; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 6
; RUN: opt < %s -passes="default<O2>" -S -mattr=+sse2 | FileCheck %s
; RUN: opt < %s -passes="default<O2>" -S -mattr=+avx  | FileCheck %s
; RUN: opt < %s -passes="default<O2>" -S -mattr=+avx2 | FileCheck %s
; RUN: opt < %s -passes="default<O2>" -S -mattr=+avx512f  | FileCheck %s
; RUN: opt < %s -passes="default<O2>" -S -mattr=+avx512bw | FileCheck %s

@a = dso_local global [1024 x i8] zeroinitializer, align 16
@b = dso_local global [1024 x i8] zeroinitializer, align 16

define dso_local i32 @sad_16i8() #0 {
; CHECK-LABEL: define dso_local i32 @sad_16i8(
; CHECK-SAME: ) local_unnamed_addr #[[ATTR0:[0-9]+]] {
; CHECK-NEXT:  [[ENTRY:.*]]:
; CHECK-NEXT:    br label %[[VECTOR_BODY:.*]]
; CHECK:       [[VECTOR_BODY]]:
; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <16 x i32> [ zeroinitializer, %[[ENTRY]] ], [ [[I10:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT:    [[I:%.*]] = getelementptr inbounds nuw i8, ptr @a, i64 [[INDEX]]
; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[I]], align 16
; CHECK-NEXT:    [[I2:%.*]] = zext <16 x i8> [[WIDE_LOAD]] to <16 x i32>
; CHECK-NEXT:    [[I3:%.*]] = getelementptr inbounds nuw i8, ptr @b, i64 [[INDEX]]
; CHECK-NEXT:    [[WIDE_LOAD1:%.*]] = load <16 x i8>, ptr [[I3]], align 16
; CHECK-NEXT:    [[I5:%.*]] = zext <16 x i8> [[WIDE_LOAD1]] to <16 x i32>
; CHECK-NEXT:    [[I6:%.*]] = sub nsw <16 x i32> [[I2]], [[I5]]
; CHECK-NEXT:    [[I9:%.*]] = tail call <16 x i32> @llvm.abs.v16i32(<16 x i32> [[I6]], i1 true)
; CHECK-NEXT:    [[I10]] = add nuw nsw <16 x i32> [[I9]], [[VEC_PHI]]
; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw nsw i64 [[INDEX]], 16
; CHECK-NEXT:    [[I11:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024
; CHECK-NEXT:    br i1 [[I11]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]]
; CHECK:       [[MIDDLE_BLOCK]]:
; CHECK-NEXT:    [[I12:%.*]] = tail call i32 @llvm.vector.reduce.add.v16i32(<16 x i32> [[I10]])
; CHECK-NEXT:    ret i32 [[I12]]
;
entry:
  br label %vector.body

vector.body:                                      ; preds = %vector.body, %entry
  %index = phi i64 [ 0, %entry ], [ %index.next, %vector.body ]
  %vec.phi = phi <16 x i32> [ zeroinitializer, %entry ], [ %i10, %vector.body ]
  %i = getelementptr inbounds [1024 x i8], ptr @a, i64 0, i64 %index
  %i1 = bitcast ptr %i to ptr
  %wide.load = load <16 x i8>, ptr %i1, align 4
  %i2 = zext <16 x i8> %wide.load to <16 x i32>
  %i3 = getelementptr inbounds [1024 x i8], ptr @b, i64 0, i64 %index
  %i4 = bitcast ptr %i3 to ptr
  %wide.load1 = load <16 x i8>, ptr %i4, align 4
  %i5 = zext <16 x i8> %wide.load1 to <16 x i32>
  %i6 = sub nsw <16 x i32> %i2, %i5
  %i7 = icmp sgt <16 x i32> %i6, splat (i32 -1)
  %i8 = sub nsw <16 x i32> zeroinitializer, %i6
  %i9 = select <16 x i1> %i7, <16 x i32> %i6, <16 x i32> %i8
  %i10 = add nsw <16 x i32> %i9, %vec.phi
  %index.next = add i64 %index, 16
  %i11 = icmp eq i64 %index.next, 1024
  br i1 %i11, label %middle.block, label %vector.body

middle.block:                                     ; preds = %vector.body
  %rdx.shuf = shufflevector <16 x i32> %i10, <16 x i32> undef, <16 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
  %bin.rdx = add <16 x i32> %i10, %rdx.shuf
  %rdx.shuf2 = shufflevector <16 x i32> %bin.rdx, <16 x i32> undef, <16 x i32> <i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
  %bin.rdx2 = add <16 x i32> %bin.rdx, %rdx.shuf2
  %rdx.shuf3 = shufflevector <16 x i32> %bin.rdx2, <16 x i32> undef, <16 x i32> <i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
  %bin.rdx3 = add <16 x i32> %bin.rdx2, %rdx.shuf3
  %rdx.shuf4 = shufflevector <16 x i32> %bin.rdx3, <16 x i32> undef, <16 x i32> <i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
  %bin.rdx4 = add <16 x i32> %bin.rdx3, %rdx.shuf4
  %i12 = extractelement <16 x i32> %bin.rdx4, i32 0
  ret i32 %i12
}

define dso_local i32 @sad_32i8() #0 {
; CHECK-LABEL: define dso_local i32 @sad_32i8(
; CHECK-SAME: ) local_unnamed_addr #[[ATTR0]] {
; CHECK-NEXT:  [[ENTRY:.*]]:
; CHECK-NEXT:    br label %[[VECTOR_BODY:.*]]
; CHECK:       [[VECTOR_BODY]]:
; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <32 x i32> [ zeroinitializer, %[[ENTRY]] ], [ [[I10:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT:    [[I:%.*]] = getelementptr inbounds nuw i8, ptr @a, i64 [[INDEX]]
; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <32 x i8>, ptr [[I]], align 32
; CHECK-NEXT:    [[I2:%.*]] = zext <32 x i8> [[WIDE_LOAD]] to <32 x i32>
; CHECK-NEXT:    [[I3:%.*]] = getelementptr inbounds nuw i8, ptr @b, i64 [[INDEX]]
; CHECK-NEXT:    [[WIDE_LOAD1:%.*]] = load <32 x i8>, ptr [[I3]], align 32
; CHECK-NEXT:    [[I5:%.*]] = zext <32 x i8> [[WIDE_LOAD1]] to <32 x i32>
; CHECK-NEXT:    [[I6:%.*]] = sub nsw <32 x i32> [[I2]], [[I5]]
; CHECK-NEXT:    [[I9:%.*]] = tail call <32 x i32> @llvm.abs.v32i32(<32 x i32> [[I6]], i1 true)
; CHECK-NEXT:    [[I10]] = add nuw nsw <32 x i32> [[I9]], [[VEC_PHI]]
; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw nsw i64 [[INDEX]], 32
; CHECK-NEXT:    [[I11:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024
; CHECK-NEXT:    br i1 [[I11]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]]
; CHECK:       [[MIDDLE_BLOCK]]:
; CHECK-NEXT:    [[I12:%.*]] = tail call i32 @llvm.vector.reduce.add.v32i32(<32 x i32> [[I10]])
; CHECK-NEXT:    ret i32 [[I12]]
;
entry:
  br label %vector.body

