %index.next, %n.vec
    br i1 %246, label %middle.block, label %vector.body

middle.block:                                     ; preds = %vector.body
    %cmp.n = icmp eq i64 %29, %n.vec
; │└
; │ @ simdloop.jl:75 within `macro expansion'
   br i1 %cmp.n, label %L167, label %scalar.ph

scalar.ph:                                        ; preds = %middle.block, %vector.memcheck, %L61.lr.ph
   %bc.resume.val = phi i64 [ %n.vec, %middle.block ], [ 0, %L61.lr.ph ], [ 0, %vector.memcheck ]
   br label %L61

L61:                                              ; preds = %scalar.ph, %L61
   %value_phi1166 = phi i64 [ %bc.resume.val, %scalar.ph ], [ %251, %L61 ]
; │ @ simdloop.jl:77 within `macro expansion' @ /data/esdrd/tqff/tmp/issue.jl:7
; │┌ @ array.jl:743 within `getindex'
    %247 = add i64 %reass.mul, %value_phi1166
    %248 = getelementptr inbounds float, float addrspace(13)* %35, i64 %247
    %249 = load float, float addrspace(13)* %248, align 4
; │└
; │┌ @ float.jl:404 within `*'
    %250 = fmul float %249, %140
; │└
; │┌ @ array.jl:743 within `getindex'
    %251 = add nuw nsw i64 %value_phi1166, 1
    %252 = add i64 %251, %reass.mul
    %253 = getelementptr inbounds float, float addrspace(13)* %35, i64 %252
    %254 = load float, float addrspace(13)* %253, align 4
; │└
; │┌ @ float.jl:404 within `*'
    %255 = fmul float %254, %141
; │└
; │┌ @ array.jl:743 within `getindex'
    %256 = add nuw nsw i64 %value_phi1166, 2
    %257 = add i64 %256, %reass.mul
    %258 = getelementptr inbounds float, float addrspace(13)* %35, i64 %257
    %259 = load float, float addrspace(13)* %258, align 4
; │└
; │┌ @ float.jl:404 within `*'
    %260 = fmul float %259, %142
; │└
; │┌ @ array.jl:743 within `getindex'
    %261 = add nuw nsw i64 %value_phi1166, 3
    %262 = add i64 %261, %reass.mul
    %263 = getelementptr inbounds float, float addrspace(13)* %35, i64 %262
    %264 = load float, float addrspace(13)* %263, align 4
; │└
; │┌ @ float.jl:404 within `*'
    %265 = fmul float %264, %143
; │└
; │┌ @ array.jl:743 within `getindex'
    %266 = add nuw nsw i64 %value_phi1166, 4
    %267 = add i64 %266, %reass.mul
    %268 = getelementptr inbounds float, float addrspace(13)* %35, i64 %267
    %269 = load float, float addrspace(13)* %268, align 4
; │└
; │┌ @ float.jl:404 within `*'
    %270 = fmul float %269, %144
; │└
; │┌ @ array.jl:743 within `getindex'
    %271 = add nuw nsw i64 %value_phi1166, 5
    %272 = add i64 %271, %reass.mul
    %273 = getelementptr inbounds float, float addrspace(13)* %35, i64 %272
    %274 = load float, float addrspace(13)* %273, align 4
; │└
; │┌ @ float.jl:404 within `*'
    %275 = fmul float %274, %145
; │└
; │┌ @ array.jl:743 within `getindex'
    %276 = add nuw i64 %value_phi1166, 6
    %277 = add i64 %276, %reass.mul
    %278 = getelementptr inbounds float, float addrspace(13)* %35, i64 %277
    %279 = load float, float addrspace(13)* %278, align 4
; │└
; │┌ @ float.jl:404 within `*'
    %280 = fmul float %279, %146
; │└
; │┌ @ array.jl:743 within `getindex'
    %281 = add nuw i64 %value_phi1166, 7
    %282 = add i64 %281, %reass.mul
    %283 = getelementptr inbounds float, float addrspace(13)* %35, i64 %282
    %284 = load float, float addrspace(13)* %283, align 4
; │└
; │┌ @ float.jl:404 within `*'
    %285 = fmul float %284, %147
; │└
; │┌ @ array.jl:743 within `getindex'
    %286 = add nuw i64 %value_phi1166, 8
    %287 = add i64 %286, %reass.mul
    %288 = getelementptr inbounds float, float addrspace(13)* %35, i64 %287
    %289 = load float, float addrspace(13)* %288, align 4
; │└
; │┌ @ float.jl:404 within `*'
    %290 = fmul float %289, %148
; │└
; │┌ @ array.jl:743 within `getindex'
    %291 = add i64 %266, %reass.mul48
    %292 = getelementptr inbounds float, float addrspace(13)* %35, i64 %291
    %293 = load float, float addrspace(13)* %292, align 4
; │└
; │┌ @ float.jl:404 within `*'
    %294 = fmul float %293, %140
; │└
; │┌ @ array.jl:743 within `getindex'
    %295 = add i64 %266, %reass.mul50
    %296 = getelementptr inbounds float, float addrspace(13)* %35, i64 %295
    %297 = load float, float addrspace(13)* %296, align 4
; │└
; │┌ @ float.jl:404 within `*'
    %298 = fmul float %297, %141
; │└
; │┌ @ array.jl:743 within `getindex'
    %299 = add i64 %266, %reass.mul52
    %300 = getelementptr inbounds float, float addrspace(13)* %35, i64 %299
    %301 = load float, float addrspace(13)* %300, align 4
; │└
; │┌ @ float.jl:404 within `*'
    %302 = fmul float %301, %142
; │└
; │┌ @ array.jl:743 within `getindex'
    %303 = add i64 %266, %reass.mul54
    %304 = getelementptr inbounds float, float addrspace(13)* %35, i64 %303
    %305 = load float, float addrspace(13)* %304, align 4
