llvm/test/Transforms/LoopVectorize/RISCV/tail-folding-reduction.ll

; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
; RUN: opt -passes=loop-vectorize \
; RUN: -tail-folding-policy=prefer-fold-tail \
; RUN: -mtriple=riscv64 -mattr=+v,+f -S < %s| FileCheck %s --check-prefix=IF-EVL

; RUN: opt -passes=loop-vectorize \
; RUN: -tail-folding-policy=dont-fold-tail \
; RUN: -mtriple=riscv64 -mattr=+v,+f -S < %s| FileCheck %s --check-prefix=NO-VP

define i32 @add(ptr %a, i64 %n, i32 %start) {
; IF-EVL-LABEL: @add(
; IF-EVL-NEXT:  entry:
; IF-EVL-NEXT:    br label [[VECTOR_PH:%.*]]
; IF-EVL:       vector.ph:
; IF-EVL-NEXT:    [[TMP9:%.*]] = insertelement <vscale x 4 x i32> zeroinitializer, i32 [[START:%.*]], i64 0
; IF-EVL-NEXT:    br label [[VECTOR_BODY:%.*]]
; IF-EVL:       vector.body:
; IF-EVL-NEXT:    [[EVL_BASED_IV:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_EVL_NEXT:%.*]], [[VECTOR_BODY]] ]
; IF-EVL-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x i32> [ [[TMP9]], [[VECTOR_PH]] ], [ [[TMP14:%.*]], [[VECTOR_BODY]] ]
; IF-EVL-NEXT:    [[AVL:%.*]] = phi i64 [ [[N:%.*]], [[VECTOR_PH]] ], [ [[AVL_NEXT:%.*]], [[VECTOR_BODY]] ]
; IF-EVL-NEXT:    [[TMP10:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[AVL]], i32 4, i1 true)
; IF-EVL-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[EVL_BASED_IV]]
; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP12]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP10]])
; IF-EVL-NEXT:    [[VP_OP:%.*]] = add <vscale x 4 x i32> [[VP_OP_LOAD]], [[VEC_PHI]]
; IF-EVL-NEXT:    [[TMP14]] = call <vscale x 4 x i32> @llvm.vp.merge.nxv4i32(<vscale x 4 x i1> splat (i1 true), <vscale x 4 x i32> [[VP_OP]], <vscale x 4 x i32> [[VEC_PHI]], i32 [[TMP10]])
; IF-EVL-NEXT:    [[TMP15:%.*]] = zext i32 [[TMP10]] to i64
; IF-EVL-NEXT:    [[INDEX_EVL_NEXT]] = add i64 [[TMP15]], [[EVL_BASED_IV]]
; IF-EVL-NEXT:    [[AVL_NEXT]] = sub nuw i64 [[AVL]], [[TMP15]]
; IF-EVL-NEXT:    [[TMP11:%.*]] = icmp eq i64 [[AVL_NEXT]], 0
; IF-EVL-NEXT:    br i1 [[TMP11]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
; IF-EVL:       middle.block:
; IF-EVL-NEXT:    [[TMP17:%.*]] = call i32 @llvm.vector.reduce.add.nxv4i32(<vscale x 4 x i32> [[TMP14]])
; IF-EVL-NEXT:    br label [[FOR_BODY:%.*]]
; IF-EVL:       for.end:
; IF-EVL-NEXT:    ret i32 [[TMP17]]
;
; NO-VP-LABEL: @add(
; NO-VP-NEXT:  entry:
; NO-VP-NEXT:    [[TMP0:%.*]] = call i64 @llvm.vscale.i64()
; NO-VP-NEXT:    [[TMP1:%.*]] = shl nuw i64 [[TMP0]], 2
; NO-VP-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], [[TMP1]]
; NO-VP-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; NO-VP:       vector.ph:
; NO-VP-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP1]]
; NO-VP-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
; NO-VP-NEXT:    [[TMP6:%.*]] = insertelement <vscale x 4 x i32> zeroinitializer, i32 [[START:%.*]], i64 0
; NO-VP-NEXT:    br label [[VECTOR_BODY:%.*]]
; NO-VP:       vector.body:
; NO-VP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; NO-VP-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x i32> [ [[TMP6]], [[VECTOR_PH]] ], [ [[TMP10:%.*]], [[VECTOR_BODY]] ]
; NO-VP-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[INDEX]]
; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP8]], align 4
; NO-VP-NEXT:    [[TMP10]] = add <vscale x 4 x i32> [[WIDE_LOAD]], [[VEC_PHI]]
; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP1]]
; NO-VP-NEXT:    [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; NO-VP-NEXT:    br i1 [[TMP11]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
; NO-VP:       middle.block:
; NO-VP-NEXT:    [[TMP12:%.*]] = call i32 @llvm.vector.reduce.add.nxv4i32(<vscale x 4 x i32> [[TMP10]])
; NO-VP-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
; NO-VP-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
; NO-VP:       scalar.ph:
; NO-VP-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
; NO-VP-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP12]], [[MIDDLE_BLOCK]] ], [ [[START]], [[ENTRY]] ]
; NO-VP-NEXT:    br label [[FOR_BODY:%.*]]
; NO-VP:       for.body:
; NO-VP-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ]
; NO-VP-NEXT:    [[RDX:%.*]] = phi i32 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[ADD:%.*]], [[FOR_BODY]] ]
; NO-VP-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[IV]]
; NO-VP-NEXT:    [[TMP13:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
; NO-VP-NEXT:    [[ADD]] = add nsw i32 [[TMP13]], [[RDX]]
; NO-VP-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
; NO-VP-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
; NO-VP-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP3:![0-9]+]]
; NO-VP:       for.end:
; NO-VP-NEXT:    [[ADD_LCSSA:%.*]] = phi i32 [ [[ADD]], [[FOR_BODY]] ], [ [[TMP12]], [[MIDDLE_BLOCK]] ]
; NO-VP-NEXT:    ret i32 [[ADD_LCSSA]]
;
entry:
  br label %for.body

for.body:
  %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]
  %rdx = phi i32 [ %start, %entry ], [ %add, %for.body ]
  %arrayidx = getelementptr inbounds i32, ptr %a, i64 %iv
  %0 = load i32, ptr %arrayidx, align 4
  %add = add nsw i32 %0, %rdx
  %iv.next = add nuw nsw i64 %iv, 1
  %exitcond.not = icmp eq i64 %iv.next, %n
  br i1 %exitcond.not, label %for.end, label %for.body, !llvm.loop !0

for.end:
  ret i32 %add
}

; not support mul reduction for scalable vector
define i32 @mul(ptr %a, i64 %n, i32 %start) {
; IF-EVL-LABEL: @mul(
; IF-EVL-NEXT:  entry:
; IF-EVL-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N_RND_UP:%.*]], 16
; IF-EVL-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[ENTRY:%.*]]
; IF-EVL:       vector.ph:
; IF-EVL-NEXT:    [[N_MOD_VF:%.*]] = and i64 [[N_RND_UP]], 15
; IF-EVL-NEXT:    [[N_VEC:%.*]] = sub i64 [[N_RND_UP]], [[N_MOD_VF]]
; IF-EVL-NEXT:    [[TMP9:%.*]] = insertelement <8 x i32> splat (i32 1), i32 [[START:%.*]], i64 0
; IF-EVL-NEXT:    br label [[FOR_BODY:%.*]]
; IF-EVL:       vector.body:
; IF-EVL-NEXT:    [[IV:%.*]] = phi i64 [ 0, [[ENTRY]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ]
; IF-EVL-NEXT:    [[VEC_PHI:%.*]] = phi <8 x i32> [ [[TMP9]], [[ENTRY]] ], [ [[TMP5:%.*]], [[FOR_BODY]] ]
; IF-EVL-NEXT:    [[VEC_PHI1:%.*]] = phi <8 x i32> [ splat (i32 1), [[ENTRY]] ], [ [[TMP4:%.*]], [[FOR_BODY]] ]
; IF-EVL-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[IV]]
; IF-EVL-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP3]], i64 8
; IF-EVL-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = load <8 x i32>, ptr [[TMP3]], align 4
; IF-EVL-NEXT:    [[WIDE_LOAD2:%.*]] = load <8 x i32>, ptr [[TMP2]], align 4
; IF-EVL-NEXT:    [[TMP5]] = mul <8 x i32> [[WIDE_MASKED_LOAD]], [[VEC_PHI]]
; IF-EVL-NEXT:    [[TMP4]] = mul <8 x i32> [[WIDE_LOAD2]], [[VEC_PHI1]]
; IF-EVL-NEXT:    [[IV_NEXT]] = add nuw i64 [[IV]], 16
; IF-EVL-NEXT:    [[TMP7:%.*]] = icmp eq i64 [[IV_NEXT]], [[N_VEC]]
; IF-EVL-NEXT:    br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[FOR_BODY]], !llvm.loop [[LOOP3:![0-9]+]]
; IF-EVL:       middle.block:
; IF-EVL-NEXT:    [[TMP6:%.*]] = mul <8 x i32> [[TMP4]], [[TMP5]]
; IF-EVL-NEXT:    [[TMP8:%.*]] = call i32 @llvm.vector.reduce.mul.v8i32(<8 x i32> [[TMP6]])
; IF-EVL-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N_RND_UP]], [[N_VEC]]
; IF-EVL-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
; IF-EVL:       scalar.ph:
; IF-EVL-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY1:%.*]] ]
; IF-EVL-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP8]], [[MIDDLE_BLOCK]] ], [ [[START]], [[ENTRY1]] ]
; IF-EVL-NEXT:    br label [[FOR_BODY1:%.*]]
; IF-EVL:       for.body:
; IF-EVL-NEXT:    [[IV1:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT1:%.*]], [[FOR_BODY1]] ]
; IF-EVL-NEXT:    [[RDX:%.*]] = phi i32 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[MUL:%.*]], [[FOR_BODY1]] ]
; IF-EVL-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[IV1]]
; IF-EVL-NEXT:    [[TMP0:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
; IF-EVL-NEXT:    [[MUL]] = mul nsw i32 [[TMP0]], [[RDX]]
; IF-EVL-NEXT:    [[IV_NEXT1]] = add nuw nsw i64 [[IV1]], 1
; IF-EVL-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT1]], [[N_RND_UP]]
; IF-EVL-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_END]], label [[FOR_BODY1]], !llvm.loop [[LOOP4:![0-9]+]]
; IF-EVL:       for.end:
; IF-EVL-NEXT:    [[MUL_LCSSA:%.*]] = phi i32 [ [[MUL]], [[FOR_BODY1]] ], [ [[TMP8]], [[MIDDLE_BLOCK]] ]
; IF-EVL-NEXT:    ret i32 [[MUL_LCSSA]]
;
; NO-VP-LABEL: @mul(
; NO-VP-NEXT:  entry:
; NO-VP-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], 16
; NO-VP-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; NO-VP:       vector.ph:
; NO-VP-NEXT:    [[N_MOD_VF:%.*]] = and i64 [[N]], 15
; NO-VP-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
; NO-VP-NEXT:    [[TMP0:%.*]] = insertelement <8 x i32> splat (i32 1), i32 [[START:%.*]], i64 0
; NO-VP-NEXT:    br label [[VECTOR_BODY:%.*]]
; NO-VP:       vector.body:
; NO-VP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; NO-VP-NEXT:    [[VEC_PHI:%.*]] = phi <8 x i32> [ [[TMP0]], [[VECTOR_PH]] ], [ [[TMP5:%.*]], [[VECTOR_BODY]] ]
; NO-VP-NEXT:    [[VEC_PHI1:%.*]] = phi <8 x i32> [ splat (i32 1), [[VECTOR_PH]] ], [ [[TMP6:%.*]], [[VECTOR_BODY]] ]
; NO-VP-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[INDEX]]
; NO-VP-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i32, ptr [[TMP2]], i64 8
; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i32>, ptr [[TMP2]], align 4
; NO-VP-NEXT:    [[WIDE_LOAD2:%.*]] = load <8 x i32>, ptr [[TMP4]], align 4
; NO-VP-NEXT:    [[TMP5]] = mul <8 x i32> [[WIDE_LOAD]], [[VEC_PHI]]
; NO-VP-NEXT:    [[TMP6]] = mul <8 x i32> [[WIDE_LOAD2]], [[VEC_PHI1]]
; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 16
; NO-VP-NEXT:    [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; NO-VP-NEXT:    br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
; NO-VP:       middle.block:
; NO-VP-NEXT:    [[BIN_RDX:%.*]] = mul <8 x i32> [[TMP6]], [[TMP5]]
; NO-VP-NEXT:    [[TMP8:%.*]] = call i32 @llvm.vector.reduce.mul.v8i32(<8 x i32> [[BIN_RDX]])
; NO-VP-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
; NO-VP-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
; NO-VP:       scalar.ph:
; NO-VP-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
; NO-VP-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP8]], [[MIDDLE_BLOCK]] ], [ [[START]], [[ENTRY]] ]
; NO-VP-NEXT:    br label [[FOR_BODY:%.*]]
; NO-VP:       for.body:
; NO-VP-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ]
; NO-VP-NEXT:    [[RDX:%.*]] = phi i32 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[MUL:%.*]], [[FOR_BODY]] ]
; NO-VP-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[IV]]
; NO-VP-NEXT:    [[TMP9:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
; NO-VP-NEXT:    [[MUL]] = mul nsw i32 [[TMP9]], [[RDX]]
; NO-VP-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
; NO-VP-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
; NO-VP-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]]
; NO-VP:       for.end:
; NO-VP-NEXT:    [[MUL_LCSSA:%.*]] = phi i32 [ [[MUL]], [[FOR_BODY]] ], [ [[TMP8]], [[MIDDLE_BLOCK]] ]
; NO-VP-NEXT:    ret i32 [[MUL_LCSSA]]
;
entry:
  br label %for.body

for.body:
  %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]
  %rdx = phi i32 [ %start, %entry ], [ %mul, %for.body ]
  %arrayidx = getelementptr inbounds i32, ptr %a, i64 %iv
  %0 = load i32, ptr %arrayidx, align 4
  %mul = mul nsw i32 %0, %rdx
  %iv.next = add nuw nsw i64 %iv, 1
  %exitcond.not = icmp eq i64 %iv.next, %n
  br i1 %exitcond.not, label %for.end, label %for.body, !llvm.loop !0