vector.body:                                      ; preds = %vector.body, %entry
  %index = phi i64 [ 0, %entry ], [ %index.next, %vector.body ]
  %vec.phi = phi <32 x i32> [ zeroinitializer, %entry ], [ %i10, %vector.body ]
  %i = getelementptr inbounds [1024 x i8], ptr @a, i64 0, i64 %index
  %i1 = bitcast ptr %i to ptr
  %wide.load = load <32 x i8>, ptr %i1, align 32
  %i2 = zext <32 x i8> %wide.load to <32 x i32>
  %i3 = getelementptr inbounds [1024 x i8], ptr @b, i64 0, i64 %index
  %i4 = bitcast ptr %i3 to ptr
  %wide.load1 = load <32 x i8>, ptr %i4, align 32
  %i5 = zext <32 x i8> %wide.load1 to <32 x i32>
  %i6 = sub nsw <32 x i32> %i2, %i5
  %i7 = icmp sgt <32 x i32> %i6, splat (i32 -1)
  %i8 = sub nsw <32 x i32> zeroinitializer, %i6
  %i9 = select <32 x i1> %i7, <32 x i32> %i6, <32 x i32> %i8
  %i10 = add nsw <32 x i32> %i9, %vec.phi
  %index.next = add i64 %index, 32
  %i11 = icmp eq i64 %index.next, 1024
  br i1 %i11, label %middle.block, label %vector.body

middle.block:                                     ; preds = %vector.body
  %rdx.shuf = shufflevector <32 x i32> %i10, <32 x i32> undef, <32 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
  %bin.rdx = add <32 x i32> %i10, %rdx.shuf
  %rdx.shuf2 = shufflevector <32 x i32> %bin.rdx, <32 x i32> undef, <32 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
  %bin.rdx2 = add <32 x i32> %bin.rdx, %rdx.shuf2
  %rdx.shuf3 = shufflevector <32 x i32> %bin.rdx2, <32 x i32> undef, <32 x i32> <i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
  %bin.rdx3 = add <32 x i32> %bin.rdx2, %rdx.shuf3
  %rdx.shuf4 = shufflevector <32 x i32> %bin.rdx3, <32 x i32> undef, <32 x i32> <i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
  %bin.rdx4 = add <32 x i32> %bin.rdx3, %rdx.shuf4
  %rdx.shuf5 = shufflevector <32 x i32> %bin.rdx4, <32 x i32> undef, <32 x i32> <i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
  %bin.rdx5 = add <32 x i32> %bin.rdx4, %rdx.shuf5
  %i12 = extractelement <32 x i32> %bin.rdx5, i32 0
  ret i32 %i12
}

define dso_local i32 @sad_avx64i8() #0 {
; CHECK-LABEL: define dso_local i32 @sad_avx64i8(
; CHECK-SAME: ) local_unnamed_addr #[[ATTR0]] {
; CHECK-NEXT:  [[ENTRY:.*]]:
; CHECK-NEXT:    br label %[[VECTOR_BODY:.*]]
; CHECK:       [[VECTOR_BODY]]:
; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <64 x i32> [ zeroinitializer, %[[ENTRY]] ], [ [[I10:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT:    [[I:%.*]] = getelementptr inbounds nuw i8, ptr @a, i64 [[INDEX]]
; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <64 x i8>, ptr [[I]], align 64
; CHECK-NEXT:    [[I2:%.*]] = zext <64 x i8> [[WIDE_LOAD]] to <64 x i32>
; CHECK-NEXT:    [[I3:%.*]] = getelementptr inbounds nuw i8, ptr @b, i64 [[INDEX]]
; CHECK-NEXT:    [[WIDE_LOAD1:%.*]] = load <64 x i8>, ptr [[I3]], align 64
; CHECK-NEXT:    [[I5:%.*]] = zext <64 x i8> [[WIDE_LOAD1]] to <64 x i32>
; CHECK-NEXT:    [[I6:%.*]] = sub nsw <64 x i32> [[I2]], [[I5]]
; CHECK-NEXT:    [[I9:%.*]] = tail call <64 x i32> @llvm.abs.v64i32(<64 x i32> [[I6]], i1 true)
; CHECK-NEXT:    [[I10]] = add nuw nsw <64 x i32> [[I9]], [[VEC_PHI]]
; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw nsw i64 [[INDEX]], 64
; CHECK-NEXT:    [[I11:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024
; CHECK-NEXT:    br i1 [[I11]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]]
; CHECK:       [[MIDDLE_BLOCK]]:
; CHECK-NEXT:    [[I12:%.*]] = tail call i32 @llvm.vector.reduce.add.v64i32(<64 x i32> [[I10]])
; CHECK-NEXT:    ret i32 [[I12]]
;
entry:
  br label %vector.body

vector.body:                                      ; preds = %vector.body, %entry
  %index = phi i64 [ 0, %entry ], [ %index.next, %vector.body ]
  %vec.phi = phi <64 x i32> [ zeroinitializer, %entry ], [ %i10, %vector.body ]
  %i = getelementptr inbounds [1024 x i8], ptr @a, i64 0, i64 %index
  %i1 = bitcast ptr %i to ptr
  %wide.load = load <64 x i8>, ptr %i1, align 64
  %i2 = zext <64 x i8> %wide.load to <64 x i32>
  %i3 = getelementptr inbounds [1024 x i8], ptr @b, i64 0, i64 %index
  %i4 = bitcast ptr %i3 to ptr
  %wide.load1 = load <64 x i8>, ptr %i4, align 64
  %i5 = zext <64 x i8> %wide.load1 to <64 x i32>
  %i6 = sub nsw <64 x i32> %i2, %i5
  %i7 = icmp sgt <64 x i32> %i6, splat (i32 -1)
  %i8 = sub nsw <64 x i32> zeroinitializer, %i6
  %i9 = select <64 x i1> %i7, <64 x i32> %i6, <64 x i32> %i8
  %i10 = add nsw <64 x i32> %i9, %vec.phi
  %index.next = add i64 %index, 64
  %i11 = icmp eq i64 %index.next, 1024
  br i1 %i11, label %middle.block, label %vector.body