; │└
; │┌ @ float.jl:404 within `*'
    %306 = fmul float %305, %143
; │└
; │┌ @ array.jl:743 within `getindex'
    %307 = add i64 %266, %reass.mul58
    %308 = getelementptr inbounds float, float addrspace(13)* %35, i64 %307
    %309 = load float, float addrspace(13)* %308, align 4
; │└
; │┌ @ float.jl:404 within `*'
    %310 = fmul float %309, %145
; │└
; │┌ @ array.jl:743 within `getindex'
    %311 = add i64 %266, %reass.mul60
    %312 = getelementptr inbounds float, float addrspace(13)* %35, i64 %311
    %313 = load float, float addrspace(13)* %312, align 4
; │└
; │┌ @ float.jl:404 within `*'
    %314 = fmul float %313, %146
; │└
; │┌ @ array.jl:743 within `getindex'
    %315 = add i64 %266, %reass.mul62
    %316 = getelementptr inbounds float, float addrspace(13)* %35, i64 %315
    %317 = load float, float addrspace(13)* %316, align 4
; │└
; │┌ @ float.jl:404 within `*'
    %318 = fmul float %317, %147
; │└
; │┌ @ operators.jl:529 within `+' @ float.jl:400
    %319 = fadd float %250, %255
    %320 = fadd float %319, %260
; ││ @ operators.jl:529 within `+'
; ││┌ @ operators.jl:517 within `afoldl'
; │││┌ @ float.jl:400 within `+'
      %321 = fadd float %320, %265
; │││└
; │││ @ operators.jl:517 within `afoldl' @ operators.jl:517
; │││┌ @ float.jl:400 within `+'
      %322 = fadd float %321, %270
; │││└
; │││ @ operators.jl:517 within `afoldl' @ operators.jl:517 @ operators.jl:517
; │││┌ @ float.jl:400 within `+'
      %323 = fadd float %322, %275
; │││└
; │││ @ operators.jl:517 within `afoldl' @ operators.jl:517 @ operators.jl:517 @ operators.jl:517
; │││┌ @ float.jl:400 within `+'
      %324 = fadd float %323, %280
; │││└
; │││ @ operators.jl:517 within `afoldl' @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517
; │││┌ @ float.jl:400 within `+'
      %325 = fadd float %324, %285
; │││└
; │││ @ operators.jl:517 within `afoldl' @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517
; │││┌ @ float.jl:400 within `+'
      %326 = fadd float %325, %290
; │││└
; │││ @ operators.jl:517 within `afoldl' @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517
; │││┌ @ float.jl:400 within `+'
      %327 = fadd float %326, %294
; │││└
; │││ @ operators.jl:517 within `afoldl' @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517
; │││┌ @ float.jl:400 within `+'
      %328 = fadd float %327, %298
; │││└
; │││ @ operators.jl:517 within `afoldl' @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517
; │││┌ @ float.jl:400 within `+'
      %329 = fadd float %328, %302
; │││└
; │││ @ operators.jl:517 within `afoldl' @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517
; │││┌ @ float.jl:400 within `+'
      %330 = fadd float %329, %306
; │││└
; │││ @ operators.jl:517 within `afoldl' @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517
; │││┌ @ float.jl:400 within `+'
      %331 = fadd float %330, %270
; │││└
; │││ @ operators.jl:517 within `afoldl' @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517
; │││┌ @ float.jl:400 within `+'
      %332 = fadd float %331, %310
; │││└
; │││ @ operators.jl:517 within `afoldl' @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517
; │││┌ @ float.jl:400 within `+'
      %333 = fadd float %332, %314
; │││└
; │││ @ operators.jl:517 within `afoldl' @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:517 @ operators.jl:516
; │││┌ @ float.jl:400 within `+'
      %334 = fadd float %333, %318
; │└└└
; │┌ @ array.jl:782 within `setindex!'
    %335 = add i64 %266, %reass.mul64
    %336 = getelementptr inbounds float, float addrspace(13)* %54, i64 %335
    store float %334, float addrspace(13)* %336, align 4
; │└
; │ @ simdloop.jl:75 within `macro expansion'
; │┌ @ int.jl:49 within `<'
    %337 = icmp ult i64 %251, %29
; │└
   br i1 %337, label %L61, label %L167

L167:                                             ; preds = %L61, %middle.block, %L55
; └
; ┌ @ range.jl:597 within `iterate'
; │┌ @ promotion.jl:403 within `=='
    %338 = icmp eq i64 %value_phi867, %23
; └└
  br i1 %338, label %L178, label %L32

L178:                                             ; preds = %L167, %L18
; ┌ @ range.jl:597 within `iterate'
; │┌ @ promotion.jl:403 within `=='
    %339 = icmp eq i64 %value_phi3, %13
; │└
; │ @ range.jl:598 within `iterate'
; │┌ @ int.jl:53 within `+'
    %340 = add i64 %value_phi3, 1
; └└
  %indvar.next = add i64 %indvar, 1
  br i1 %339, label %L189, label %L18

L189:                                             ; preds = %L178, %top
;  @ /data/esdrd/tqff/tmp/issue.jl:13 within `does_vectorize!'
  ret %jl_value_t addrspace(10)* addrspacecast (%jl_value_t* inttoptr (i64 47938844788480 to %jl_value_t*) to %jl_value_t addrspace(10)*)
}