for.end:
  ret i32 %mul
}

define i32 @or(ptr %a, i64 %n, i32 %start) {
; IF-EVL-LABEL: @or(
; IF-EVL-NEXT:  entry:
; IF-EVL-NEXT:    br label [[VECTOR_PH:%.*]]
; IF-EVL:       vector.ph:
; IF-EVL-NEXT:    [[TMP9:%.*]] = insertelement <vscale x 4 x i32> zeroinitializer, i32 [[START:%.*]], i64 0
; IF-EVL-NEXT:    br label [[VECTOR_BODY:%.*]]
; IF-EVL:       vector.body:
; IF-EVL-NEXT:    [[EVL_BASED_IV:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_EVL_NEXT:%.*]], [[VECTOR_BODY]] ]
; IF-EVL-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x i32> [ [[TMP9]], [[VECTOR_PH]] ], [ [[TMP14:%.*]], [[VECTOR_BODY]] ]
; IF-EVL-NEXT:    [[AVL:%.*]] = phi i64 [ [[N:%.*]], [[VECTOR_PH]] ], [ [[AVL_NEXT:%.*]], [[VECTOR_BODY]] ]
; IF-EVL-NEXT:    [[TMP10:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[AVL]], i32 4, i1 true)
; IF-EVL-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[EVL_BASED_IV]]
; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP12]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP10]])
; IF-EVL-NEXT:    [[VP_OP:%.*]] = or <vscale x 4 x i32> [[VP_OP_LOAD]], [[VEC_PHI]]
; IF-EVL-NEXT:    [[TMP14]] = call <vscale x 4 x i32> @llvm.vp.merge.nxv4i32(<vscale x 4 x i1> splat (i1 true), <vscale x 4 x i32> [[VP_OP]], <vscale x 4 x i32> [[VEC_PHI]], i32 [[TMP10]])
; IF-EVL-NEXT:    [[TMP15:%.*]] = zext i32 [[TMP10]] to i64
; IF-EVL-NEXT:    [[INDEX_EVL_NEXT]] = add i64 [[TMP15]], [[EVL_BASED_IV]]
; IF-EVL-NEXT:    [[AVL_NEXT]] = sub nuw i64 [[AVL]], [[TMP15]]
; IF-EVL-NEXT:    [[TMP11:%.*]] = icmp eq i64 [[AVL_NEXT]], 0
; IF-EVL-NEXT:    br i1 [[TMP11]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]]
; IF-EVL:       middle.block:
; IF-EVL-NEXT:    [[TMP17:%.*]] = call i32 @llvm.vector.reduce.or.nxv4i32(<vscale x 4 x i32> [[TMP14]])
; IF-EVL-NEXT:    br label [[FOR_BODY:%.*]]
; IF-EVL:       for.end:
; IF-EVL-NEXT:    ret i32 [[TMP17]]
;
; NO-VP-LABEL: @or(
; NO-VP-NEXT:  entry:
; NO-VP-NEXT:    [[TMP0:%.*]] = call i64 @llvm.vscale.i64()
; NO-VP-NEXT:    [[TMP1:%.*]] = shl nuw i64 [[TMP0]], 2
; NO-VP-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], [[TMP1]]
; NO-VP-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; NO-VP:       vector.ph:
; NO-VP-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP1]]
; NO-VP-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
; NO-VP-NEXT:    [[TMP6:%.*]] = insertelement <vscale x 4 x i32> zeroinitializer, i32 [[START:%.*]], i64 0
; NO-VP-NEXT:    br label [[VECTOR_BODY:%.*]]
; NO-VP:       vector.body:
; NO-VP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; NO-VP-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x i32> [ [[TMP6]], [[VECTOR_PH]] ], [ [[TMP10:%.*]], [[VECTOR_BODY]] ]
; NO-VP-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[INDEX]]
; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP8]], align 4
; NO-VP-NEXT:    [[TMP10]] = or <vscale x 4 x i32> [[WIDE_LOAD]], [[VEC_PHI]]
; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP1]]
; NO-VP-NEXT:    [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; NO-VP-NEXT:    br i1 [[TMP11]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
; NO-VP:       middle.block:
; NO-VP-NEXT:    [[TMP12:%.*]] = call i32 @llvm.vector.reduce.or.nxv4i32(<vscale x 4 x i32> [[TMP10]])
; NO-VP-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
; NO-VP-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
; NO-VP:       scalar.ph:
; NO-VP-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
; NO-VP-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP12]], [[MIDDLE_BLOCK]] ], [ [[START]], [[ENTRY]] ]
; NO-VP-NEXT:    br label [[FOR_BODY:%.*]]
; NO-VP:       for.body:
; NO-VP-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ]
; NO-VP-NEXT:    [[RDX:%.*]] = phi i32 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[OR:%.*]], [[FOR_BODY]] ]
; NO-VP-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[IV]]
; NO-VP-NEXT:    [[TMP13:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
; NO-VP-NEXT:    [[OR]] = or i32 [[TMP13]], [[RDX]]
; NO-VP-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
; NO-VP-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
; NO-VP-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP7:![0-9]+]]
; NO-VP:       for.end:
; NO-VP-NEXT:    [[OR_LCSSA:%.*]] = phi i32 [ [[OR]], [[FOR_BODY]] ], [ [[TMP12]], [[MIDDLE_BLOCK]] ]
; NO-VP-NEXT:    ret i32 [[OR_LCSSA]]
;
entry:
  br label %for.body

for.body:
  %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]
  %rdx = phi i32 [ %start, %entry ], [ %or, %for.body ]
  %arrayidx = getelementptr inbounds i32, ptr %a, i64 %iv
  %0 = load i32, ptr %arrayidx, align 4
  %or = or i32 %0, %rdx
  %iv.next = add nuw nsw i64 %iv, 1
  %exitcond.not = icmp eq i64 %iv.next, %n
  br i1 %exitcond.not, label %for.end, label %for.body, !llvm.loop !0

for.end:
  ret i32 %or
}

define i32 @and(ptr %a, i64 %n, i32 %start) {
; IF-EVL-LABEL: @and(
; IF-EVL-NEXT:  entry:
; IF-EVL-NEXT:    br label [[VECTOR_PH:%.*]]
; IF-EVL:       vector.ph:
; IF-EVL-NEXT:    [[TMP9:%.*]] = insertelement <vscale x 4 x i32> splat (i32 -1), i32 [[START:%.*]], i64 0
; IF-EVL-NEXT:    br label [[VECTOR_BODY:%.*]]
; IF-EVL:       vector.body:
; IF-EVL-NEXT:    [[EVL_BASED_IV:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_EVL_NEXT:%.*]], [[VECTOR_BODY]] ]
; IF-EVL-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x i32> [ [[TMP9]], [[VECTOR_PH]] ], [ [[TMP14:%.*]], [[VECTOR_BODY]] ]
; IF-EVL-NEXT:    [[AVL:%.*]] = phi i64 [ [[N:%.*]], [[VECTOR_PH]] ], [ [[AVL_NEXT:%.*]], [[VECTOR_BODY]] ]
; IF-EVL-NEXT:    [[TMP10:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[AVL]], i32 4, i1 true)
; IF-EVL-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[EVL_BASED_IV]]
; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP12]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP10]])
; IF-EVL-NEXT:    [[VP_OP:%.*]] = and <vscale x 4 x i32> [[VP_OP_LOAD]], [[VEC_PHI]]
; IF-EVL-NEXT:    [[TMP14]] = call <vscale x 4 x i32> @llvm.vp.merge.nxv4i32(<vscale x 4 x i1> splat (i1 true), <vscale x 4 x i32> [[VP_OP]], <vscale x 4 x i32> [[VEC_PHI]], i32 [[TMP10]])
; IF-EVL-NEXT:    [[TMP15:%.*]] = zext i32 [[TMP10]] to i64
; IF-EVL-NEXT:    [[INDEX_EVL_NEXT]] = add i64 [[TMP15]], [[EVL_BASED_IV]]
; IF-EVL-NEXT:    [[AVL_NEXT]] = sub nuw i64 [[AVL]], [[TMP15]]
; IF-EVL-NEXT:    [[TMP11:%.*]] = icmp eq i64 [[AVL_NEXT]], 0
; IF-EVL-NEXT:    br i1 [[TMP11]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
; IF-EVL:       middle.block:
; IF-EVL-NEXT:    [[TMP17:%.*]] = call i32 @llvm.vector.reduce.and.nxv4i32(<vscale x 4 x i32> [[TMP14]])
; IF-EVL-NEXT:    br label [[FOR_BODY:%.*]]
; IF-EVL:       for.end:
; IF-EVL-NEXT:    ret i32 [[TMP17]]
;
; NO-VP-LABEL: @and(
; NO-VP-NEXT:  entry:
; NO-VP-NEXT:    [[TMP0:%.*]] = call i64 @llvm.vscale.i64()
; NO-VP-NEXT:    [[TMP1:%.*]] = shl nuw i64 [[TMP0]], 2
; NO-VP-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], [[TMP1]]
; NO-VP-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; NO-VP:       vector.ph:
; NO-VP-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP1]]
; NO-VP-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
; NO-VP-NEXT:    [[TMP6:%.*]] = insertelement <vscale x 4 x i32> splat (i32 -1), i32 [[START:%.*]], i64 0
; NO-VP-NEXT:    br label [[VECTOR_BODY:%.*]]
; NO-VP:       vector.body:
; NO-VP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; NO-VP-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x i32> [ [[TMP6]], [[VECTOR_PH]] ], [ [[TMP10:%.*]], [[VECTOR_BODY]] ]
; NO-VP-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[INDEX]]
; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP8]], align 4
; NO-VP-NEXT:    [[TMP10]] = and <vscale x 4 x i32> [[WIDE_LOAD]], [[VEC_PHI]]
; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP1]]
; NO-VP-NEXT:    [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; NO-VP-NEXT:    br i1 [[TMP11]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
; NO-VP:       middle.block:
; NO-VP-NEXT:    [[TMP12:%.*]] = call i32 @llvm.vector.reduce.and.nxv4i32(<vscale x 4 x i32> [[TMP10]])
; NO-VP-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
; NO-VP-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
; NO-VP:       scalar.ph:
; NO-VP-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
; NO-VP-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP12]], [[MIDDLE_BLOCK]] ], [ [[START]], [[ENTRY]] ]
; NO-VP-NEXT:    br label [[FOR_BODY:%.*]]
; NO-VP:       for.body:
; NO-VP-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ]
; NO-VP-NEXT:    [[RDX:%.*]] = phi i32 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[AND:%.*]], [[FOR_BODY]] ]
; NO-VP-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[IV]]
; NO-VP-NEXT:    [[TMP13:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
; NO-VP-NEXT:    [[AND]] = and i32 [[TMP13]], [[RDX]]
; NO-VP-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
; NO-VP-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
; NO-VP-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP9:![0-9]+]]
; NO-VP:       for.end:
; NO-VP-NEXT:    [[AND_LCSSA:%.*]] = phi i32 [ [[AND]], [[FOR_BODY]] ], [ [[TMP12]], [[MIDDLE_BLOCK]] ]
; NO-VP-NEXT:    ret i32 [[AND_LCSSA]]
;
entry:
  br label %for.body

for.body:
  %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]
  %rdx = phi i32 [ %start, %entry ], [ %and, %for.body ]
  %arrayidx = getelementptr inbounds i32, ptr %a, i64 %iv
  %0 = load i32, ptr %arrayidx, align 4
  %and = and i32 %0, %rdx
  %iv.next = add nuw nsw i64 %iv, 1
  %exitcond.not = icmp eq i64 %iv.next, %n
  br i1 %exitcond.not, label %for.end, label %for.body, !llvm.loop !0

for.end:
  ret i32 %and
}

define i32 @xor(ptr %a, i64 %n, i32 %start) {
; IF-EVL-LABEL: @xor(
; IF-EVL-NEXT:  entry:
; IF-EVL-NEXT:    br label [[VECTOR_PH:%.*]]
; IF-EVL:       vector.ph:
; IF-EVL-NEXT:    [[TMP9:%.*]] = insertelement <vscale x 4 x i32> zeroinitializer, i32 [[START:%.*]], i64 0
; IF-EVL-NEXT:    br label [[VECTOR_BODY:%.*]]
; IF-EVL:       vector.body:
; IF-EVL-NEXT:    [[EVL_BASED_IV:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_EVL_NEXT:%.*]], [[VECTOR_BODY]] ]
; IF-EVL-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x i32> [ [[TMP9]], [[VECTOR_PH]] ], [ [[TMP14:%.*]], [[VECTOR_BODY]] ]
; IF-EVL-NEXT:    [[AVL:%.*]] = phi i64 [ [[N:%.*]], [[VECTOR_PH]] ], [ [[AVL_NEXT:%.*]], [[VECTOR_BODY]] ]
; IF-EVL-NEXT:    [[TMP10:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[AVL]], i32 4, i1 true)
; IF-EVL-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[EVL_BASED_IV]]
; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP12]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP10]])
; IF-EVL-NEXT:    [[VP_OP:%.*]] = xor <vscale x 4 x i32> [[VP_OP_LOAD]], [[VEC_PHI]]
; IF-EVL-NEXT:    [[TMP14]] = call <vscale x 4 x i32> @llvm.vp.merge.nxv4i32(<vscale x 4 x i1> splat (i1 true), <vscale x 4 x i32> [[VP_OP]], <vscale x 4 x i32> [[VEC_PHI]], i32 [[TMP10]])
; IF-EVL-NEXT:    [[TMP15:%.*]] = zext i32 [[TMP10]] to i64
; IF-EVL-NEXT:    [[INDEX_EVL_NEXT]] = add i64 [[TMP15]], [[EVL_BASED_IV]]
; IF-EVL-NEXT:    [[AVL_NEXT]] = sub nuw i64 [[AVL]], [[TMP15]]
; IF-EVL-NEXT:    [[TMP11:%.*]] = icmp eq i64 [[AVL_NEXT]], 0
; IF-EVL-NEXT:    br i1 [[TMP11]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP7:![0-9]+]]
; IF-EVL:       middle.block:
; IF-EVL-NEXT:    [[TMP17:%.*]] = call i32 @llvm.vector.reduce.xor.nxv4i32(<vscale x 4 x i32> [[TMP14]])
; IF-EVL-NEXT:    br label [[FOR_BODY:%.*]]
; IF-EVL:       for.end:
; IF-EVL-NEXT:    ret i32 [[TMP17]]
;
; NO-VP-LABEL: @xor(
; NO-VP-NEXT:  entry:
; NO-VP-NEXT:    [[TMP0:%.*]] = call i64 @llvm.vscale.i64()
; NO-VP-NEXT:    [[TMP1:%.*]] = shl nuw i64 [[TMP0]], 2
; NO-VP-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], [[TMP1]]
; NO-VP-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; NO-VP:       vector.ph:
; NO-VP-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP1]]
; NO-VP-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
; NO-VP-NEXT:    [[TMP6:%.*]] = insertelement <vscale x 4 x i32> zeroinitializer, i32 [[START:%.*]], i64 0
; NO-VP-NEXT:    br label [[VECTOR_BODY:%.*]]
; NO-VP:       vector.body:
; NO-VP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; NO-VP-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x i32> [ [[TMP6]], [[VECTOR_PH]] ], [ [[TMP10:%.*]], [[VECTOR_BODY]] ]
; NO-VP-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[INDEX]]
; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP8]], align 4
; NO-VP-NEXT:    [[TMP10]] = xor <vscale x 4 x i32> [[WIDE_LOAD]], [[VEC_PHI]]
; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP1]]
; NO-VP-NEXT:    [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; NO-VP-NEXT:    br i1 [[TMP11]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]
; NO-VP:       middle.block:
; NO-VP-NEXT:    [[TMP12:%.*]] = call i32 @llvm.vector.reduce.xor.nxv4i32(<vscale x 4 x i32> [[TMP10]])
; NO-VP-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
; NO-VP-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
; NO-VP:       scalar.ph:
; NO-VP-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
; NO-VP-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP12]], [[MIDDLE_BLOCK]] ], [ [[START]], [[ENTRY]] ]
; NO-VP-NEXT:    br label [[FOR_BODY:%.*]]
; NO-VP:       for.body:
; NO-VP-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ]
; NO-VP-NEXT:    [[RDX:%.*]] = phi i32 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[XOR:%.*]], [[FOR_BODY]] ]
; NO-VP-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[IV]]
; NO-VP-NEXT:    [[TMP13:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
; NO-VP-NEXT:    [[XOR]] = xor i32 [[TMP13]], [[RDX]]
; NO-VP-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
; NO-VP-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
; NO-VP-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP11:![0-9]+]]
; NO-VP:       for.end:
; NO-VP-NEXT:    [[XOR_LCSSA:%.*]] = phi i32 [ [[XOR]], [[FOR_BODY]] ], [ [[TMP12]], [[MIDDLE_BLOCK]] ]
; NO-VP-NEXT:    ret i32 [[XOR_LCSSA]]
;
entry:
  br label %for.body