middle.block:                                     ; preds = %vector.body
  %rdx.shuf = shufflevector <64 x i32> %i10, <64 x i32> undef, <64 x i32> <i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47, i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
  %bin.rdx = add <64 x i32> %i10, %rdx.shuf
  %rdx.shuf2 = shufflevector <64 x i32> %bin.rdx, <64 x i32> undef, <64 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
  %bin.rdx2 = add <64 x i32> %bin.rdx, %rdx.shuf2
  %rdx.shuf3 = shufflevector <64 x i32> %bin.rdx2, <64 x i32> undef, <64 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
  %bin.rdx3 = add <64 x i32> %bin.rdx2, %rdx.shuf3
  %rdx.shuf4 = shufflevector <64 x i32> %bin.rdx3, <64 x i32> undef, <64 x i32> <i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
  %bin.rdx4 = add <64 x i32> %bin.rdx3, %rdx.shuf4
  %rdx.shuf5 = shufflevector <64 x i32> %bin.rdx4, <64 x i32> undef, <64 x i32> <i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
  %bin.rdx5 = add <64 x i32> %bin.rdx4, %rdx.shuf5
  %rdx.shuf6 = shufflevector <64 x i32> %bin.rdx5, <64 x i32> undef, <64 x i32> <i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
  %bin.rdx6 = add <64 x i32> %bin.rdx5, %rdx.shuf6
  %i12 = extractelement <64 x i32> %bin.rdx6, i32 0
  ret i32 %i12
}

define dso_local i32 @sad_2i8() #0 {
; CHECK-LABEL: define dso_local i32 @sad_2i8(
; CHECK-SAME: ) local_unnamed_addr #[[ATTR0]] {
; CHECK-NEXT:  [[ENTRY:.*]]:
; CHECK-NEXT:    br label %[[VECTOR_BODY:.*]]
; CHECK:       [[VECTOR_BODY]]:
; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <2 x i32> [ zeroinitializer, %[[ENTRY]] ], [ [[I10:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT:    [[I:%.*]] = getelementptr inbounds nuw i8, ptr @a, i64 [[INDEX]]
; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i8>, ptr [[I]], align 4
; CHECK-NEXT:    [[I2:%.*]] = zext <2 x i8> [[WIDE_LOAD]] to <2 x i32>
; CHECK-NEXT:    [[I3:%.*]] = getelementptr inbounds nuw i8, ptr @b, i64 [[INDEX]]
; CHECK-NEXT:    [[WIDE_LOAD1:%.*]] = load <2 x i8>, ptr [[I3]], align 4
; CHECK-NEXT:    [[I5:%.*]] = zext <2 x i8> [[WIDE_LOAD1]] to <2 x i32>
; CHECK-NEXT:    [[I6:%.*]] = sub nsw <2 x i32> [[I2]], [[I5]]
; CHECK-NEXT:    [[I9:%.*]] = tail call <2 x i32> @llvm.abs.v2i32(<2 x i32> [[I6]], i1 true)
; CHECK-NEXT:    [[I10]] = add nuw nsw <2 x i32> [[I9]], [[VEC_PHI]]
; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw nsw i64 [[INDEX]], 2
; CHECK-NEXT:    [[I11:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024
; CHECK-NEXT:    br i1 [[I11]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]]
; CHECK:       [[MIDDLE_BLOCK]]:
; CHECK-NEXT:    [[I12:%.*]] = tail call i32 @llvm.vector.reduce.add.v2i32(<2 x i32> [[I10]])
; CHECK-NEXT:    ret i32 [[I12]]
;
entry:
  br label %vector.body

vector.body:                                      ; preds = %vector.body, %entry
  %index = phi i64 [ 0, %entry ], [ %index.next, %vector.body ]
  %vec.phi = phi <2 x i32> [ zeroinitializer, %entry ], [ %i10, %vector.body ]
  %i = getelementptr inbounds [1024 x i8], ptr @a, i64 0, i64 %index
  %i1 = bitcast ptr %i to ptr
  %wide.load = load <2 x i8>, ptr %i1, align 4
  %i2 = zext <2 x i8> %wide.load to <2 x i32>
  %i3 = getelementptr inbounds [1024 x i8], ptr @b, i64 0, i64 %index
  %i4 = bitcast ptr %i3 to ptr
  %wide.load1 = load <2 x i8>, ptr %i4, align 4
  %i5 = zext <2 x i8> %wide.load1 to <2 x i32>
  %i6 = sub nsw <2 x i32> %i2, %i5
  %i7 = icmp sgt <2 x i32> %i6, splat (i32 -1)
  %i8 = sub nsw <2 x i32> zeroinitializer, %i6
  %i9 = select <2 x i1> %i7, <2 x i32> %i6, <2 x i32> %i8
  %i10 = add nsw <2 x i32> %i9, %vec.phi
  %index.next = add i64 %index, 2
  %i11 = icmp eq i64 %index.next, 1024
  br i1 %i11, label %middle.block, label %vector.body

middle.block:                                     ; preds = %vector.body
  %rdx.shuf = shufflevector <2 x i32> %i10, <2 x i32> undef, <2 x i32> <i32 1, i32 poison>
  %bin.rdx = add <2 x i32> %i10, %rdx.shuf
  %i12 = extractelement <2 x i32> %bin.rdx, i32 0
  ret i32 %i12
}

define dso_local i32 @sad_4i8() #0 {
; CHECK-LABEL: define dso_local i32 @sad_4i8(
; CHECK-SAME: ) local_unnamed_addr #[[ATTR0]] {
; CHECK-NEXT:  [[ENTRY:.*]]:
; CHECK-NEXT:    br label %[[VECTOR_BODY:.*]]
; CHECK:       [[VECTOR_BODY]]:
; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i32> [ zeroinitializer, %[[ENTRY]] ], [ [[I10:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT:    [[I:%.*]] = getelementptr inbounds nuw i8, ptr @a, i64 [[INDEX]]
; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i8>, ptr [[I]], align 4
; CHECK-NEXT:    [[I2:%.*]] = zext <4 x i8> [[WIDE_LOAD]] to <4 x i32>
; CHECK-NEXT:    [[I3:%.*]] = getelementptr inbounds nuw i8, ptr @b, i64 [[INDEX]]
; CHECK-NEXT:    [[WIDE_LOAD1:%.*]] = load <4 x i8>, ptr [[I3]], align 4
; CHECK-NEXT:    [[I5:%.*]] = zext <4 x i8> [[WIDE_LOAD1]] to <4 x i32>
; CHECK-NEXT:    [[I6:%.*]] = sub nsw <4 x i32> [[I2]], [[I5]]
; CHECK-NEXT:    [[I9:%.*]] = tail call <4 x i32> @llvm.abs.v4i32(<4 x i32> [[I6]], i1 true)
; CHECK-NEXT:    [[I10]] = add nuw nsw <4 x i32> [[I9]], [[VEC_PHI]]
; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw nsw i64 [[INDEX]], 4
; CHECK-NEXT:    [[I11:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024
; CHECK-NEXT:    br i1 [[I11]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]]
; CHECK:       [[MIDDLE_BLOCK]]:
; CHECK-NEXT:    [[SUM:%.*]] = tail call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> [[I10]])
; CHECK-NEXT:    ret i32 [[SUM]]
;
entry:
  br label %vector.body

vector.body:                                      ; preds = %vector.body, %entry
  %index = phi i64 [ 0, %entry ], [ %index.next, %vector.body ]
  %vec.phi = phi <4 x i32> [ zeroinitializer, %entry ], [ %i10, %vector.body ]
  %i = getelementptr inbounds [1024 x i8], ptr @a, i64 0, i64 %index
  %i1 = bitcast ptr %i to ptr
  %wide.load = load <4 x i8>, ptr %i1, align 4
  %i2 = zext <4 x i8> %wide.load to <4 x i32>
  %i3 = getelementptr inbounds [1024 x i8], ptr @b, i64 0, i64 %index
  %i4 = bitcast ptr %i3 to ptr
  %wide.load1 = load <4 x i8>, ptr %i4, align 4
  %i5 = zext <4 x i8> %wide.load1 to <4 x i32>
  %i6 = sub nsw <4 x i32> %i2, %i5
  %i7 = icmp sgt <4 x i32> %i6, splat (i32 -1)
  %i8 = sub nsw <4 x i32> zeroinitializer, %i6
  %i9 = select <4 x i1> %i7, <4 x i32> %i6, <4 x i32> %i8
  %i10 = add nsw <4 x i32> %i9, %vec.phi
  %index.next = add i64 %index, 4
  %i11 = icmp eq i64 %index.next, 1024
  br i1 %i11, label %middle.block, label %vector.body

middle.block:                                     ; preds = %vector.body
  %h2 = shufflevector <4 x i32> %i10, <4 x i32> undef, <4 x i32> <i32 2, i32 3, i32 poison, i32 poison>
  %sum2 = add <4 x i32> %i10, %h2
  %h3 = shufflevector <4 x i32> %sum2, <4 x i32> undef, <4 x i32> <i32 1, i32 poison, i32 poison, i32 poison>
  %sum3 = add <4 x i32> %sum2, %h3
  %sum = extractelement <4 x i32> %sum3, i32 0
  ret i32 %sum
}

define dso_local i32 @sad_nonloop_4i8(ptr readonly captures(none) %p, i64 %arg, ptr readonly captures(none) %q) local_unnamed_addr #1 {
; CHECK-LABEL: define dso_local i32 @sad_nonloop_4i8(
; CHECK-SAME: ptr readonly captures(none) [[P:%.*]], i64 [[ARG:%.*]], ptr readonly captures(none) [[Q:%.*]]) local_unnamed_addr #[[ATTR1:[0-9]+]] {
; CHECK-NEXT:  [[BB:.*:]]
; CHECK-NEXT:    [[V1:%.*]] = load <4 x i8>, ptr [[P]], align 1
; CHECK-NEXT:    [[Z1:%.*]] = zext <4 x i8> [[V1]] to <4 x i32>
; CHECK-NEXT:    [[V2:%.*]] = load <4 x i8>, ptr [[Q]], align 1
; CHECK-NEXT:    [[Z2:%.*]] = zext <4 x i8> [[V2]] to <4 x i32>
; CHECK-NEXT:    [[SUB:%.*]] = sub nsw <4 x i32> [[Z1]], [[Z2]]
; CHECK-NEXT:    [[ABS:%.*]] = tail call <4 x i32> @llvm.abs.v4i32(<4 x i32> [[SUB]], i1 true)
; CHECK-NEXT:    [[SUM:%.*]] = tail call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> [[ABS]])
; CHECK-NEXT:    ret i32 [[SUM]]
;
bb:
  %v1 = load <4 x i8>, ptr %p, align 1
  %z1 = zext <4 x i8> %v1 to <4 x i32>
  %v2 = load <4 x i8>, ptr %q, align 1
  %z2 = zext <4 x i8> %v2 to <4 x i32>
  %sub = sub nsw <4 x i32> %z1, %z2
  %isneg = icmp sgt <4 x i32> %sub, splat (i32 -1)
  %neg = sub nsw <4 x i32> zeroinitializer, %sub
  %abs = select <4 x i1> %isneg, <4 x i32> %sub, <4 x i32> %neg
  %h2 = shufflevector <4 x i32> %abs, <4 x i32> undef, <4 x i32> <i32 2, i32 3, i32 poison, i32 poison>
  %sum2 = add <4 x i32> %abs, %h2
  %h3 = shufflevector <4 x i32> %sum2, <4 x i32> undef, <4 x i32> <i32 1, i32 poison, i32 poison, i32 poison>
  %sum3 = add <4 x i32> %sum2, %h3
  %sum = extractelement <4 x i32> %sum3, i32 0
  ret i32 %sum
}

define dso_local i32 @sad_nonloop_8i8(ptr readonly captures(none) %p, i64 %arg, ptr readonly captures(none) %q) local_unnamed_addr #1 {
; CHECK-LABEL: define dso_local i32 @sad_nonloop_8i8(
; CHECK-SAME: ptr readonly captures(none) [[P:%.*]], i64 [[ARG:%.*]], ptr readonly captures(none) [[Q:%.*]]) local_unnamed_addr #[[ATTR1]] {
; CHECK-NEXT:  [[BB:.*:]]
; CHECK-NEXT:    [[V1:%.*]] = load <8 x i8>, ptr [[P]], align 1
; CHECK-NEXT:    [[Z1:%.*]] = zext <8 x i8> [[V1]] to <8 x i32>
; CHECK-NEXT:    [[V2:%.*]] = load <8 x i8>, ptr [[Q]], align 1
; CHECK-NEXT:    [[Z2:%.*]] = zext <8 x i8> [[V2]] to <8 x i32>
; CHECK-NEXT:    [[SUB:%.*]] = sub nsw <8 x i32> [[Z1]], [[Z2]]
; CHECK-NEXT:    [[ABS:%.*]] = tail call <8 x i32> @llvm.abs.v8i32(<8 x i32> [[SUB]], i1 true)
; CHECK-NEXT:    [[SUM:%.*]] = tail call i32 @llvm.vector.reduce.add.v8i32(<8 x i32> [[ABS]])
; CHECK-NEXT:    ret i32 [[SUM]]
;
bb:
  %v1 = load <8 x i8>, ptr %p, align 1
  %z1 = zext <8 x i8> %v1 to <8 x i32>
  %v2 = load <8 x i8>, ptr %q, align 1
  %z2 = zext <8 x i8> %v2 to <8 x i32>
  %sub = sub nsw <8 x i32> %z1, %z2
  %isneg = icmp sgt <8 x i32> %sub, splat (i32 -1)
  %neg = sub nsw <8 x i32> zeroinitializer, %sub
  %abs = select <8 x i1> %isneg, <8 x i32> %sub, <8 x i32> %neg
  %h1 = shufflevector <8 x i32> %abs, <8 x i32> undef, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison>
  %sum1 = add <8 x i32> %abs, %h1
  %h2 = shufflevector <8 x i32> %sum1, <8 x i32> undef, <8 x i32> <i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
  %sum2 = add <8 x i32> %sum1, %h2
  %h3 = shufflevector <8 x i32> %sum2, <8 x i32> undef, <8 x i32> <i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
  %sum3 = add <8 x i32> %sum2, %h3
  %sum = extractelement <8 x i32> %sum3, i32 0
  ret i32 %sum
}