for.body:
  %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]
  %rdx = phi i32 [ %start, %entry ], [ %xor, %for.body ]
  %arrayidx = getelementptr inbounds i32, ptr %a, i64 %iv
  %0 = load i32, ptr %arrayidx, align 4
  %xor = xor i32 %0, %rdx
  %iv.next = add nuw nsw i64 %iv, 1
  %exitcond.not = icmp eq i64 %iv.next, %n
  br i1 %exitcond.not, label %for.end, label %for.body, !llvm.loop !0

for.end:
  ret i32 %xor
}

define i32 @smin(ptr %a, i64 %n, i32 %start) {
; IF-EVL-LABEL: @smin(
; IF-EVL-NEXT:  entry:
; IF-EVL-NEXT:    br label [[VECTOR_PH:%.*]]
; IF-EVL:       vector.ph:
; IF-EVL-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 4 x i32> poison, i32 [[START:%.*]], i64 0
; IF-EVL-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 4 x i32> [[BROADCAST_SPLATINSERT]], <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer
; IF-EVL-NEXT:    br label [[VECTOR_BODY:%.*]]
; IF-EVL:       vector.body:
; IF-EVL-NEXT:    [[EVL_BASED_IV:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_EVL_NEXT:%.*]], [[VECTOR_BODY]] ]
; IF-EVL-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x i32> [ [[BROADCAST_SPLAT]], [[VECTOR_PH]] ], [ [[TMP15:%.*]], [[VECTOR_BODY]] ]
; IF-EVL-NEXT:    [[AVL:%.*]] = phi i64 [ [[N:%.*]], [[VECTOR_PH]] ], [ [[AVL_NEXT:%.*]], [[VECTOR_BODY]] ]
; IF-EVL-NEXT:    [[TMP9:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[AVL]], i32 4, i1 true)
; IF-EVL-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[EVL_BASED_IV]]
; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP11]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP9]])
; IF-EVL-NEXT:    [[TMP14:%.*]] = call <vscale x 4 x i32> @llvm.smin.nxv4i32(<vscale x 4 x i32> [[VP_OP_LOAD]], <vscale x 4 x i32> [[VEC_PHI]])
; IF-EVL-NEXT:    [[TMP15]] = call <vscale x 4 x i32> @llvm.vp.merge.nxv4i32(<vscale x 4 x i1> splat (i1 true), <vscale x 4 x i32> [[TMP14]], <vscale x 4 x i32> [[VEC_PHI]], i32 [[TMP9]])
; IF-EVL-NEXT:    [[TMP16:%.*]] = zext i32 [[TMP9]] to i64
; IF-EVL-NEXT:    [[INDEX_EVL_NEXT]] = add i64 [[TMP16]], [[EVL_BASED_IV]]
; IF-EVL-NEXT:    [[AVL_NEXT]] = sub nuw i64 [[AVL]], [[TMP16]]
; IF-EVL-NEXT:    [[TMP12:%.*]] = icmp eq i64 [[AVL_NEXT]], 0
; IF-EVL-NEXT:    br i1 [[TMP12]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
; IF-EVL:       middle.block:
; IF-EVL-NEXT:    [[TMP18:%.*]] = call i32 @llvm.vector.reduce.smin.nxv4i32(<vscale x 4 x i32> [[TMP15]])
; IF-EVL-NEXT:    br label [[FOR_BODY:%.*]]
; IF-EVL:       for.end:
; IF-EVL-NEXT:    ret i32 [[TMP18]]
;
; NO-VP-LABEL: @smin(
; NO-VP-NEXT:  entry:
; NO-VP-NEXT:    [[TMP0:%.*]] = call i64 @llvm.vscale.i64()
; NO-VP-NEXT:    [[TMP1:%.*]] = shl nuw i64 [[TMP0]], 2
; NO-VP-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], [[TMP1]]
; NO-VP-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; NO-VP:       vector.ph:
; NO-VP-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP1]]
; NO-VP-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
; NO-VP-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 4 x i32> poison, i32 [[START:%.*]], i64 0
; NO-VP-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 4 x i32> [[BROADCAST_SPLATINSERT]], <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer
; NO-VP-NEXT:    br label [[VECTOR_BODY:%.*]]
; NO-VP:       vector.body:
; NO-VP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; NO-VP-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x i32> [ [[BROADCAST_SPLAT]], [[VECTOR_PH]] ], [ [[TMP10:%.*]], [[VECTOR_BODY]] ]
; NO-VP-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[INDEX]]
; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP7]], align 4
; NO-VP-NEXT:    [[TMP10]] = call <vscale x 4 x i32> @llvm.smin.nxv4i32(<vscale x 4 x i32> [[WIDE_LOAD]], <vscale x 4 x i32> [[VEC_PHI]])
; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP1]]
; NO-VP-NEXT:    [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; NO-VP-NEXT:    br i1 [[TMP11]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP12:![0-9]+]]
; NO-VP:       middle.block:
; NO-VP-NEXT:    [[TMP12:%.*]] = call i32 @llvm.vector.reduce.smin.nxv4i32(<vscale x 4 x i32> [[TMP10]])
; NO-VP-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
; NO-VP-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
; NO-VP:       scalar.ph:
; NO-VP-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
; NO-VP-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP12]], [[MIDDLE_BLOCK]] ], [ [[START]], [[ENTRY]] ]
; NO-VP-NEXT:    br label [[FOR_BODY:%.*]]
; NO-VP:       for.body:
; NO-VP-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ]
; NO-VP-NEXT:    [[RDX:%.*]] = phi i32 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[SMIN:%.*]], [[FOR_BODY]] ]
; NO-VP-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[IV]]
; NO-VP-NEXT:    [[TMP13:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
; NO-VP-NEXT:    [[SMIN]] = call i32 @llvm.smin.i32(i32 [[TMP13]], i32 [[RDX]])
; NO-VP-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
; NO-VP-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
; NO-VP-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP13:![0-9]+]]
; NO-VP:       for.end:
; NO-VP-NEXT:    [[SMIN_LCSSA:%.*]] = phi i32 [ [[SMIN]], [[FOR_BODY]] ], [ [[TMP12]], [[MIDDLE_BLOCK]] ]
; NO-VP-NEXT:    ret i32 [[SMIN_LCSSA]]
;
entry:
  br label %for.body

for.body:
  %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]
  %rdx = phi i32 [ %start, %entry ], [ %smin, %for.body ]
  %arrayidx = getelementptr inbounds i32, ptr %a, i64 %iv
  %0 = load i32, ptr %arrayidx, align 4
  %smin = call i32 @llvm.smin.i32(i32 %0, i32 %rdx)
  %iv.next = add nuw nsw i64 %iv, 1
  %exitcond.not = icmp eq i64 %iv.next, %n
  br i1 %exitcond.not, label %for.end, label %for.body, !llvm.loop !0

for.end:
  ret i32 %smin
}

define i32 @smax(ptr %a, i64 %n, i32 %start) {
; IF-EVL-LABEL: @smax(
; IF-EVL-NEXT:  entry:
; IF-EVL-NEXT:    br label [[VECTOR_PH:%.*]]
; IF-EVL:       vector.ph:
; IF-EVL-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 4 x i32> poison, i32 [[START:%.*]], i64 0
; IF-EVL-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 4 x i32> [[BROADCAST_SPLATINSERT]], <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer
; IF-EVL-NEXT:    br label [[VECTOR_BODY:%.*]]
; IF-EVL:       vector.body:
; IF-EVL-NEXT:    [[EVL_BASED_IV:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_EVL_NEXT:%.*]], [[VECTOR_BODY]] ]
; IF-EVL-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x i32> [ [[BROADCAST_SPLAT]], [[VECTOR_PH]] ], [ [[TMP15:%.*]], [[VECTOR_BODY]] ]
; IF-EVL-NEXT:    [[AVL:%.*]] = phi i64 [ [[N:%.*]], [[VECTOR_PH]] ], [ [[AVL_NEXT:%.*]], [[VECTOR_BODY]] ]
; IF-EVL-NEXT:    [[TMP9:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[AVL]], i32 4, i1 true)
; IF-EVL-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[EVL_BASED_IV]]
; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP11]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP9]])
; IF-EVL-NEXT:    [[TMP14:%.*]] = call <vscale x 4 x i32> @llvm.smax.nxv4i32(<vscale x 4 x i32> [[VP_OP_LOAD]], <vscale x 4 x i32> [[VEC_PHI]])
; IF-EVL-NEXT:    [[TMP15]] = call <vscale x 4 x i32> @llvm.vp.merge.nxv4i32(<vscale x 4 x i1> splat (i1 true), <vscale x 4 x i32> [[TMP14]], <vscale x 4 x i32> [[VEC_PHI]], i32 [[TMP9]])
; IF-EVL-NEXT:    [[TMP16:%.*]] = zext i32 [[TMP9]] to i64
; IF-EVL-NEXT:    [[INDEX_EVL_NEXT]] = add i64 [[TMP16]], [[EVL_BASED_IV]]
; IF-EVL-NEXT:    [[AVL_NEXT]] = sub nuw i64 [[AVL]], [[TMP16]]
; IF-EVL-NEXT:    [[TMP12:%.*]] = icmp eq i64 [[AVL_NEXT]], 0
; IF-EVL-NEXT:    br i1 [[TMP12]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP9:![0-9]+]]
; IF-EVL:       middle.block:
; IF-EVL-NEXT:    [[TMP18:%.*]] = call i32 @llvm.vector.reduce.smax.nxv4i32(<vscale x 4 x i32> [[TMP15]])
; IF-EVL-NEXT:    br label [[FOR_BODY:%.*]]
; IF-EVL:       for.end:
; IF-EVL-NEXT:    ret i32 [[TMP18]]
;
; NO-VP-LABEL: @smax(
; NO-VP-NEXT:  entry:
; NO-VP-NEXT:    [[TMP0:%.*]] = call i64 @llvm.vscale.i64()
; NO-VP-NEXT:    [[TMP1:%.*]] = shl nuw i64 [[TMP0]], 2
; NO-VP-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], [[TMP1]]
; NO-VP-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; NO-VP:       vector.ph:
; NO-VP-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP1]]
; NO-VP-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
; NO-VP-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 4 x i32> poison, i32 [[START:%.*]], i64 0
; NO-VP-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 4 x i32> [[BROADCAST_SPLATINSERT]], <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer
; NO-VP-NEXT:    br label [[VECTOR_BODY:%.*]]
; NO-VP:       vector.body:
; NO-VP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; NO-VP-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x i32> [ [[BROADCAST_SPLAT]], [[VECTOR_PH]] ], [ [[TMP10:%.*]], [[VECTOR_BODY]] ]
; NO-VP-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[INDEX]]
; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP7]], align 4
; NO-VP-NEXT:    [[TMP10]] = call <vscale x 4 x i32> @llvm.smax.nxv4i32(<vscale x 4 x i32> [[WIDE_LOAD]], <vscale x 4 x i32> [[VEC_PHI]])
; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP1]]
; NO-VP-NEXT:    [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; NO-VP-NEXT:    br i1 [[TMP11]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP14:![0-9]+]]
; NO-VP:       middle.block:
; NO-VP-NEXT:    [[TMP12:%.*]] = call i32 @llvm.vector.reduce.smax.nxv4i32(<vscale x 4 x i32> [[TMP10]])
; NO-VP-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
; NO-VP-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
; NO-VP:       scalar.ph:
; NO-VP-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
; NO-VP-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP12]], [[MIDDLE_BLOCK]] ], [ [[START]], [[ENTRY]] ]
; NO-VP-NEXT:    br label [[FOR_BODY:%.*]]
; NO-VP:       for.body:
; NO-VP-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ]
; NO-VP-NEXT:    [[RDX:%.*]] = phi i32 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[SMAX:%.*]], [[FOR_BODY]] ]
; NO-VP-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[IV]]
; NO-VP-NEXT:    [[TMP13:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
; NO-VP-NEXT:    [[SMAX]] = call i32 @llvm.smax.i32(i32 [[TMP13]], i32 [[RDX]])
; NO-VP-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
; NO-VP-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
; NO-VP-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP15:![0-9]+]]
; NO-VP:       for.end:
; NO-VP-NEXT:    [[SMAX_LCSSA:%.*]] = phi i32 [ [[SMAX]], [[FOR_BODY]] ], [ [[TMP12]], [[MIDDLE_BLOCK]] ]
; NO-VP-NEXT:    ret i32 [[SMAX_LCSSA]]
;
entry:
  br label %for.body

for.body:
  %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]
  %rdx = phi i32 [ %start, %entry ], [ %smax, %for.body ]
  %arrayidx = getelementptr inbounds i32, ptr %a, i64 %iv
  %0 = load i32, ptr %arrayidx, align 4
  %smax = call i32 @llvm.smax.i32(i32 %0, i32 %rdx)
  %iv.next = add nuw nsw i64 %iv, 1
  %exitcond.not = icmp eq i64 %iv.next, %n
  br i1 %exitcond.not, label %for.end, label %for.body, !llvm.loop !0