define dso_local i32 @sad_nonloop_16i8(ptr readonly captures(none) %p, i64 %arg, ptr readonly captures(none) %q) local_unnamed_addr #1 {
; CHECK-LABEL: define dso_local i32 @sad_nonloop_16i8(
; CHECK-SAME: ptr readonly captures(none) [[P:%.*]], i64 [[ARG:%.*]], ptr readonly captures(none) [[Q:%.*]]) local_unnamed_addr #[[ATTR1]] {
; CHECK-NEXT:  [[BB:.*:]]
; CHECK-NEXT:    [[V1:%.*]] = load <16 x i8>, ptr [[P]], align 1
; CHECK-NEXT:    [[Z1:%.*]] = zext <16 x i8> [[V1]] to <16 x i32>
; CHECK-NEXT:    [[V2:%.*]] = load <16 x i8>, ptr [[Q]], align 1
; CHECK-NEXT:    [[Z2:%.*]] = zext <16 x i8> [[V2]] to <16 x i32>
; CHECK-NEXT:    [[SUB:%.*]] = sub nsw <16 x i32> [[Z1]], [[Z2]]
; CHECK-NEXT:    [[ABS:%.*]] = tail call <16 x i32> @llvm.abs.v16i32(<16 x i32> [[SUB]], i1 true)
; CHECK-NEXT:    [[SUM:%.*]] = tail call i32 @llvm.vector.reduce.add.v16i32(<16 x i32> [[ABS]])
; CHECK-NEXT:    ret i32 [[SUM]]
;
bb:
  %v1 = load <16 x i8>, ptr %p, align 1
  %z1 = zext <16 x i8> %v1 to <16 x i32>
  %v2 = load <16 x i8>, ptr %q, align 1
  %z2 = zext <16 x i8> %v2 to <16 x i32>
  %sub = sub nsw <16 x i32> %z1, %z2
  %isneg = icmp sgt <16 x i32> %sub, splat (i32 -1)
  %neg = sub nsw <16 x i32> zeroinitializer, %sub
  %abs = select <16 x i1> %isneg, <16 x i32> %sub, <16 x i32> %neg
  %h0 = shufflevector <16 x i32> %abs, <16 x i32> undef, <16 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
  %sum0 = add <16 x i32> %abs, %h0
  %h1 = shufflevector <16 x i32> %sum0, <16 x i32> undef, <16 x i32> <i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
  %sum1 = add <16 x i32> %sum0, %h1
  %h2 = shufflevector <16 x i32> %sum1, <16 x i32> undef, <16 x i32> <i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
  %sum2 = add <16 x i32> %sum1, %h2
  %h3 = shufflevector <16 x i32> %sum2, <16 x i32> undef, <16 x i32> <i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
  %sum3 = add <16 x i32> %sum2, %h3
  %sum = extractelement <16 x i32> %sum3, i32 0
  ret i32 %sum
}

define dso_local i32 @sad_nonloop_32i8(ptr readonly captures(none) %p, i64 %arg, ptr readonly captures(none) %q) local_unnamed_addr #1 {
; CHECK-LABEL: define dso_local i32 @sad_nonloop_32i8(
; CHECK-SAME: ptr readonly captures(none) [[P:%.*]], i64 [[ARG:%.*]], ptr readonly captures(none) [[Q:%.*]]) local_unnamed_addr #[[ATTR1]] {
; CHECK-NEXT:  [[BB:.*:]]
; CHECK-NEXT:    [[V1:%.*]] = load <32 x i8>, ptr [[P]], align 1
; CHECK-NEXT:    [[Z1:%.*]] = zext <32 x i8> [[V1]] to <32 x i32>
; CHECK-NEXT:    [[V2:%.*]] = load <32 x i8>, ptr [[Q]], align 1
; CHECK-NEXT:    [[Z2:%.*]] = zext <32 x i8> [[V2]] to <32 x i32>
; CHECK-NEXT:    [[SUB:%.*]] = sub nsw <32 x i32> [[Z1]], [[Z2]]
; CHECK-NEXT:    [[ABS:%.*]] = tail call <32 x i32> @llvm.abs.v32i32(<32 x i32> [[SUB]], i1 true)
; CHECK-NEXT:    [[SUM:%.*]] = tail call i32 @llvm.vector.reduce.add.v32i32(<32 x i32> [[ABS]])
; CHECK-NEXT:    ret i32 [[SUM]]
;
bb:
  %v1 = load <32 x i8>, ptr %p, align 1
  %z1 = zext <32 x i8> %v1 to <32 x i32>
  %v2 = load <32 x i8>, ptr %q, align 1
  %z2 = zext <32 x i8> %v2 to <32 x i32>
  %sub = sub nsw <32 x i32> %z1, %z2
  %isneg = icmp sgt <32 x i32> %sub, splat (i32 -1)
  %neg = sub nsw <32 x i32> zeroinitializer, %sub
  %abs = select <32 x i1> %isneg, <32 x i32> %sub, <32 x i32> %neg
  %h32 = shufflevector <32 x i32> %abs, <32 x i32> undef, <32 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
  %sum32 = add <32 x i32> %abs, %h32
  %h0 = shufflevector <32 x i32> %sum32, <32 x i32> undef, <32 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
  %sum0 = add <32 x i32> %sum32, %h0
  %h1 = shufflevector <32 x i32> %sum0, <32 x i32> undef, <32 x i32> <i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
  %sum1 = add <32 x i32> %sum0, %h1
  %h2 = shufflevector <32 x i32> %sum1, <32 x i32> undef, <32 x i32> <i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
  %sum2 = add <32 x i32> %sum1, %h2
  %h3 = shufflevector <32 x i32> %sum2, <32 x i32> undef, <32 x i32> <i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
  %sum3 = add <32 x i32> %sum2, %h3
  %sum = extractelement <32 x i32> %sum3, i32 0
  ret i32 %sum
}

define dso_local i32 @sad_nonloop_64i8(ptr readonly captures(none) %p, i64 %arg, ptr readonly captures(none) %q) local_unnamed_addr #1 {
; CHECK-LABEL: define dso_local i32 @sad_nonloop_64i8(
; CHECK-SAME: ptr readonly captures(none) [[P:%.*]], i64 [[ARG:%.*]], ptr readonly captures(none) [[Q:%.*]]) local_unnamed_addr #[[ATTR1]] {
; CHECK-NEXT:  [[BB:.*:]]
; CHECK-NEXT:    [[V1:%.*]] = load <64 x i8>, ptr [[P]], align 1
; CHECK-NEXT:    [[Z1:%.*]] = zext <64 x i8> [[V1]] to <64 x i32>
; CHECK-NEXT:    [[V2:%.*]] = load <64 x i8>, ptr [[Q]], align 1
; CHECK-NEXT:    [[Z2:%.*]] = zext <64 x i8> [[V2]] to <64 x i32>
; CHECK-NEXT:    [[SUB:%.*]] = sub nsw <64 x i32> [[Z1]], [[Z2]]
; CHECK-NEXT:    [[ABS:%.*]] = tail call <64 x i32> @llvm.abs.v64i32(<64 x i32> [[SUB]], i1 true)
; CHECK-NEXT:    [[SUM:%.*]] = tail call i32 @llvm.vector.reduce.add.v64i32(<64 x i32> [[ABS]])
; CHECK-NEXT:    ret i32 [[SUM]]
;
bb:
  %v1 = load <64 x i8>, ptr %p, align 1
  %z1 = zext <64 x i8> %v1 to <64 x i32>
  %v2 = load <64 x i8>, ptr %q, align 1
  %z2 = zext <64 x i8> %v2 to <64 x i32>
  %sub = sub nsw <64 x i32> %z1, %z2
  %isneg = icmp sgt <64 x i32> %sub, splat (i32 -1)
  %neg = sub nsw <64 x i32> zeroinitializer, %sub
  %abs = select <64 x i1> %isneg, <64 x i32> %sub, <64 x i32> %neg
  %h64 = shufflevector <64 x i32> %abs, <64 x i32> undef, <64 x i32> <i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47, i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
  %sum64 = add <64 x i32> %abs, %h64
  %h32 = shufflevector <64 x i32> %sum64, <64 x i32> undef, <64 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
  %sum32 = add <64 x i32> %sum64, %h32
  %h0 = shufflevector <64 x i32> %sum32, <64 x i32> undef, <64 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
  %sum0 = add <64 x i32> %sum32, %h0
  %h1 = shufflevector <64 x i32> %sum0, <64 x i32> undef, <64 x i32> <i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
  %sum1 = add <64 x i32> %sum0, %h1
  %h2 = shufflevector <64 x i32> %sum1, <64 x i32> undef, <64 x i32> <i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
  %sum2 = add <64 x i32> %sum1, %h2
  %h3 = shufflevector <64 x i32> %sum2, <64 x i32> undef, <64 x i32> <i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
  %sum3 = add <64 x i32> %sum2, %h3
  %sum = extractelement <64 x i32> %sum3, i32 0
  ret i32 %sum
}

define dso_local i32 @sad_unroll_nonzero_initial(ptr %arg, ptr %arg1, ptr %arg2, ptr %arg3) {
; CHECK-LABEL: define dso_local i32 @sad_unroll_nonzero_initial(
; CHECK-SAME: ptr readonly captures(none) [[ARG:%.*]], ptr readonly captures(none) [[ARG1:%.*]], ptr readonly captures(none) [[ARG2:%.*]], ptr readonly captures(none) [[ARG3:%.*]]) local_unnamed_addr #[[ATTR2:[0-9]+]] {
; CHECK-NEXT:  [[BB:.*:]]
; CHECK-NEXT:    [[TMP:%.*]] = load <16 x i8>, ptr [[ARG]], align 1
; CHECK-NEXT:    [[TMP4:%.*]] = load <16 x i8>, ptr [[ARG1]], align 1
; CHECK-NEXT:    [[TMP5:%.*]] = zext <16 x i8> [[TMP]] to <16 x i32>
; CHECK-NEXT:    [[TMP6:%.*]] = zext <16 x i8> [[TMP4]] to <16 x i32>
; CHECK-NEXT:    [[TMP7:%.*]] = sub nsw <16 x i32> [[TMP5]], [[TMP6]]
; CHECK-NEXT:    [[TMP10:%.*]] = tail call <16 x i32> @llvm.abs.v16i32(<16 x i32> [[TMP7]], i1 true)
; CHECK-NEXT:    [[TMP11:%.*]] = add nuw nsw <16 x i32> [[TMP10]], <i32 1, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0>
; CHECK-NEXT:    [[TMP12:%.*]] = load <16 x i8>, ptr [[ARG2]], align 1
; CHECK-NEXT:    [[TMP13:%.*]] = load <16 x i8>, ptr [[ARG3]], align 1
; CHECK-NEXT:    [[TMP14:%.*]] = zext <16 x i8> [[TMP12]] to <16 x i32>
; CHECK-NEXT:    [[TMP15:%.*]] = zext <16 x i8> [[TMP13]] to <16 x i32>
; CHECK-NEXT:    [[TMP16:%.*]] = sub nsw <16 x i32> [[TMP14]], [[TMP15]]
; CHECK-NEXT:    [[TMP19:%.*]] = tail call <16 x i32> @llvm.abs.v16i32(<16 x i32> [[TMP16]], i1 true)
; CHECK-NEXT:    [[TMP20:%.*]] = add nuw nsw <16 x i32> [[TMP11]], [[TMP19]]
; CHECK-NEXT:    [[TMP29:%.*]] = tail call i32 @llvm.vector.reduce.add.v16i32(<16 x i32> [[TMP20]])
; CHECK-NEXT:    ret i32 [[TMP29]]
;
bb:
  %i = load <16 x i8>, ptr %arg, align 1
  %i4 = load <16 x i8>, ptr %arg1, align 1
  %i5 = zext <16 x i8> %i to <16 x i32>
  %i6 = zext <16 x i8> %i4 to <16 x i32>
  %i7 = sub nsw <16 x i32> %i5, %i6
  %i8 = icmp slt <16 x i32> %i7, zeroinitializer
  %i9 = sub nsw <16 x i32> zeroinitializer, %i7
  %i10 = select <16 x i1> %i8, <16 x i32> %i9, <16 x i32> %i7
  %i11 = add nuw nsw <16 x i32> %i10, <i32 1, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0>
  %i12 = load <16 x i8>, ptr %arg2, align 1
  %i13 = load <16 x i8>, ptr %arg3, align 1
  %i14 = zext <16 x i8> %i12 to <16 x i32>
  %i15 = zext <16 x i8> %i13 to <16 x i32>
  %i16 = sub nsw <16 x i32> %i14, %i15
  %i17 = icmp slt <16 x i32> %i16, zeroinitializer
  %i18 = sub nsw <16 x i32> zeroinitializer, %i16
  %i19 = select <16 x i1> %i17, <16 x i32> %i18, <16 x i32> %i16
  %i20 = add nuw nsw <16 x i32> %i19, %i11
  %i21 = shufflevector <16 x i32> %i20, <16 x i32> undef, <16 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
  %i22 = add <16 x i32> %i20, %i21
  %i23 = shufflevector <16 x i32> %i22, <16 x i32> undef, <16 x i32> <i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
  %i24 = add <16 x i32> %i22, %i23
  %i25 = shufflevector <16 x i32> %i24, <16 x i32> undef, <16 x i32> <i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
  %i26 = add <16 x i32> %i24, %i25
  %i27 = shufflevector <16 x i32> %i26, <16 x i32> undef, <16 x i32> <i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
  %i28 = add <16 x i32> %i26, %i27
  %i29 = extractelement <16 x i32> %i28, i64 0
  ret i32 %i29
}