for.end:
  ret i32 %smax
}

define i32 @umin(ptr %a, i64 %n, i32 %start) {
; IF-EVL-LABEL: @umin(
; IF-EVL-NEXT:  entry:
; IF-EVL-NEXT:    br label [[VECTOR_PH:%.*]]
; IF-EVL:       vector.ph:
; IF-EVL-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 4 x i32> poison, i32 [[START:%.*]], i64 0
; IF-EVL-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 4 x i32> [[BROADCAST_SPLATINSERT]], <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer
; IF-EVL-NEXT:    br label [[VECTOR_BODY:%.*]]
; IF-EVL:       vector.body:
; IF-EVL-NEXT:    [[EVL_BASED_IV:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_EVL_NEXT:%.*]], [[VECTOR_BODY]] ]
; IF-EVL-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x i32> [ [[BROADCAST_SPLAT]], [[VECTOR_PH]] ], [ [[TMP15:%.*]], [[VECTOR_BODY]] ]
; IF-EVL-NEXT:    [[AVL:%.*]] = phi i64 [ [[N:%.*]], [[VECTOR_PH]] ], [ [[AVL_NEXT:%.*]], [[VECTOR_BODY]] ]
; IF-EVL-NEXT:    [[TMP9:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[AVL]], i32 4, i1 true)
; IF-EVL-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[EVL_BASED_IV]]
; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP11]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP9]])
; IF-EVL-NEXT:    [[TMP14:%.*]] = call <vscale x 4 x i32> @llvm.umin.nxv4i32(<vscale x 4 x i32> [[VP_OP_LOAD]], <vscale x 4 x i32> [[VEC_PHI]])
; IF-EVL-NEXT:    [[TMP15]] = call <vscale x 4 x i32> @llvm.vp.merge.nxv4i32(<vscale x 4 x i1> splat (i1 true), <vscale x 4 x i32> [[TMP14]], <vscale x 4 x i32> [[VEC_PHI]], i32 [[TMP9]])
; IF-EVL-NEXT:    [[TMP16:%.*]] = zext i32 [[TMP9]] to i64
; IF-EVL-NEXT:    [[INDEX_EVL_NEXT]] = add i64 [[TMP16]], [[EVL_BASED_IV]]
; IF-EVL-NEXT:    [[AVL_NEXT]] = sub nuw i64 [[AVL]], [[TMP16]]
; IF-EVL-NEXT:    [[TMP12:%.*]] = icmp eq i64 [[AVL_NEXT]], 0
; IF-EVL-NEXT:    br i1 [[TMP12]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]
; IF-EVL:       middle.block:
; IF-EVL-NEXT:    [[TMP18:%.*]] = call i32 @llvm.vector.reduce.umin.nxv4i32(<vscale x 4 x i32> [[TMP15]])
; IF-EVL-NEXT:    br label [[FOR_BODY:%.*]]
; IF-EVL:       for.end:
; IF-EVL-NEXT:    ret i32 [[TMP18]]
;
; NO-VP-LABEL: @umin(
; NO-VP-NEXT:  entry:
; NO-VP-NEXT:    [[TMP0:%.*]] = call i64 @llvm.vscale.i64()
; NO-VP-NEXT:    [[TMP1:%.*]] = shl nuw i64 [[TMP0]], 2
; NO-VP-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], [[TMP1]]
; NO-VP-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; NO-VP:       vector.ph:
; NO-VP-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP1]]
; NO-VP-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
; NO-VP-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 4 x i32> poison, i32 [[START:%.*]], i64 0
; NO-VP-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 4 x i32> [[BROADCAST_SPLATINSERT]], <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer
; NO-VP-NEXT:    br label [[VECTOR_BODY:%.*]]
; NO-VP:       vector.body:
; NO-VP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; NO-VP-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x i32> [ [[BROADCAST_SPLAT]], [[VECTOR_PH]] ], [ [[TMP10:%.*]], [[VECTOR_BODY]] ]
; NO-VP-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[INDEX]]
; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP7]], align 4
; NO-VP-NEXT:    [[TMP10]] = call <vscale x 4 x i32> @llvm.umin.nxv4i32(<vscale x 4 x i32> [[WIDE_LOAD]], <vscale x 4 x i32> [[VEC_PHI]])
; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP1]]
; NO-VP-NEXT:    [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; NO-VP-NEXT:    br i1 [[TMP11]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP16:![0-9]+]]
; NO-VP:       middle.block:
; NO-VP-NEXT:    [[TMP12:%.*]] = call i32 @llvm.vector.reduce.umin.nxv4i32(<vscale x 4 x i32> [[TMP10]])
; NO-VP-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
; NO-VP-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
; NO-VP:       scalar.ph:
; NO-VP-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
; NO-VP-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP12]], [[MIDDLE_BLOCK]] ], [ [[START]], [[ENTRY]] ]
; NO-VP-NEXT:    br label [[FOR_BODY:%.*]]
; NO-VP:       for.body:
; NO-VP-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ]
; NO-VP-NEXT:    [[RDX:%.*]] = phi i32 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[UMIN:%.*]], [[FOR_BODY]] ]
; NO-VP-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[IV]]
; NO-VP-NEXT:    [[TMP13:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
; NO-VP-NEXT:    [[UMIN]] = call i32 @llvm.umin.i32(i32 [[TMP13]], i32 [[RDX]])
; NO-VP-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
; NO-VP-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
; NO-VP-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP17:![0-9]+]]
; NO-VP:       for.end:
; NO-VP-NEXT:    [[UMIN_LCSSA:%.*]] = phi i32 [ [[UMIN]], [[FOR_BODY]] ], [ [[TMP12]], [[MIDDLE_BLOCK]] ]
; NO-VP-NEXT:    ret i32 [[UMIN_LCSSA]]
;
entry:
  br label %for.body

for.body:
  %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]
  %rdx = phi i32 [ %start, %entry ], [ %umin, %for.body ]
  %arrayidx = getelementptr inbounds i32, ptr %a, i64 %iv
  %0 = load i32, ptr %arrayidx, align 4
  %umin = call i32 @llvm.umin.i32(i32 %0, i32 %rdx)
  %iv.next = add nuw nsw i64 %iv, 1
  %exitcond.not = icmp eq i64 %iv.next, %n
  br i1 %exitcond.not, label %for.end, label %for.body, !llvm.loop !0

for.end:
  ret i32 %umin
}

define i32 @umax(ptr %a, i64 %n, i32 %start) {
; IF-EVL-LABEL: @umax(
; IF-EVL-NEXT:  entry:
; IF-EVL-NEXT:    br label [[VECTOR_PH:%.*]]
; IF-EVL:       vector.ph:
; IF-EVL-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 4 x i32> poison, i32 [[START:%.*]], i64 0
; IF-EVL-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 4 x i32> [[BROADCAST_SPLATINSERT]], <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer
; IF-EVL-NEXT:    br label [[VECTOR_BODY:%.*]]
; IF-EVL:       vector.body:
; IF-EVL-NEXT:    [[EVL_BASED_IV:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_EVL_NEXT:%.*]], [[VECTOR_BODY]] ]
; IF-EVL-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x i32> [ [[BROADCAST_SPLAT]], [[VECTOR_PH]] ], [ [[TMP15:%.*]], [[VECTOR_BODY]] ]
; IF-EVL-NEXT:    [[AVL:%.*]] = phi i64 [ [[N:%.*]], [[VECTOR_PH]] ], [ [[AVL_NEXT:%.*]], [[VECTOR_BODY]] ]
; IF-EVL-NEXT:    [[TMP9:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[AVL]], i32 4, i1 true)
; IF-EVL-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[EVL_BASED_IV]]
; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP11]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP9]])
; IF-EVL-NEXT:    [[TMP14:%.*]] = call <vscale x 4 x i32> @llvm.umax.nxv4i32(<vscale x 4 x i32> [[VP_OP_LOAD]], <vscale x 4 x i32> [[VEC_PHI]])
; IF-EVL-NEXT:    [[TMP15]] = call <vscale x 4 x i32> @llvm.vp.merge.nxv4i32(<vscale x 4 x i1> splat (i1 true), <vscale x 4 x i32> [[TMP14]], <vscale x 4 x i32> [[VEC_PHI]], i32 [[TMP9]])
; IF-EVL-NEXT:    [[TMP16:%.*]] = zext i32 [[TMP9]] to i64
; IF-EVL-NEXT:    [[INDEX_EVL_NEXT]] = add i64 [[TMP16]], [[EVL_BASED_IV]]
; IF-EVL-NEXT:    [[AVL_NEXT]] = sub nuw i64 [[AVL]], [[TMP16]]
; IF-EVL-NEXT:    [[TMP12:%.*]] = icmp eq i64 [[AVL_NEXT]], 0
; IF-EVL-NEXT:    br i1 [[TMP12]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP11:![0-9]+]]
; IF-EVL:       middle.block:
; IF-EVL-NEXT:    [[TMP18:%.*]] = call i32 @llvm.vector.reduce.umax.nxv4i32(<vscale x 4 x i32> [[TMP15]])
; IF-EVL-NEXT:    br label [[FOR_BODY:%.*]]
; IF-EVL:       for.end:
; IF-EVL-NEXT:    ret i32 [[TMP18]]
;
; NO-VP-LABEL: @umax(
; NO-VP-NEXT:  entry:
; NO-VP-NEXT:    [[TMP0:%.*]] = call i64 @llvm.vscale.i64()
; NO-VP-NEXT:    [[TMP1:%.*]] = shl nuw i64 [[TMP0]], 2
; NO-VP-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], [[TMP1]]
; NO-VP-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; NO-VP:       vector.ph:
; NO-VP-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP1]]
; NO-VP-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
; NO-VP-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 4 x i32> poison, i32 [[START:%.*]], i64 0
; NO-VP-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 4 x i32> [[BROADCAST_SPLATINSERT]], <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer
; NO-VP-NEXT:    br label [[VECTOR_BODY:%.*]]
; NO-VP:       vector.body:
; NO-VP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; NO-VP-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x i32> [ [[BROADCAST_SPLAT]], [[VECTOR_PH]] ], [ [[TMP10:%.*]], [[VECTOR_BODY]] ]
; NO-VP-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[INDEX]]
; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP7]], align 4
; NO-VP-NEXT:    [[TMP10]] = call <vscale x 4 x i32> @llvm.umax.nxv4i32(<vscale x 4 x i32> [[WIDE_LOAD]], <vscale x 4 x i32> [[VEC_PHI]])
; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP1]]
; NO-VP-NEXT:    [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; NO-VP-NEXT:    br i1 [[TMP11]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP18:![0-9]+]]
; NO-VP:       middle.block:
; NO-VP-NEXT:    [[TMP12:%.*]] = call i32 @llvm.vector.reduce.umax.nxv4i32(<vscale x 4 x i32> [[TMP10]])
; NO-VP-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
; NO-VP-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
; NO-VP:       scalar.ph:
; NO-VP-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
; NO-VP-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP12]], [[MIDDLE_BLOCK]] ], [ [[START]], [[ENTRY]] ]
; NO-VP-NEXT:    br label [[FOR_BODY:%.*]]
; NO-VP:       for.body:
; NO-VP-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ]
; NO-VP-NEXT:    [[RDX:%.*]] = phi i32 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[UMAX:%.*]], [[FOR_BODY]] ]
; NO-VP-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[IV]]
; NO-VP-NEXT:    [[TMP13:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
; NO-VP-NEXT:    [[UMAX]] = call i32 @llvm.umax.i32(i32 [[TMP13]], i32 [[RDX]])
; NO-VP-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
; NO-VP-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
; NO-VP-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP19:![0-9]+]]
; NO-VP:       for.end:
; NO-VP-NEXT:    [[UMAX_LCSSA:%.*]] = phi i32 [ [[UMAX]], [[FOR_BODY]] ], [ [[TMP12]], [[MIDDLE_BLOCK]] ]
; NO-VP-NEXT:    ret i32 [[UMAX_LCSSA]]
;
entry:
  br label %for.body

for.body:
  %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]
  %rdx = phi i32 [ %start, %entry ], [ %umax, %for.body ]
  %arrayidx = getelementptr inbounds i32, ptr %a, i64 %iv
  %0 = load i32, ptr %arrayidx, align 4
  %umax = call i32 @llvm.umax.i32(i32 %0, i32 %rdx)
  %iv.next = add nuw nsw i64 %iv, 1
  %exitcond.not = icmp eq i64 %iv.next, %n
  br i1 %exitcond.not, label %for.end, label %for.body, !llvm.loop !0

for.end:
  ret i32 %umax
}

define float @fadd(ptr %a, i64 %n, float %start) {
; IF-EVL-LABEL: @fadd(
; IF-EVL-NEXT:  entry:
; IF-EVL-NEXT:    br label [[VECTOR_PH:%.*]]
; IF-EVL:       vector.ph:
; IF-EVL-NEXT:    [[TMP9:%.*]] = insertelement <vscale x 4 x float> splat (float -0.000000e+00), float [[START:%.*]], i64 0
; IF-EVL-NEXT:    br label [[VECTOR_BODY:%.*]]
; IF-EVL:       vector.body:
; IF-EVL-NEXT:    [[EVL_BASED_IV:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_EVL_NEXT:%.*]], [[VECTOR_BODY]] ]
; IF-EVL-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x float> [ [[TMP9]], [[VECTOR_PH]] ], [ [[TMP14:%.*]], [[VECTOR_BODY]] ]
; IF-EVL-NEXT:    [[AVL:%.*]] = phi i64 [ [[N:%.*]], [[VECTOR_PH]] ], [ [[AVL_NEXT:%.*]], [[VECTOR_BODY]] ]
; IF-EVL-NEXT:    [[TMP10:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[AVL]], i32 4, i1 true)
; IF-EVL-NEXT:    [[TMP12:%.*]] = getelementptr inbounds float, ptr [[A:%.*]], i64 [[EVL_BASED_IV]]
; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x float> @llvm.vp.load.nxv4f32.p0(ptr align 4 [[TMP12]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP10]])
; IF-EVL-NEXT:    [[VP_OP:%.*]] = fadd reassoc <vscale x 4 x float> [[VP_OP_LOAD]], [[VEC_PHI]]
; IF-EVL-NEXT:    [[TMP14]] = call <vscale x 4 x float> @llvm.vp.merge.nxv4f32(<vscale x 4 x i1> splat (i1 true), <vscale x 4 x float> [[VP_OP]], <vscale x 4 x float> [[VEC_PHI]], i32 [[TMP10]])
; IF-EVL-NEXT:    [[TMP15:%.*]] = zext i32 [[TMP10]] to i64
; IF-EVL-NEXT:    [[INDEX_EVL_NEXT]] = add i64 [[TMP15]], [[EVL_BASED_IV]]
; IF-EVL-NEXT:    [[AVL_NEXT]] = sub nuw i64 [[AVL]], [[TMP15]]
; IF-EVL-NEXT:    [[TMP11:%.*]] = icmp eq i64 [[AVL_NEXT]], 0
; IF-EVL-NEXT:    br i1 [[TMP11]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP12:![0-9]+]]
; IF-EVL:       middle.block:
; IF-EVL-NEXT:    [[TMP17:%.*]] = call reassoc float @llvm.vector.reduce.fadd.nxv4f32(float -0.000000e+00, <vscale x 4 x float> [[TMP14]])
; IF-EVL-NEXT:    br label [[FOR_BODY:%.*]]
; IF-EVL:       for.end:
; IF-EVL-NEXT:    ret float [[TMP17]]
;
; NO-VP-LABEL: @fadd(
; NO-VP-NEXT:  entry:
; NO-VP-NEXT:    [[TMP0:%.*]] = call i64 @llvm.vscale.i64()
; NO-VP-NEXT:    [[TMP1:%.*]] = shl nuw i64 [[TMP0]], 2
; NO-VP-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], [[TMP1]]
; NO-VP-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; NO-VP:       vector.ph:
; NO-VP-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP1]]
; NO-VP-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
; NO-VP-NEXT:    [[TMP6:%.*]] = insertelement <vscale x 4 x float> splat (float -0.000000e+00), float [[START:%.*]], i64 0
; NO-VP-NEXT:    br label [[VECTOR_BODY:%.*]]
; NO-VP:       vector.body:
; NO-VP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; NO-VP-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x float> [ [[TMP6]], [[VECTOR_PH]] ], [ [[TMP10:%.*]], [[VECTOR_BODY]] ]
; NO-VP-NEXT:    [[TMP8:%.*]] = getelementptr inbounds float, ptr [[A:%.*]], i64 [[INDEX]]
; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[TMP8]], align 4
; NO-VP-NEXT:    [[TMP10]] = fadd reassoc <vscale x 4 x float> [[WIDE_LOAD]], [[VEC_PHI]]
; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP1]]
; NO-VP-NEXT:    [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; NO-VP-NEXT:    br i1 [[TMP11]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP20:![0-9]+]]
; NO-VP:       middle.block:
; NO-VP-NEXT:    [[TMP12:%.*]] = call reassoc float @llvm.vector.reduce.fadd.nxv4f32(float -0.000000e+00, <vscale x 4 x float> [[TMP10]])
; NO-VP-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
; NO-VP-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
; NO-VP:       scalar.ph:
; NO-VP-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
; NO-VP-NEXT:    [[BC_MERGE_RDX:%.*]] = phi float [ [[TMP12]], [[MIDDLE_BLOCK]] ], [ [[START]], [[ENTRY]] ]
; NO-VP-NEXT:    br label [[FOR_BODY:%.*]]
; NO-VP:       for.body:
; NO-VP-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ]
; NO-VP-NEXT:    [[RDX:%.*]] = phi float [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[ADD:%.*]], [[FOR_BODY]] ]
; NO-VP-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[IV]]
; NO-VP-NEXT:    [[TMP13:%.*]] = load float, ptr [[ARRAYIDX]], align 4
; NO-VP-NEXT:    [[ADD]] = fadd reassoc float [[TMP13]], [[RDX]]
; NO-VP-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
; NO-VP-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
; NO-VP-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP21:![0-9]+]]
; NO-VP:       for.end:
; NO-VP-NEXT:    [[ADD_LCSSA:%.*]] = phi float [ [[ADD]], [[FOR_BODY]] ], [ [[TMP12]], [[MIDDLE_BLOCK]] ]
; NO-VP-NEXT:    ret float [[ADD_LCSSA]]
;
entry:
  br label %for.body