define dso_local i32 @sad_double_reduction(ptr %arg, ptr %arg1, ptr %arg2, ptr %arg3) {
; CHECK-LABEL: define dso_local i32 @sad_double_reduction(
; CHECK-SAME: ptr readonly captures(none) [[ARG:%.*]], ptr readonly captures(none) [[ARG1:%.*]], ptr readonly captures(none) [[ARG2:%.*]], ptr readonly captures(none) [[ARG3:%.*]]) local_unnamed_addr #[[ATTR2]] {
; CHECK-NEXT:  [[BB:.*:]]
; CHECK-NEXT:    [[TMP:%.*]] = load <16 x i8>, ptr [[ARG]], align 1
; CHECK-NEXT:    [[TMP4:%.*]] = load <16 x i8>, ptr [[ARG1]], align 1
; CHECK-NEXT:    [[TMP5:%.*]] = zext <16 x i8> [[TMP]] to <16 x i32>
; CHECK-NEXT:    [[TMP6:%.*]] = zext <16 x i8> [[TMP4]] to <16 x i32>
; CHECK-NEXT:    [[TMP7:%.*]] = sub nsw <16 x i32> [[TMP5]], [[TMP6]]
; CHECK-NEXT:    [[TMP10:%.*]] = tail call <16 x i32> @llvm.abs.v16i32(<16 x i32> [[TMP7]], i1 true)
; CHECK-NEXT:    [[TMP11:%.*]] = load <16 x i8>, ptr [[ARG2]], align 1
; CHECK-NEXT:    [[TMP12:%.*]] = load <16 x i8>, ptr [[ARG3]], align 1
; CHECK-NEXT:    [[TMP13:%.*]] = zext <16 x i8> [[TMP11]] to <16 x i32>
; CHECK-NEXT:    [[TMP14:%.*]] = zext <16 x i8> [[TMP12]] to <16 x i32>
; CHECK-NEXT:    [[TMP15:%.*]] = sub nsw <16 x i32> [[TMP13]], [[TMP14]]
; CHECK-NEXT:    [[TMP18:%.*]] = tail call <16 x i32> @llvm.abs.v16i32(<16 x i32> [[TMP15]], i1 true)
; CHECK-NEXT:    [[TMP19:%.*]] = add nuw nsw <16 x i32> [[TMP18]], [[TMP10]]
; CHECK-NEXT:    [[TMP28:%.*]] = tail call i32 @llvm.vector.reduce.add.v16i32(<16 x i32> [[TMP19]])
; CHECK-NEXT:    ret i32 [[TMP28]]
;
bb:
  %i = load <16 x i8>, ptr %arg, align 1
  %i4 = load <16 x i8>, ptr %arg1, align 1
  %i5 = zext <16 x i8> %i to <16 x i32>
  %i6 = zext <16 x i8> %i4 to <16 x i32>
  %i7 = sub nsw <16 x i32> %i5, %i6
  %i8 = icmp slt <16 x i32> %i7, zeroinitializer
  %i9 = sub nsw <16 x i32> zeroinitializer, %i7
  %i10 = select <16 x i1> %i8, <16 x i32> %i9, <16 x i32> %i7
  %i11 = load <16 x i8>, ptr %arg2, align 1
  %i12 = load <16 x i8>, ptr %arg3, align 1
  %i13 = zext <16 x i8> %i11 to <16 x i32>
  %i14 = zext <16 x i8> %i12 to <16 x i32>
  %i15 = sub nsw <16 x i32> %i13, %i14
  %i16 = icmp slt <16 x i32> %i15, zeroinitializer
  %i17 = sub nsw <16 x i32> zeroinitializer, %i15
  %i18 = select <16 x i1> %i16, <16 x i32> %i17, <16 x i32> %i15
  %i19 = add nuw nsw <16 x i32> %i18, %i10
  %i20 = shufflevector <16 x i32> %i19, <16 x i32> undef, <16 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
  %i21 = add <16 x i32> %i19, %i20
  %i22 = shufflevector <16 x i32> %i21, <16 x i32> undef, <16 x i32> <i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
  %i23 = add <16 x i32> %i21, %i22
  %i24 = shufflevector <16 x i32> %i23, <16 x i32> undef, <16 x i32> <i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
  %i25 = add <16 x i32> %i23, %i24
  %i26 = shufflevector <16 x i32> %i25, <16 x i32> undef, <16 x i32> <i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
  %i27 = add <16 x i32> %i25, %i26
  %i28 = extractelement <16 x i32> %i27, i64 0
  ret i32 %i28
}

define dso_local i32 @sad_double_reduction_abs(ptr %arg, ptr %arg1, ptr %arg2, ptr %arg3) {
; CHECK-LABEL: define dso_local i32 @sad_double_reduction_abs(
; CHECK-SAME: ptr readonly captures(none) [[ARG:%.*]], ptr readonly captures(none) [[ARG1:%.*]], ptr readonly captures(none) [[ARG2:%.*]], ptr readonly captures(none) [[ARG3:%.*]]) local_unnamed_addr #[[ATTR2]] {
; CHECK-NEXT:  [[BB:.*:]]
; CHECK-NEXT:    [[TMP:%.*]] = load <16 x i8>, ptr [[ARG]], align 1
; CHECK-NEXT:    [[TMP4:%.*]] = load <16 x i8>, ptr [[ARG1]], align 1
; CHECK-NEXT:    [[TMP5:%.*]] = zext <16 x i8> [[TMP]] to <16 x i32>
; CHECK-NEXT:    [[TMP6:%.*]] = zext <16 x i8> [[TMP4]] to <16 x i32>
; CHECK-NEXT:    [[TMP7:%.*]] = sub nsw <16 x i32> [[TMP5]], [[TMP6]]
; CHECK-NEXT:    [[TMP10:%.*]] = tail call <16 x i32> @llvm.abs.v16i32(<16 x i32> [[TMP7]], i1 true)
; CHECK-NEXT:    [[TMP11:%.*]] = load <16 x i8>, ptr [[ARG2]], align 1
; CHECK-NEXT:    [[TMP12:%.*]] = load <16 x i8>, ptr [[ARG3]], align 1
; CHECK-NEXT:    [[TMP13:%.*]] = zext <16 x i8> [[TMP11]] to <16 x i32>
; CHECK-NEXT:    [[TMP14:%.*]] = zext <16 x i8> [[TMP12]] to <16 x i32>
; CHECK-NEXT:    [[TMP15:%.*]] = sub nsw <16 x i32> [[TMP13]], [[TMP14]]
; CHECK-NEXT:    [[TMP18:%.*]] = tail call <16 x i32> @llvm.abs.v16i32(<16 x i32> [[TMP15]], i1 true)
; CHECK-NEXT:    [[TMP19:%.*]] = add nuw nsw <16 x i32> [[TMP18]], [[TMP10]]
; CHECK-NEXT:    [[TMP28:%.*]] = tail call i32 @llvm.vector.reduce.add.v16i32(<16 x i32> [[TMP19]])
; CHECK-NEXT:    ret i32 [[TMP28]]
;
bb:
  %i = load <16 x i8>, ptr %arg, align 1
  %i4 = load <16 x i8>, ptr %arg1, align 1
  %i5 = zext <16 x i8> %i to <16 x i32>
  %i6 = zext <16 x i8> %i4 to <16 x i32>
  %i7 = sub nsw <16 x i32> %i5, %i6
  %i10 = call <16 x i32> @llvm.abs.v16i32(<16 x i32> %i7, i1 false)
  %i11 = load <16 x i8>, ptr %arg2, align 1
  %i12 = load <16 x i8>, ptr %arg3, align 1
  %i13 = zext <16 x i8> %i11 to <16 x i32>
  %i14 = zext <16 x i8> %i12 to <16 x i32>
  %i15 = sub nsw <16 x i32> %i13, %i14
  %i18 = call <16 x i32> @llvm.abs.v16i32(<16 x i32> %i15, i1 false)
  %i19 = add nuw nsw <16 x i32> %i18, %i10
  %i20 = shufflevector <16 x i32> %i19, <16 x i32> undef, <16 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
  %i21 = add <16 x i32> %i19, %i20
  %i22 = shufflevector <16 x i32> %i21, <16 x i32> undef, <16 x i32> <i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
  %i23 = add <16 x i32> %i21, %i22
  %i24 = shufflevector <16 x i32> %i23, <16 x i32> undef, <16 x i32> <i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
  %i25 = add <16 x i32> %i23, %i24
  %i26 = shufflevector <16 x i32> %i25, <16 x i32> undef, <16 x i32> <i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
  %i27 = add <16 x i32> %i25, %i26
  %i28 = extractelement <16 x i32> %i27, i64 0
  ret i32 %i28
}

define i32 @PR143456(ptr %p0, ptr %p1) {
; CHECK-LABEL: define range(i32 0, 256) i32 @PR143456(
; CHECK-SAME: ptr readonly captures(none) [[P0:%.*]], ptr readonly captures(none) [[P1:%.*]]) local_unnamed_addr #[[ATTR2]] {
; CHECK-NEXT:  [[BB:.*:]]
; CHECK-NEXT:    [[V0:%.*]] = load <8 x i8>, ptr [[P0]], align 1
; CHECK-NEXT:    [[V1:%.*]] = load <8 x i8>, ptr [[P1]], align 1
; CHECK-NEXT:    [[MAX:%.*]] = tail call <8 x i8> @llvm.umax.v8i8(<8 x i8> [[V0]], <8 x i8> [[V1]])
; CHECK-NEXT:    [[MIN:%.*]] = tail call <8 x i8> @llvm.umin.v8i8(<8 x i8> [[V0]], <8 x i8> [[V1]])
; CHECK-NEXT:    [[ABD:%.*]] = sub nuw <8 x i8> [[MAX]], [[MIN]]
; CHECK-NEXT:    [[ELT:%.*]] = tail call i8 @llvm.vector.reduce.add.v8i8(<8 x i8> [[ABD]])
; CHECK-NEXT:    [[RES:%.*]] = zext i8 [[ELT]] to i32
; CHECK-NEXT:    ret i32 [[RES]]
;
bb:
  %v0 = load <8 x i8>, ptr %p0, align 1
  %v1 = load <8 x i8>, ptr %p1, align 1
  %max = tail call <8 x i8> @llvm.umax.v8i8(<8 x i8> %v0, <8 x i8> %v1)
  %min = tail call <8 x i8> @llvm.umin.v8i8(<8 x i8> %v0, <8 x i8> %v1)
  %abd = sub nuw <8 x i8> %max, %min
  %rdx.shuf = shufflevector <8 x i8> %abd, <8 x i8> poison, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison>
  %bin.rdx = add <8 x i8> %abd, %rdx.shuf
  %rdx.shuf15 = shufflevector <8 x i8> %bin.rdx, <8 x i8> poison, <8 x i32> <i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
  %bin.rdx16 = add <8 x i8> %bin.rdx, %rdx.shuf15
  %rdx.shuf17 = shufflevector <8 x i8> %bin.rdx16, <8 x i8> poison, <8 x i32> <i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
  %bin.rdx18 = add <8 x i8> %bin.rdx16, %rdx.shuf17
  %elt = extractelement <8 x i8> %bin.rdx18, i32 0
  %res = zext i8 %elt to i32
  ret i32 %res
}

attributes #0 = { nounwind }
attributes #1 = { nofree nosync nounwind speculatable willreturn memory(none) }
attributes #2 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }
attributes #3 = { nocallback nocreateundeforpoison nofree nosync nounwind speculatable willreturn memory(none) }