for.body:
  %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]
  %rdx = phi float [ %start, %entry ], [ %add, %for.body ]
  %arrayidx = getelementptr inbounds float, ptr %a, i64 %iv
  %0 = load float, ptr %arrayidx, align 4
  %add = fadd reassoc float %0, %rdx
  %iv.next = add nuw nsw i64 %iv, 1
  %exitcond.not = icmp eq i64 %iv.next, %n
  br i1 %exitcond.not, label %for.end, label %for.body, !llvm.loop !0

for.end:
  ret float %add
}

; not support fmul reduction for scalable vector
define float @fmul(ptr %a, i64 %n, float %start) {
; IF-EVL-LABEL: @fmul(
; IF-EVL-NEXT:  entry:
; IF-EVL-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N_RND_UP:%.*]], 16
; IF-EVL-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[ENTRY:%.*]]
; IF-EVL:       vector.ph:
; IF-EVL-NEXT:    [[N_MOD_VF:%.*]] = and i64 [[N_RND_UP]], 15
; IF-EVL-NEXT:    [[N_VEC:%.*]] = sub i64 [[N_RND_UP]], [[N_MOD_VF]]
; IF-EVL-NEXT:    [[TMP9:%.*]] = insertelement <8 x float> splat (float 1.000000e+00), float [[START:%.*]], i64 0
; IF-EVL-NEXT:    br label [[FOR_BODY:%.*]]
; IF-EVL:       vector.body:
; IF-EVL-NEXT:    [[IV:%.*]] = phi i64 [ 0, [[ENTRY]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ]
; IF-EVL-NEXT:    [[VEC_PHI:%.*]] = phi <8 x float> [ [[TMP9]], [[ENTRY]] ], [ [[TMP5:%.*]], [[FOR_BODY]] ]
; IF-EVL-NEXT:    [[VEC_PHI1:%.*]] = phi <8 x float> [ splat (float 1.000000e+00), [[ENTRY]] ], [ [[TMP4:%.*]], [[FOR_BODY]] ]
; IF-EVL-NEXT:    [[TMP3:%.*]] = getelementptr inbounds float, ptr [[A:%.*]], i64 [[IV]]
; IF-EVL-NEXT:    [[TMP2:%.*]] = getelementptr inbounds float, ptr [[TMP3]], i64 8
; IF-EVL-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = load <8 x float>, ptr [[TMP3]], align 4
; IF-EVL-NEXT:    [[WIDE_LOAD2:%.*]] = load <8 x float>, ptr [[TMP2]], align 4
; IF-EVL-NEXT:    [[TMP5]] = fmul reassoc <8 x float> [[WIDE_MASKED_LOAD]], [[VEC_PHI]]
; IF-EVL-NEXT:    [[TMP4]] = fmul reassoc <8 x float> [[WIDE_LOAD2]], [[VEC_PHI1]]
; IF-EVL-NEXT:    [[IV_NEXT]] = add nuw i64 [[IV]], 16
; IF-EVL-NEXT:    [[TMP7:%.*]] = icmp eq i64 [[IV_NEXT]], [[N_VEC]]
; IF-EVL-NEXT:    br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[FOR_BODY]], !llvm.loop [[LOOP13:![0-9]+]]
; IF-EVL:       middle.block:
; IF-EVL-NEXT:    [[TMP6:%.*]] = fmul reassoc <8 x float> [[TMP4]], [[TMP5]]
; IF-EVL-NEXT:    [[TMP8:%.*]] = call reassoc float @llvm.vector.reduce.fmul.v8f32(float 1.000000e+00, <8 x float> [[TMP6]])
; IF-EVL-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N_RND_UP]], [[N_VEC]]
; IF-EVL-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
; IF-EVL:       scalar.ph:
; IF-EVL-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY1:%.*]] ]
; IF-EVL-NEXT:    [[BC_MERGE_RDX:%.*]] = phi float [ [[TMP8]], [[MIDDLE_BLOCK]] ], [ [[START]], [[ENTRY1]] ]
; IF-EVL-NEXT:    br label [[FOR_BODY1:%.*]]
; IF-EVL:       for.body:
; IF-EVL-NEXT:    [[IV1:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT1:%.*]], [[FOR_BODY1]] ]
; IF-EVL-NEXT:    [[RDX:%.*]] = phi float [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[MUL:%.*]], [[FOR_BODY1]] ]
; IF-EVL-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[IV1]]
; IF-EVL-NEXT:    [[TMP0:%.*]] = load float, ptr [[ARRAYIDX]], align 4
; IF-EVL-NEXT:    [[MUL]] = fmul reassoc float [[TMP0]], [[RDX]]
; IF-EVL-NEXT:    [[IV_NEXT1]] = add nuw nsw i64 [[IV1]], 1
; IF-EVL-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT1]], [[N_RND_UP]]
; IF-EVL-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_END]], label [[FOR_BODY1]], !llvm.loop [[LOOP14:![0-9]+]]
; IF-EVL:       for.end:
; IF-EVL-NEXT:    [[MUL_LCSSA:%.*]] = phi float [ [[MUL]], [[FOR_BODY1]] ], [ [[TMP8]], [[MIDDLE_BLOCK]] ]
; IF-EVL-NEXT:    ret float [[MUL_LCSSA]]
;
; NO-VP-LABEL: @fmul(
; NO-VP-NEXT:  entry:
; NO-VP-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], 16
; NO-VP-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; NO-VP:       vector.ph:
; NO-VP-NEXT:    [[N_MOD_VF:%.*]] = and i64 [[N]], 15
; NO-VP-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
; NO-VP-NEXT:    [[TMP0:%.*]] = insertelement <8 x float> splat (float 1.000000e+00), float [[START:%.*]], i64 0
; NO-VP-NEXT:    br label [[VECTOR_BODY:%.*]]
; NO-VP:       vector.body:
; NO-VP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; NO-VP-NEXT:    [[VEC_PHI:%.*]] = phi <8 x float> [ [[TMP0]], [[VECTOR_PH]] ], [ [[TMP5:%.*]], [[VECTOR_BODY]] ]
; NO-VP-NEXT:    [[VEC_PHI1:%.*]] = phi <8 x float> [ splat (float 1.000000e+00), [[VECTOR_PH]] ], [ [[TMP6:%.*]], [[VECTOR_BODY]] ]
; NO-VP-NEXT:    [[TMP2:%.*]] = getelementptr inbounds float, ptr [[A:%.*]], i64 [[INDEX]]
; NO-VP-NEXT:    [[TMP4:%.*]] = getelementptr inbounds float, ptr [[TMP2]], i64 8
; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x float>, ptr [[TMP2]], align 4
; NO-VP-NEXT:    [[WIDE_LOAD2:%.*]] = load <8 x float>, ptr [[TMP4]], align 4
; NO-VP-NEXT:    [[TMP5]] = fmul reassoc <8 x float> [[WIDE_LOAD]], [[VEC_PHI]]
; NO-VP-NEXT:    [[TMP6]] = fmul reassoc <8 x float> [[WIDE_LOAD2]], [[VEC_PHI1]]
; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 16
; NO-VP-NEXT:    [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; NO-VP-NEXT:    br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP22:![0-9]+]]
; NO-VP:       middle.block:
; NO-VP-NEXT:    [[BIN_RDX:%.*]] = fmul reassoc <8 x float> [[TMP6]], [[TMP5]]
; NO-VP-NEXT:    [[TMP8:%.*]] = call reassoc float @llvm.vector.reduce.fmul.v8f32(float 1.000000e+00, <8 x float> [[BIN_RDX]])
; NO-VP-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
; NO-VP-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
; NO-VP:       scalar.ph:
; NO-VP-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
; NO-VP-NEXT:    [[BC_MERGE_RDX:%.*]] = phi float [ [[TMP8]], [[MIDDLE_BLOCK]] ], [ [[START]], [[ENTRY]] ]
; NO-VP-NEXT:    br label [[FOR_BODY:%.*]]
; NO-VP:       for.body:
; NO-VP-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ]
; NO-VP-NEXT:    [[RDX:%.*]] = phi float [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[MUL:%.*]], [[FOR_BODY]] ]
; NO-VP-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[IV]]
; NO-VP-NEXT:    [[TMP9:%.*]] = load float, ptr [[ARRAYIDX]], align 4
; NO-VP-NEXT:    [[MUL]] = fmul reassoc float [[TMP9]], [[RDX]]
; NO-VP-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
; NO-VP-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
; NO-VP-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP23:![0-9]+]]
; NO-VP:       for.end:
; NO-VP-NEXT:    [[MUL_LCSSA:%.*]] = phi float [ [[MUL]], [[FOR_BODY]] ], [ [[TMP8]], [[MIDDLE_BLOCK]] ]
; NO-VP-NEXT:    ret float [[MUL_LCSSA]]
;
entry:
  br label %for.body

for.body:
  %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]
  %rdx = phi float [ %start, %entry ], [ %mul, %for.body ]
  %arrayidx = getelementptr inbounds float, ptr %a, i64 %iv
  %0 = load float, ptr %arrayidx, align 4
  %mul = fmul reassoc float %0, %rdx
  %iv.next = add nuw nsw i64 %iv, 1
  %exitcond.not = icmp eq i64 %iv.next, %n
  br i1 %exitcond.not, label %for.end, label %for.body, !llvm.loop !0

for.end:
  ret float %mul
}

define float @fmin(ptr %a, i64 %n, float %start) {
; IF-EVL-LABEL: @fmin(
; IF-EVL-NEXT:  entry:
; IF-EVL-NEXT:    br label [[VECTOR_PH:%.*]]
; IF-EVL:       vector.ph:
; IF-EVL-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 4 x float> poison, float [[START:%.*]], i64 0
; IF-EVL-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 4 x float> [[BROADCAST_SPLATINSERT]], <vscale x 4 x float> poison, <vscale x 4 x i32> zeroinitializer
; IF-EVL-NEXT:    br label [[VECTOR_BODY:%.*]]
; IF-EVL:       vector.body:
; IF-EVL-NEXT:    [[EVL_BASED_IV:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_EVL_NEXT:%.*]], [[VECTOR_BODY]] ]
; IF-EVL-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x float> [ [[BROADCAST_SPLAT]], [[VECTOR_PH]] ], [ [[TMP15:%.*]], [[VECTOR_BODY]] ]
; IF-EVL-NEXT:    [[AVL:%.*]] = phi i64 [ [[N:%.*]], [[VECTOR_PH]] ], [ [[AVL_NEXT:%.*]], [[VECTOR_BODY]] ]
; IF-EVL-NEXT:    [[TMP9:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[AVL]], i32 4, i1 true)
; IF-EVL-NEXT:    [[TMP11:%.*]] = getelementptr inbounds float, ptr [[A:%.*]], i64 [[EVL_BASED_IV]]
; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x float> @llvm.vp.load.nxv4f32.p0(ptr align 4 [[TMP11]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP9]])
; IF-EVL-NEXT:    [[TMP13:%.*]] = fcmp fast olt <vscale x 4 x float> [[VP_OP_LOAD]], [[VEC_PHI]]
; IF-EVL-NEXT:    [[TMP15]] = call <vscale x 4 x float> @llvm.vp.merge.nxv4f32(<vscale x 4 x i1> [[TMP13]], <vscale x 4 x float> [[VP_OP_LOAD]], <vscale x 4 x float> [[VEC_PHI]], i32 [[TMP9]])
; IF-EVL-NEXT:    [[TMP16:%.*]] = zext i32 [[TMP9]] to i64
; IF-EVL-NEXT:    [[INDEX_EVL_NEXT]] = add i64 [[TMP16]], [[EVL_BASED_IV]]
; IF-EVL-NEXT:    [[AVL_NEXT]] = sub nuw i64 [[AVL]], [[TMP16]]
; IF-EVL-NEXT:    [[TMP12:%.*]] = icmp eq i64 [[AVL_NEXT]], 0
; IF-EVL-NEXT:    br i1 [[TMP12]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP15:![0-9]+]]
; IF-EVL:       middle.block:
; IF-EVL-NEXT:    [[TMP18:%.*]] = call fast float @llvm.vector.reduce.fmin.nxv4f32(<vscale x 4 x float> [[TMP15]])
; IF-EVL-NEXT:    br label [[FOR_BODY:%.*]]
; IF-EVL:       for.end:
; IF-EVL-NEXT:    ret float [[TMP18]]
;
; NO-VP-LABEL: @fmin(
; NO-VP-NEXT:  entry:
; NO-VP-NEXT:    [[TMP0:%.*]] = call i64 @llvm.vscale.i64()
; NO-VP-NEXT:    [[TMP1:%.*]] = shl nuw i64 [[TMP0]], 2
; NO-VP-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], [[TMP1]]
; NO-VP-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; NO-VP:       vector.ph:
; NO-VP-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP1]]
; NO-VP-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
; NO-VP-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 4 x float> poison, float [[START:%.*]], i64 0
; NO-VP-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 4 x float> [[BROADCAST_SPLATINSERT]], <vscale x 4 x float> poison, <vscale x 4 x i32> zeroinitializer
; NO-VP-NEXT:    br label [[VECTOR_BODY:%.*]]
; NO-VP:       vector.body:
; NO-VP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; NO-VP-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x float> [ [[BROADCAST_SPLAT]], [[VECTOR_PH]] ], [ [[TMP10:%.*]], [[VECTOR_BODY]] ]
; NO-VP-NEXT:    [[TMP7:%.*]] = getelementptr inbounds float, ptr [[A:%.*]], i64 [[INDEX]]
; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[TMP7]], align 4
; NO-VP-NEXT:    [[TMP9:%.*]] = fcmp fast olt <vscale x 4 x float> [[WIDE_LOAD]], [[VEC_PHI]]
; NO-VP-NEXT:    [[TMP10]] = select nnan nsz <vscale x 4 x i1> [[TMP9]], <vscale x 4 x float> [[WIDE_LOAD]], <vscale x 4 x float> [[VEC_PHI]]
; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP1]]
; NO-VP-NEXT:    [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; NO-VP-NEXT:    br i1 [[TMP11]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP24:![0-9]+]]
; NO-VP:       middle.block:
; NO-VP-NEXT:    [[TMP12:%.*]] = call fast float @llvm.vector.reduce.fmin.nxv4f32(<vscale x 4 x float> [[TMP10]])
; NO-VP-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
; NO-VP-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
; NO-VP:       scalar.ph:
; NO-VP-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
; NO-VP-NEXT:    [[BC_MERGE_RDX:%.*]] = phi float [ [[TMP12]], [[MIDDLE_BLOCK]] ], [ [[START]], [[ENTRY]] ]
; NO-VP-NEXT:    br label [[FOR_BODY:%.*]]
; NO-VP:       for.body:
; NO-VP-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ]
; NO-VP-NEXT:    [[RDX:%.*]] = phi float [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[MIN:%.*]], [[FOR_BODY]] ]
; NO-VP-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[IV]]
; NO-VP-NEXT:    [[TMP13:%.*]] = load float, ptr [[ARRAYIDX]], align 4
; NO-VP-NEXT:    [[CMP:%.*]] = fcmp fast olt float [[TMP13]], [[RDX]]
; NO-VP-NEXT:    [[MIN]] = select nnan nsz i1 [[CMP]], float [[TMP13]], float [[RDX]]
; NO-VP-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
; NO-VP-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
; NO-VP-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP25:![0-9]+]]
; NO-VP:       for.end:
; NO-VP-NEXT:    [[MIN_LCSSA:%.*]] = phi float [ [[MIN]], [[FOR_BODY]] ], [ [[TMP12]], [[MIDDLE_BLOCK]] ]
; NO-VP-NEXT:    ret float [[MIN_LCSSA]]
;
entry:
  br label %for.body

for.body:
  %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]
  %rdx = phi float [ %start, %entry ], [ %min, %for.body ]
  %arrayidx = getelementptr inbounds float, ptr %a, i64 %iv
  %0 = load float, ptr %arrayidx, align 4
  %cmp = fcmp fast olt float %0, %rdx
  %min = select nnan nsz i1 %cmp, float %0, float %rdx
  %iv.next = add nuw nsw i64 %iv, 1
  %exitcond.not = icmp eq i64 %iv.next, %n
  br i1 %exitcond.not, label %for.end, label %for.body, !llvm.loop !0

for.end:
  ret float %min
}

define float @fmax(ptr %a, i64 %n, float %start) {
; IF-EVL-LABEL: @fmax(
; IF-EVL-NEXT:  entry:
; IF-EVL-NEXT:    br label [[VECTOR_PH:%.*]]
; IF-EVL:       vector.ph:
; IF-EVL-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 4 x float> poison, float [[START:%.*]], i64 0
; IF-EVL-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 4 x float> [[BROADCAST_SPLATINSERT]], <vscale x 4 x float> poison, <vscale x 4 x i32> zeroinitializer
; IF-EVL-NEXT:    br label [[VECTOR_BODY:%.*]]
; IF-EVL:       vector.body:
; IF-EVL-NEXT:    [[EVL_BASED_IV:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_EVL_NEXT:%.*]], [[VECTOR_BODY]] ]
; IF-EVL-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x float> [ [[BROADCAST_SPLAT]], [[VECTOR_PH]] ], [ [[TMP15:%.*]], [[VECTOR_BODY]] ]
; IF-EVL-NEXT:    [[AVL:%.*]] = phi i64 [ [[N:%.*]], [[VECTOR_PH]] ], [ [[AVL_NEXT:%.*]], [[VECTOR_BODY]] ]
; IF-EVL-NEXT:    [[TMP9:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[AVL]], i32 4, i1 true)
; IF-EVL-NEXT:    [[TMP11:%.*]] = getelementptr inbounds float, ptr [[A:%.*]], i64 [[EVL_BASED_IV]]
; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x float> @llvm.vp.load.nxv4f32.p0(ptr align 4 [[TMP11]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP9]])
; IF-EVL-NEXT:    [[TMP13:%.*]] = fcmp fast ogt <vscale x 4 x float> [[VP_OP_LOAD]], [[VEC_PHI]]
; IF-EVL-NEXT:    [[TMP15]] = call <vscale x 4 x float> @llvm.vp.merge.nxv4f32(<vscale x 4 x i1> [[TMP13]], <vscale x 4 x float> [[VP_OP_LOAD]], <vscale x 4 x float> [[VEC_PHI]], i32 [[TMP9]])
; IF-EVL-NEXT:    [[TMP16:%.*]] = zext i32 [[TMP9]] to i64
; IF-EVL-NEXT:    [[INDEX_EVL_NEXT]] = add i64 [[TMP16]], [[EVL_BASED_IV]]
; IF-EVL-NEXT:    [[AVL_NEXT]] = sub nuw i64 [[AVL]], [[TMP16]]
; IF-EVL-NEXT:    [[TMP12:%.*]] = icmp eq i64 [[AVL_NEXT]], 0
; IF-EVL-NEXT:    br i1 [[TMP12]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP16:![0-9]+]]
; IF-EVL:       middle.block:
; IF-EVL-NEXT:    [[TMP18:%.*]] = call fast float @llvm.vector.reduce.fmax.nxv4f32(<vscale x 4 x float> [[TMP15]])
; IF-EVL-NEXT:    br label [[FOR_BODY:%.*]]
; IF-EVL:       for.end:
; IF-EVL-NEXT:    ret float [[TMP18]]
;
; NO-VP-LABEL: @fmax(
; NO-VP-NEXT:  entry:
; NO-VP-NEXT:    [[TMP0:%.*]] = call i64 @llvm.vscale.i64()
; NO-VP-NEXT:    [[TMP1:%.*]] = shl nuw i64 [[TMP0]], 2
; NO-VP-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], [[TMP1]]
; NO-VP-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; NO-VP:       vector.ph:
; NO-VP-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP1]]
; NO-VP-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
; NO-VP-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 4 x float> poison, float [[START:%.*]], i64 0
; NO-VP-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 4 x float> [[BROADCAST_SPLATINSERT]], <vscale x 4 x float> poison, <vscale x 4 x i32> zeroinitializer
; NO-VP-NEXT:    br label [[VECTOR_BODY:%.*]]
; NO-VP:       vector.body:
; NO-VP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; NO-VP-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x float> [ [[BROADCAST_SPLAT]], [[VECTOR_PH]] ], [ [[TMP10:%.*]], [[VECTOR_BODY]] ]
; NO-VP-NEXT:    [[TMP7:%.*]] = getelementptr inbounds float, ptr [[A:%.*]], i64 [[INDEX]]
; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[TMP7]], align 4
; NO-VP-NEXT:    [[TMP9:%.*]] = fcmp fast ogt <vscale x 4 x float> [[WIDE_LOAD]], [[VEC_PHI]]
; NO-VP-NEXT:    [[TMP10]] = select nnan nsz <vscale x 4 x i1> [[TMP9]], <vscale x 4 x float> [[WIDE_LOAD]], <vscale x 4 x float> [[VEC_PHI]]
; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP1]]
; NO-VP-NEXT:    [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; NO-VP-NEXT:    br i1 [[TMP11]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP26:![0-9]+]]
; NO-VP:       middle.block:
; NO-VP-NEXT:    [[TMP12:%.*]] = call fast float @llvm.vector.reduce.fmax.nxv4f32(<vscale x 4 x float> [[TMP10]])
; NO-VP-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
; NO-VP-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
; NO-VP:       scalar.ph:
; NO-VP-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
; NO-VP-NEXT:    [[BC_MERGE_RDX:%.*]] = phi float [ [[TMP12]], [[MIDDLE_BLOCK]] ], [ [[START]], [[ENTRY]] ]
; NO-VP-NEXT:    br label [[FOR_BODY:%.*]]
; NO-VP:       for.body:
; NO-VP-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ]
; NO-VP-NEXT:    [[RDX:%.*]] = phi float [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[MAX:%.*]], [[FOR_BODY]] ]
; NO-VP-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[IV]]
; NO-VP-NEXT:    [[TMP13:%.*]] = load float, ptr [[ARRAYIDX]], align 4
; NO-VP-NEXT:    [[CMP:%.*]] = fcmp fast ogt float [[TMP13]], [[RDX]]
; NO-VP-NEXT:    [[MAX]] = select nnan nsz i1 [[CMP]], float [[TMP13]], float [[RDX]]
; NO-VP-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
; NO-VP-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
; NO-VP-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP27:![0-9]+]]
; NO-VP:       for.end:
; NO-VP-NEXT:    [[MAX_LCSSA:%.*]] = phi float [ [[MAX]], [[FOR_BODY]] ], [ [[TMP12]], [[MIDDLE_BLOCK]] ]
; NO-VP-NEXT:    ret float [[MAX_LCSSA]]
;
entry:
  br label %for.body

for.body:
  %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]
  %rdx = phi float [ %start, %entry ], [ %max, %for.body ]
  %arrayidx = getelementptr inbounds float, ptr %a, i64 %iv
  %0 = load float, ptr %arrayidx, align 4
  %cmp = fcmp fast ogt float %0, %rdx
  %max = select nnan nsz i1 %cmp, float %0, float %rdx
  %iv.next = add nuw nsw i64 %iv, 1
  %exitcond.not = icmp eq i64 %iv.next, %n
  br i1 %exitcond.not, label %for.end, label %for.body, !llvm.loop !0

for.end:
  ret float %max
}

define float @fminimum(ptr %a, i64 %n, float %start) {
; IF-EVL-LABEL: @fminimum(
; IF-EVL-NEXT:  entry:
; IF-EVL-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N_RND_UP:%.*]], 16
; IF-EVL-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[ENTRY:%.*]]
; IF-EVL:       vector.ph:
; IF-EVL-NEXT:    [[N_MOD_VF:%.*]] = and i64 [[N_RND_UP]], 15
; IF-EVL-NEXT:    [[N_VEC:%.*]] = sub i64 [[N_RND_UP]], [[N_MOD_VF]]
; IF-EVL-NEXT:    [[BROADCAST_SPLATINSERT2:%.*]] = insertelement <8 x float> poison, float [[START:%.*]], i64 0
; IF-EVL-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <8 x float> [[BROADCAST_SPLATINSERT2]], <8 x float> poison, <8 x i32> zeroinitializer
; IF-EVL-NEXT:    br label [[FOR_BODY:%.*]]
; IF-EVL:       vector.body:
; IF-EVL-NEXT:    [[IV:%.*]] = phi i64 [ 0, [[ENTRY]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ]
; IF-EVL-NEXT:    [[VEC_PHI:%.*]] = phi <8 x float> [ [[BROADCAST_SPLAT]], [[ENTRY]] ], [ [[TMP4:%.*]], [[FOR_BODY]] ]
; IF-EVL-NEXT:    [[VEC_PHI1:%.*]] = phi <8 x float> [ [[BROADCAST_SPLAT]], [[ENTRY]] ], [ [[TMP3:%.*]], [[FOR_BODY]] ]
; IF-EVL-NEXT:    [[TMP2:%.*]] = getelementptr inbounds float, ptr [[A:%.*]], i64 [[IV]]
; IF-EVL-NEXT:    [[TMP1:%.*]] = getelementptr inbounds float, ptr [[TMP2]], i64 8
; IF-EVL-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = load <8 x float>, ptr [[TMP2]], align 4
; IF-EVL-NEXT:    [[WIDE_LOAD2:%.*]] = load <8 x float>, ptr [[TMP1]], align 4
; IF-EVL-NEXT:    [[TMP4]] = call <8 x float> @llvm.minimum.v8f32(<8 x float> [[VEC_PHI]], <8 x float> [[WIDE_MASKED_LOAD]])
; IF-EVL-NEXT:    [[TMP3]] = call <8 x float> @llvm.minimum.v8f32(<8 x float> [[VEC_PHI1]], <8 x float> [[WIDE_LOAD2]])
; IF-EVL-NEXT:    [[IV_NEXT]] = add nuw i64 [[IV]], 16
; IF-EVL-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[IV_NEXT]], [[N_VEC]]
; IF-EVL-NEXT:    br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[FOR_BODY]], !llvm.loop [[LOOP17:![0-9]+]]
; IF-EVL:       middle.block:
; IF-EVL-NEXT:    [[TMP5:%.*]] = call <8 x float> @llvm.minimum.v8f32(<8 x float> [[TMP4]], <8 x float> [[TMP3]])
; IF-EVL-NEXT:    [[TMP7:%.*]] = call float @llvm.vector.reduce.fminimum.v8f32(<8 x float> [[TMP5]])
; IF-EVL-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N_RND_UP]], [[N_VEC]]
; IF-EVL-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
; IF-EVL:       scalar.ph:
; IF-EVL-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY1:%.*]] ]
; IF-EVL-NEXT:    [[BC_MERGE_RDX:%.*]] = phi float [ [[TMP7]], [[MIDDLE_BLOCK]] ], [ [[START]], [[ENTRY1]] ]
; IF-EVL-NEXT:    br label [[FOR_BODY1:%.*]]
; IF-EVL:       for.body:
; IF-EVL-NEXT:    [[IV1:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT1:%.*]], [[FOR_BODY1]] ]
; IF-EVL-NEXT:    [[RDX:%.*]] = phi float [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[MIN:%.*]], [[FOR_BODY1]] ]
; IF-EVL-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[IV1]]
; IF-EVL-NEXT:    [[TMP0:%.*]] = load float, ptr [[ARRAYIDX]], align 4
; IF-EVL-NEXT:    [[MIN]] = tail call float @llvm.minimum.f32(float [[RDX]], float [[TMP0]])
; IF-EVL-NEXT:    [[IV_NEXT1]] = add nuw nsw i64 [[IV1]], 1
; IF-EVL-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT1]], [[N_RND_UP]]
; IF-EVL-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_END]], label [[FOR_BODY1]], !llvm.loop [[LOOP18:![0-9]+]]
; IF-EVL:       for.end:
; IF-EVL-NEXT:    [[MIN_LCSSA:%.*]] = phi float [ [[MIN]], [[FOR_BODY1]] ], [ [[TMP7]], [[MIDDLE_BLOCK]] ]
; IF-EVL-NEXT:    ret float [[MIN_LCSSA]]
;
; NO-VP-LABEL: @fminimum(
; NO-VP-NEXT:  entry:
; NO-VP-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], 16
; NO-VP-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; NO-VP:       vector.ph:
; NO-VP-NEXT:    [[N_MOD_VF:%.*]] = and i64 [[N]], 15
; NO-VP-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
; NO-VP-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <8 x float> poison, float [[START:%.*]], i64 0
; NO-VP-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <8 x float> [[BROADCAST_SPLATINSERT]], <8 x float> poison, <8 x i32> zeroinitializer
; NO-VP-NEXT:    br label [[VECTOR_BODY:%.*]]
; NO-VP:       vector.body:
; NO-VP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; NO-VP-NEXT:    [[VEC_PHI:%.*]] = phi <8 x float> [ [[BROADCAST_SPLAT]], [[VECTOR_PH]] ], [ [[TMP4:%.*]], [[VECTOR_BODY]] ]
; NO-VP-NEXT:    [[VEC_PHI1:%.*]] = phi <8 x float> [ [[BROADCAST_SPLAT]], [[VECTOR_PH]] ], [ [[TMP5:%.*]], [[VECTOR_BODY]] ]
; NO-VP-NEXT:    [[TMP1:%.*]] = getelementptr inbounds float, ptr [[A:%.*]], i64 [[INDEX]]
; NO-VP-NEXT:    [[TMP3:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i64 8
; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x float>, ptr [[TMP1]], align 4
; NO-VP-NEXT:    [[WIDE_LOAD2:%.*]] = load <8 x float>, ptr [[TMP3]], align 4
; NO-VP-NEXT:    [[TMP4]] = call <8 x float> @llvm.minimum.v8f32(<8 x float> [[VEC_PHI]], <8 x float> [[WIDE_LOAD]])
; NO-VP-NEXT:    [[TMP5]] = call <8 x float> @llvm.minimum.v8f32(<8 x float> [[VEC_PHI1]], <8 x float> [[WIDE_LOAD2]])
; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 16
; NO-VP-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; NO-VP-NEXT:    br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP28:![0-9]+]]
; NO-VP:       middle.block:
; NO-VP-NEXT:    [[RDX_MINMAX:%.*]] = call <8 x float> @llvm.minimum.v8f32(<8 x float> [[TMP4]], <8 x float> [[TMP5]])
; NO-VP-NEXT:    [[TMP7:%.*]] = call float @llvm.vector.reduce.fminimum.v8f32(<8 x float> [[RDX_MINMAX]])
; NO-VP-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
; NO-VP-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
; NO-VP:       scalar.ph:
; NO-VP-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
; NO-VP-NEXT:    [[BC_MERGE_RDX:%.*]] = phi float [ [[TMP7]], [[MIDDLE_BLOCK]] ], [ [[START]], [[ENTRY]] ]
; NO-VP-NEXT:    br label [[FOR_BODY:%.*]]
; NO-VP:       for.body:
; NO-VP-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ]
; NO-VP-NEXT:    [[RDX:%.*]] = phi float [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[MIN:%.*]], [[FOR_BODY]] ]
; NO-VP-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[IV]]
; NO-VP-NEXT:    [[TMP8:%.*]] = load float, ptr [[ARRAYIDX]], align 4
; NO-VP-NEXT:    [[MIN]] = tail call float @llvm.minimum.f32(float [[RDX]], float [[TMP8]])
; NO-VP-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
; NO-VP-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
; NO-VP-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP29:![0-9]+]]
; NO-VP:       for.end:
; NO-VP-NEXT:    [[MIN_LCSSA:%.*]] = phi float [ [[MIN]], [[FOR_BODY]] ], [ [[TMP7]], [[MIDDLE_BLOCK]] ]
; NO-VP-NEXT:    ret float [[MIN_LCSSA]]
;
entry:
  br label %for.body

for.body:
  %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]
  %rdx = phi float [ %start, %entry ], [ %min, %for.body ]
  %arrayidx = getelementptr inbounds float, ptr %a, i64 %iv
  %0 = load float, ptr %arrayidx, align 4
  %min = tail call float @llvm.minimum.f32(float %rdx, float %0)
  %iv.next = add nuw nsw i64 %iv, 1
  %exitcond.not = icmp eq i64 %iv.next, %n
  br i1 %exitcond.not, label %for.end, label %for.body, !llvm.loop !0

for.end:
  ret float %min
}

define float @fmaximum(ptr %a, i64 %n, float %start) {
; IF-EVL-LABEL: @fmaximum(
; IF-EVL-NEXT:  entry:
; IF-EVL-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N_RND_UP:%.*]], 16
; IF-EVL-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[ENTRY:%.*]]
; IF-EVL:       vector.ph:
; IF-EVL-NEXT:    [[N_MOD_VF:%.*]] = and i64 [[N_RND_UP]], 15
; IF-EVL-NEXT:    [[N_VEC:%.*]] = sub i64 [[N_RND_UP]], [[N_MOD_VF]]
; IF-EVL-NEXT:    [[BROADCAST_SPLATINSERT2:%.*]] = insertelement <8 x float> poison, float [[START:%.*]], i64 0
; IF-EVL-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <8 x float> [[BROADCAST_SPLATINSERT2]], <8 x float> poison, <8 x i32> zeroinitializer
; IF-EVL-NEXT:    br label [[FOR_BODY:%.*]]
; IF-EVL:       vector.body:
; IF-EVL-NEXT:    [[IV:%.*]] = phi i64 [ 0, [[ENTRY]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ]
; IF-EVL-NEXT:    [[VEC_PHI:%.*]] = phi <8 x float> [ [[BROADCAST_SPLAT]], [[ENTRY]] ], [ [[TMP4:%.*]], [[FOR_BODY]] ]
; IF-EVL-NEXT:    [[VEC_PHI1:%.*]] = phi <8 x float> [ [[BROADCAST_SPLAT]], [[ENTRY]] ], [ [[TMP3:%.*]], [[FOR_BODY]] ]
; IF-EVL-NEXT:    [[TMP2:%.*]] = getelementptr inbounds float, ptr [[A:%.*]], i64 [[IV]]
; IF-EVL-NEXT:    [[TMP1:%.*]] = getelementptr inbounds float, ptr [[TMP2]], i64 8
; IF-EVL-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = load <8 x float>, ptr [[TMP2]], align 4
; IF-EVL-NEXT:    [[WIDE_LOAD2:%.*]] = load <8 x float>, ptr [[TMP1]], align 4
; IF-EVL-NEXT:    [[TMP4]] = call <8 x float> @llvm.maximum.v8f32(<8 x float> [[VEC_PHI]], <8 x float> [[WIDE_MASKED_LOAD]])
; IF-EVL-NEXT:    [[TMP3]] = call <8 x float> @llvm.maximum.v8f32(<8 x float> [[VEC_PHI1]], <8 x float> [[WIDE_LOAD2]])
; IF-EVL-NEXT:    [[IV_NEXT]] = add nuw i64 [[IV]], 16
; IF-EVL-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[IV_NEXT]], [[N_VEC]]
; IF-EVL-NEXT:    br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[FOR_BODY]], !llvm.loop [[LOOP19:![0-9]+]]
; IF-EVL:       middle.block:
; IF-EVL-NEXT:    [[TMP5:%.*]] = call <8 x float> @llvm.maximum.v8f32(<8 x float> [[TMP4]], <8 x float> [[TMP3]])
; IF-EVL-NEXT:    [[TMP7:%.*]] = call float @llvm.vector.reduce.fmaximum.v8f32(<8 x float> [[TMP5]])
; IF-EVL-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N_RND_UP]], [[N_VEC]]
; IF-EVL-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
; IF-EVL:       scalar.ph:
; IF-EVL-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY1:%.*]] ]
; IF-EVL-NEXT:    [[BC_MERGE_RDX:%.*]] = phi float [ [[TMP7]], [[MIDDLE_BLOCK]] ], [ [[START]], [[ENTRY1]] ]
; IF-EVL-NEXT:    br label [[FOR_BODY1:%.*]]
; IF-EVL:       for.body:
; IF-EVL-NEXT:    [[IV1:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT1:%.*]], [[FOR_BODY1]] ]
; IF-EVL-NEXT:    [[RDX:%.*]] = phi float [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[MAX:%.*]], [[FOR_BODY1]] ]
; IF-EVL-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[IV1]]
; IF-EVL-NEXT:    [[TMP0:%.*]] = load float, ptr [[ARRAYIDX]], align 4
; IF-EVL-NEXT:    [[MAX]] = tail call float @llvm.maximum.f32(float [[RDX]], float [[TMP0]])
; IF-EVL-NEXT:    [[IV_NEXT1]] = add nuw nsw i64 [[IV1]], 1
; IF-EVL-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT1]], [[N_RND_UP]]
; IF-EVL-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_END]], label [[FOR_BODY1]], !llvm.loop [[LOOP20:![0-9]+]]
; IF-EVL:       for.end:
; IF-EVL-NEXT:    [[MAX_LCSSA:%.*]] = phi float [ [[MAX]], [[FOR_BODY1]] ], [ [[TMP7]], [[MIDDLE_BLOCK]] ]
; IF-EVL-NEXT:    ret float [[MAX_LCSSA]]
;
; NO-VP-LABEL: @fmaximum(
; NO-VP-NEXT:  entry:
; NO-VP-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], 16
; NO-VP-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; NO-VP:       vector.ph:
; NO-VP-NEXT:    [[N_MOD_VF:%.*]] = and i64 [[N]], 15
; NO-VP-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
; NO-VP-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <8 x float> poison, float [[START:%.*]], i64 0
; NO-VP-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <8 x float> [[BROADCAST_SPLATINSERT]], <8 x float> poison, <8 x i32> zeroinitializer
; NO-VP-NEXT:    br label [[VECTOR_BODY:%.*]]
; NO-VP:       vector.body:
; NO-VP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; NO-VP-NEXT:    [[VEC_PHI:%.*]] = phi <8 x float> [ [[BROADCAST_SPLAT]], [[VECTOR_PH]] ], [ [[TMP4:%.*]], [[VECTOR_BODY]] ]
; NO-VP-NEXT:    [[VEC_PHI1:%.*]] = phi <8 x float> [ [[BROADCAST_SPLAT]], [[VECTOR_PH]] ], [ [[TMP5:%.*]], [[VECTOR_BODY]] ]
; NO-VP-NEXT:    [[TMP1:%.*]] = getelementptr inbounds float, ptr [[A:%.*]], i64 [[INDEX]]
; NO-VP-NEXT:    [[TMP3:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i64 8
; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x float>, ptr [[TMP1]], align 4
; NO-VP-NEXT:    [[WIDE_LOAD2:%.*]] = load <8 x float>, ptr [[TMP3]], align 4
; NO-VP-NEXT:    [[TMP4]] = call <8 x float> @llvm.maximum.v8f32(<8 x float> [[VEC_PHI]], <8 x float> [[WIDE_LOAD]])
; NO-VP-NEXT:    [[TMP5]] = call <8 x float> @llvm.maximum.v8f32(<8 x float> [[VEC_PHI1]], <8 x float> [[WIDE_LOAD2]])
; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 16
; NO-VP-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; NO-VP-NEXT:    br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP30:![0-9]+]]
; NO-VP:       middle.block:
; NO-VP-NEXT:    [[RDX_MINMAX:%.*]] = call <8 x float> @llvm.maximum.v8f32(<8 x float> [[TMP4]], <8 x float> [[TMP5]])
; NO-VP-NEXT:    [[TMP7:%.*]] = call float @llvm.vector.reduce.fmaximum.v8f32(<8 x float> [[RDX_MINMAX]])
; NO-VP-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
; NO-VP-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
; NO-VP:       scalar.ph:
; NO-VP-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
; NO-VP-NEXT:    [[BC_MERGE_RDX:%.*]] = phi float [ [[TMP7]], [[MIDDLE_BLOCK]] ], [ [[START]], [[ENTRY]] ]
; NO-VP-NEXT:    br label [[FOR_BODY:%.*]]
; NO-VP:       for.body:
; NO-VP-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ]
; NO-VP-NEXT:    [[RDX:%.*]] = phi float [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[MAX:%.*]], [[FOR_BODY]] ]
; NO-VP-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[IV]]
; NO-VP-NEXT:    [[TMP8:%.*]] = load float, ptr [[ARRAYIDX]], align 4
; NO-VP-NEXT:    [[MAX]] = tail call float @llvm.maximum.f32(float [[RDX]], float [[TMP8]])
; NO-VP-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
; NO-VP-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
; NO-VP-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP31:![0-9]+]]
; NO-VP:       for.end:
; NO-VP-NEXT:    [[MAX_LCSSA:%.*]] = phi float [ [[MAX]], [[FOR_BODY]] ], [ [[TMP7]], [[MIDDLE_BLOCK]] ]
; NO-VP-NEXT:    ret float [[MAX_LCSSA]]
;
entry:
  br label %for.body

for.body:
  %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]
  %rdx = phi float [ %start, %entry ], [ %max, %for.body ]
  %arrayidx = getelementptr inbounds float, ptr %a, i64 %iv
  %0 = load float, ptr %arrayidx, align 4
  %max = tail call float @llvm.maximum.f32(float %rdx, float %0)
  %iv.next = add nuw nsw i64 %iv, 1
  %exitcond.not = icmp eq i64 %iv.next, %n
  br i1 %exitcond.not, label %for.end, label %for.body, !llvm.loop !0

for.end:
  ret float %max
}

define float @fmuladd(ptr %a, ptr %b, i64 %n, float %start) {
; IF-EVL-LABEL: @fmuladd(
; IF-EVL-NEXT:  entry:
; IF-EVL-NEXT:    br label [[VECTOR_PH:%.*]]
; IF-EVL:       vector.ph:
; IF-EVL-NEXT:    [[TMP9:%.*]] = insertelement <vscale x 4 x float> splat (float -0.000000e+00), float [[START:%.*]], i64 0
; IF-EVL-NEXT:    br label [[VECTOR_BODY:%.*]]
; IF-EVL:       vector.body:
; IF-EVL-NEXT:    [[EVL_BASED_IV:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_EVL_NEXT:%.*]], [[VECTOR_BODY]] ]
; IF-EVL-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x float> [ [[TMP9]], [[VECTOR_PH]] ], [ [[TMP17:%.*]], [[VECTOR_BODY]] ]
; IF-EVL-NEXT:    [[AVL:%.*]] = phi i64 [ [[N:%.*]], [[VECTOR_PH]] ], [ [[AVL_NEXT:%.*]], [[VECTOR_BODY]] ]
; IF-EVL-NEXT:    [[TMP10:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[AVL]], i32 4, i1 true)
; IF-EVL-NEXT:    [[TMP12:%.*]] = getelementptr inbounds float, ptr [[A:%.*]], i64 [[EVL_BASED_IV]]
; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x float> @llvm.vp.load.nxv4f32.p0(ptr align 4 [[TMP12]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP10]])
; IF-EVL-NEXT:    [[TMP14:%.*]] = getelementptr inbounds float, ptr [[B:%.*]], i64 [[EVL_BASED_IV]]
; IF-EVL-NEXT:    [[VP_OP_LOAD1:%.*]] = call <vscale x 4 x float> @llvm.vp.load.nxv4f32.p0(ptr align 4 [[TMP14]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP10]])
; IF-EVL-NEXT:    [[TMP16:%.*]] = call reassoc <vscale x 4 x float> @llvm.fmuladd.nxv4f32(<vscale x 4 x float> [[VP_OP_LOAD]], <vscale x 4 x float> [[VP_OP_LOAD1]], <vscale x 4 x float> [[VEC_PHI]])
; IF-EVL-NEXT:    [[TMP17]] = call <vscale x 4 x float> @llvm.vp.merge.nxv4f32(<vscale x 4 x i1> splat (i1 true), <vscale x 4 x float> [[TMP16]], <vscale x 4 x float> [[VEC_PHI]], i32 [[TMP10]])
; IF-EVL-NEXT:    [[TMP18:%.*]] = zext i32 [[TMP10]] to i64
; IF-EVL-NEXT:    [[INDEX_EVL_NEXT]] = add i64 [[TMP18]], [[EVL_BASED_IV]]
; IF-EVL-NEXT:    [[AVL_NEXT]] = sub nuw i64 [[AVL]], [[TMP18]]
; IF-EVL-NEXT:    [[TMP13:%.*]] = icmp eq i64 [[AVL_NEXT]], 0
; IF-EVL-NEXT:    br i1 [[TMP13]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP21:![0-9]+]]
; IF-EVL:       middle.block:
; IF-EVL-NEXT:    [[TMP20:%.*]] = call reassoc float @llvm.vector.reduce.fadd.nxv4f32(float -0.000000e+00, <vscale x 4 x float> [[TMP17]])
; IF-EVL-NEXT:    br label [[FOR_BODY:%.*]]
; IF-EVL:       for.end:
; IF-EVL-NEXT:    ret float [[TMP20]]
;
; NO-VP-LABEL: @fmuladd(
; NO-VP-NEXT:  entry:
; NO-VP-NEXT:    [[TMP0:%.*]] = call i64 @llvm.vscale.i64()
; NO-VP-NEXT:    [[TMP1:%.*]] = shl nuw i64 [[TMP0]], 2
; NO-VP-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], [[TMP1]]
; NO-VP-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; NO-VP:       vector.ph:
; NO-VP-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP1]]
; NO-VP-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
; NO-VP-NEXT:    [[TMP6:%.*]] = insertelement <vscale x 4 x float> splat (float -0.000000e+00), float [[START:%.*]], i64 0
; NO-VP-NEXT:    br label [[VECTOR_BODY:%.*]]
; NO-VP:       vector.body:
; NO-VP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; NO-VP-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x float> [ [[TMP6]], [[VECTOR_PH]] ], [ [[TMP12:%.*]], [[VECTOR_BODY]] ]
; NO-VP-NEXT:    [[TMP8:%.*]] = getelementptr inbounds float, ptr [[A:%.*]], i64 [[INDEX]]
; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[TMP8]], align 4
; NO-VP-NEXT:    [[TMP10:%.*]] = getelementptr inbounds float, ptr [[B:%.*]], i64 [[INDEX]]
; NO-VP-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 4 x float>, ptr [[TMP10]], align 4
; NO-VP-NEXT:    [[TMP12]] = call reassoc <vscale x 4 x float> @llvm.fmuladd.nxv4f32(<vscale x 4 x float> [[WIDE_LOAD]], <vscale x 4 x float> [[WIDE_LOAD1]], <vscale x 4 x float> [[VEC_PHI]])
; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP1]]
; NO-VP-NEXT:    [[TMP13:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; NO-VP-NEXT:    br i1 [[TMP13]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP32:![0-9]+]]
; NO-VP:       middle.block:
; NO-VP-NEXT:    [[TMP14:%.*]] = call reassoc float @llvm.vector.reduce.fadd.nxv4f32(float -0.000000e+00, <vscale x 4 x float> [[TMP12]])
; NO-VP-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
; NO-VP-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
; NO-VP:       scalar.ph:
; NO-VP-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
; NO-VP-NEXT:    [[BC_MERGE_RDX:%.*]] = phi float [ [[TMP14]], [[MIDDLE_BLOCK]] ], [ [[START]], [[ENTRY]] ]
; NO-VP-NEXT:    br label [[FOR_BODY:%.*]]
; NO-VP:       for.body:
; NO-VP-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ]
; NO-VP-NEXT:    [[RDX:%.*]] = phi float [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[MULADD:%.*]], [[FOR_BODY]] ]
; NO-VP-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[IV]]
; NO-VP-NEXT:    [[TMP15:%.*]] = load float, ptr [[ARRAYIDX]], align 4
; NO-VP-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[IV]]
; NO-VP-NEXT:    [[TMP16:%.*]] = load float, ptr [[ARRAYIDX2]], align 4
; NO-VP-NEXT:    [[MULADD]] = tail call reassoc float @llvm.fmuladd.f32(float [[TMP15]], float [[TMP16]], float [[RDX]])
; NO-VP-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
; NO-VP-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
; NO-VP-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP33:![0-9]+]]
; NO-VP:       for.end:
; NO-VP-NEXT:    [[MULADD_LCSSA:%.*]] = phi float [ [[MULADD]], [[FOR_BODY]] ], [ [[TMP14]], [[MIDDLE_BLOCK]] ]
; NO-VP-NEXT:    ret float [[MULADD_LCSSA]]
;
entry:
  br label %for.body

for.body:
  %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]
  %rdx = phi float [ %start, %entry ], [ %muladd, %for.body ]
  %arrayidx = getelementptr inbounds float, ptr %a, i64 %iv
  %0 = load float, ptr %arrayidx, align 4
  %arrayidx2 = getelementptr inbounds float, ptr %b, i64 %iv
  %1 = load float, ptr %arrayidx2, align 4
  %muladd = tail call reassoc float @llvm.fmuladd.f32(float %0, float %1, float %rdx)
  %iv.next = add nuw nsw i64 %iv, 1
  %exitcond.not = icmp eq i64 %iv.next, %n
  br i1 %exitcond.not, label %for.end, label %for.body, !llvm.loop !0

for.end:
  ret float %muladd
}

define i32 @anyof_icmp(ptr %a, i64 %n, i32 %start, i32 %inv) {
; IF-EVL-LABEL: @anyof_icmp(
; IF-EVL-NEXT:  entry:
; IF-EVL-NEXT:    br label [[VECTOR_PH:%.*]]
; IF-EVL:       vector.ph:
; IF-EVL-NEXT:    br label [[VECTOR_BODY:%.*]]
; IF-EVL:       vector.body:
; IF-EVL-NEXT:    [[EVL_BASED_IV:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_EVL_NEXT:%.*]], [[VECTOR_BODY]] ]
; IF-EVL-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x i1> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP15:%.*]], [[VECTOR_BODY]] ]
; IF-EVL-NEXT:    [[AVL:%.*]] = phi i64 [ [[N:%.*]], [[VECTOR_PH]] ], [ [[AVL_NEXT:%.*]], [[VECTOR_BODY]] ]
; IF-EVL-NEXT:    [[TMP9:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[AVL]], i32 4, i1 true)
; IF-EVL-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[EVL_BASED_IV]]
; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP11]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP9]])
; IF-EVL-NEXT:    [[TMP13:%.*]] = icmp slt <vscale x 4 x i32> [[VP_OP_LOAD]], splat (i32 3)
; IF-EVL-NEXT:    [[TMP15]] = call <vscale x 4 x i1> @llvm.vp.merge.nxv4i1(<vscale x 4 x i1> [[TMP13]], <vscale x 4 x i1> splat (i1 true), <vscale x 4 x i1> [[VEC_PHI]], i32 [[TMP9]])
; IF-EVL-NEXT:    [[TMP16:%.*]] = zext i32 [[TMP9]] to i64
; IF-EVL-NEXT:    [[INDEX_EVL_NEXT]] = add i64 [[TMP16]], [[EVL_BASED_IV]]
; IF-EVL-NEXT:    [[AVL_NEXT]] = sub nuw i64 [[AVL]], [[TMP16]]
; IF-EVL-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[AVL_NEXT]], 0
; IF-EVL-NEXT:    br i1 [[TMP10]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP22:![0-9]+]]
; IF-EVL:       middle.block:
; IF-EVL-NEXT:    [[TMP18:%.*]] = call i1 @llvm.vector.reduce.or.nxv4i1(<vscale x 4 x i1> [[TMP15]])
; IF-EVL-NEXT:    [[TMP19:%.*]] = freeze i1 [[TMP18]]
; IF-EVL-NEXT:    [[RDX_SELECT:%.*]] = select i1 [[TMP19]], i32 [[INV:%.*]], i32 [[START:%.*]]
; IF-EVL-NEXT:    br label [[FOR_BODY:%.*]]
; IF-EVL:       for.end:
; IF-EVL-NEXT:    ret i32 [[RDX_SELECT]]
;
; NO-VP-LABEL: @anyof_icmp(
; NO-VP-NEXT:  entry:
; NO-VP-NEXT:    [[TMP0:%.*]] = call i64 @llvm.vscale.i64()
; NO-VP-NEXT:    [[TMP1:%.*]] = shl nuw i64 [[TMP0]], 2
; NO-VP-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], [[TMP1]]
; NO-VP-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; NO-VP:       vector.ph:
; NO-VP-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP1]]
; NO-VP-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
; NO-VP-NEXT:    br label [[VECTOR_BODY:%.*]]
; NO-VP:       vector.body:
; NO-VP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; NO-VP-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x i1> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP10:%.*]], [[VECTOR_BODY]] ]
; NO-VP-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[INDEX]]
; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP7]], align 4
; NO-VP-NEXT:    [[TMP9:%.*]] = icmp slt <vscale x 4 x i32> [[WIDE_LOAD]], splat (i32 3)
; NO-VP-NEXT:    [[TMP10]] = or <vscale x 4 x i1> [[VEC_PHI]], [[TMP9]]
; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP1]]
; NO-VP-NEXT:    [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; NO-VP-NEXT:    br i1 [[TMP11]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP34:![0-9]+]]
; NO-VP:       middle.block:
; NO-VP-NEXT:    [[TMP12:%.*]] = call i1 @llvm.vector.reduce.or.nxv4i1(<vscale x 4 x i1> [[TMP10]])
; NO-VP-NEXT:    [[TMP13:%.*]] = freeze i1 [[TMP12]]
; NO-VP-NEXT:    [[RDX_SELECT:%.*]] = select i1 [[TMP13]], i32 [[INV:%.*]], i32 [[START:%.*]]
; NO-VP-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
; NO-VP-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
; NO-VP:       scalar.ph:
; NO-VP-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
; NO-VP-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ [[RDX_SELECT]], [[MIDDLE_BLOCK]] ], [ [[START]], [[ENTRY]] ]
; NO-VP-NEXT:    br label [[FOR_BODY:%.*]]
; NO-VP:       for.body:
; NO-VP-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ]
; NO-VP-NEXT:    [[RDX:%.*]] = phi i32 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[ANYOF:%.*]], [[FOR_BODY]] ]
; NO-VP-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[IV]]
; NO-VP-NEXT:    [[TMP14:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
; NO-VP-NEXT:    [[CMP_I:%.*]] = icmp slt i32 [[TMP14]], 3
; NO-VP-NEXT:    [[ANYOF]] = select i1 [[CMP_I]], i32 [[INV]], i32 [[RDX]]
; NO-VP-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
; NO-VP-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
; NO-VP-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP35:![0-9]+]]
; NO-VP:       for.end:
; NO-VP-NEXT:    [[ANYOF_LCSSA:%.*]] = phi i32 [ [[ANYOF]], [[FOR_BODY]] ], [ [[RDX_SELECT]], [[MIDDLE_BLOCK]] ]
; NO-VP-NEXT:    ret i32 [[ANYOF_LCSSA]]
;
entry:
  br label %for.body

for.body:
  %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]
  %rdx = phi i32 [ %start, %entry ], [ %anyof, %for.body ]
  %arrayidx = getelementptr inbounds i32, ptr %a, i64 %iv
  %0 = load i32, ptr %arrayidx, align 4
  %cmp.i = icmp slt i32 %0, 3
  %anyof = select i1 %cmp.i, i32 %inv, i32 %rdx
  %iv.next = add nuw nsw i64 %iv, 1
  %exitcond.not = icmp eq i64 %iv.next, %n
  br i1 %exitcond.not, label %for.end, label %for.body, !llvm.loop !0

for.end:
  ret i32 %anyof
}

define i32 @anyof_fcmp(ptr %a, i64 %n, i32 %start, i32 %inv) {
; IF-EVL-LABEL: @anyof_fcmp(
; IF-EVL-NEXT:  entry:
; IF-EVL-NEXT:    br label [[VECTOR_PH:%.*]]
; IF-EVL:       vector.ph:
; IF-EVL-NEXT:    br label [[VECTOR_BODY:%.*]]
; IF-EVL:       vector.body:
; IF-EVL-NEXT:    [[EVL_BASED_IV:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_EVL_NEXT:%.*]], [[VECTOR_BODY]] ]
; IF-EVL-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x i1> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP15:%.*]], [[VECTOR_BODY]] ]
; IF-EVL-NEXT:    [[AVL:%.*]] = phi i64 [ [[N:%.*]], [[VECTOR_PH]] ], [ [[AVL_NEXT:%.*]], [[VECTOR_BODY]] ]
; IF-EVL-NEXT:    [[TMP9:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[AVL]], i32 4, i1 true)
; IF-EVL-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[EVL_BASED_IV]]
; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x float> @llvm.vp.load.nxv4f32.p0(ptr align 4 [[TMP11]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP9]])
; IF-EVL-NEXT:    [[TMP13:%.*]] = fcmp fast olt <vscale x 4 x float> [[VP_OP_LOAD]], splat (float 3.000000e+00)
; IF-EVL-NEXT:    [[TMP15]] = call <vscale x 4 x i1> @llvm.vp.merge.nxv4i1(<vscale x 4 x i1> [[TMP13]], <vscale x 4 x i1> splat (i1 true), <vscale x 4 x i1> [[VEC_PHI]], i32 [[TMP9]])
; IF-EVL-NEXT:    [[TMP16:%.*]] = zext i32 [[TMP9]] to i64
; IF-EVL-NEXT:    [[INDEX_EVL_NEXT]] = add i64 [[TMP16]], [[EVL_BASED_IV]]
; IF-EVL-NEXT:    [[AVL_NEXT]] = sub nuw i64 [[AVL]], [[TMP16]]
; IF-EVL-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[AVL_NEXT]], 0
; IF-EVL-NEXT:    br i1 [[TMP10]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP23:![0-9]+]]
; IF-EVL:       middle.block:
; IF-EVL-NEXT:    [[TMP18:%.*]] = call i1 @llvm.vector.reduce.or.nxv4i1(<vscale x 4 x i1> [[TMP15]])
; IF-EVL-NEXT:    [[TMP19:%.*]] = freeze i1 [[TMP18]]
; IF-EVL-NEXT:    [[RDX_SELECT:%.*]] = select i1 [[TMP19]], i32 [[INV:%.*]], i32 [[START:%.*]]
; IF-EVL-NEXT:    br label [[FOR_BODY:%.*]]
; IF-EVL:       for.end:
; IF-EVL-NEXT:    ret i32 [[RDX_SELECT]]
;
; NO-VP-LABEL: @anyof_fcmp(
; NO-VP-NEXT:  entry:
; NO-VP-NEXT:    [[TMP0:%.*]] = call i64 @llvm.vscale.i64()
; NO-VP-NEXT:    [[TMP1:%.*]] = shl nuw i64 [[TMP0]], 2
; NO-VP-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], [[TMP1]]
; NO-VP-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; NO-VP:       vector.ph:
; NO-VP-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP1]]
; NO-VP-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
; NO-VP-NEXT:    br label [[VECTOR_BODY:%.*]]
; NO-VP:       vector.body:
; NO-VP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; NO-VP-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x i1> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP10:%.*]], [[VECTOR_BODY]] ]
; NO-VP-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[INDEX]]
; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[TMP7]], align 4
; NO-VP-NEXT:    [[TMP9:%.*]] = fcmp fast olt <vscale x 4 x float> [[WIDE_LOAD]], splat (float 3.000000e+00)
; NO-VP-NEXT:    [[TMP10]] = or <vscale x 4 x i1> [[VEC_PHI]], [[TMP9]]
; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP1]]
; NO-VP-NEXT:    [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; NO-VP-NEXT:    br i1 [[TMP11]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP36:![0-9]+]]
; NO-VP:       middle.block:
; NO-VP-NEXT:    [[TMP12:%.*]] = call i1 @llvm.vector.reduce.or.nxv4i1(<vscale x 4 x i1> [[TMP10]])
; NO-VP-NEXT:    [[TMP13:%.*]] = freeze i1 [[TMP12]]
; NO-VP-NEXT:    [[RDX_SELECT:%.*]] = select i1 [[TMP13]], i32 [[INV:%.*]], i32 [[START:%.*]]
; NO-VP-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
; NO-VP-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
; NO-VP:       scalar.ph:
; NO-VP-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
; NO-VP-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ [[RDX_SELECT]], [[MIDDLE_BLOCK]] ], [ [[START]], [[ENTRY]] ]
; NO-VP-NEXT:    br label [[FOR_BODY:%.*]]
; NO-VP:       for.body:
; NO-VP-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ]
; NO-VP-NEXT:    [[RDX:%.*]] = phi i32 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[ANYOF:%.*]], [[FOR_BODY]] ]
; NO-VP-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[IV]]
; NO-VP-NEXT:    [[TMP14:%.*]] = load float, ptr [[ARRAYIDX]], align 4
; NO-VP-NEXT:    [[CMP_I:%.*]] = fcmp fast olt float [[TMP14]], 3.000000e+00
; NO-VP-NEXT:    [[ANYOF]] = select i1 [[CMP_I]], i32 [[INV]], i32 [[RDX]]
; NO-VP-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
; NO-VP-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
; NO-VP-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP37:![0-9]+]]
; NO-VP:       for.end:
; NO-VP-NEXT:    [[ANYOF_LCSSA:%.*]] = phi i32 [ [[ANYOF]], [[FOR_BODY]] ], [ [[RDX_SELECT]], [[MIDDLE_BLOCK]] ]
; NO-VP-NEXT:    ret i32 [[ANYOF_LCSSA]]
;
entry:
  br label %for.body

for.body:
  %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]
  %rdx = phi i32 [ %start, %entry ], [ %anyof, %for.body ]
  %arrayidx = getelementptr inbounds i32, ptr %a, i64 %iv
  %0 = load float, ptr %arrayidx, align 4
  %cmp.i = fcmp fast olt float %0, 3.0
  %anyof = select i1 %cmp.i, i32 %inv, i32 %rdx
  %iv.next = add nuw nsw i64 %iv, 1
  %exitcond.not = icmp eq i64 %iv.next, %n
  br i1 %exitcond.not, label %for.end, label %for.body, !llvm.loop !0

for.end:
  ret i32 %anyof
}


; Test for multiuse-reductions that are tailfolded using strict predicates.
define i64 @argmin_first_index(ptr %arr, i64 %n, i64 %start) {
; IF-EVL-LABEL: @argmin_first_index(
; IF-EVL-NEXT:  entry:
; IF-EVL-NEXT:    br label [[FOR_BODY:%.*]]
; IF-EVL:       vector.ph:
; IF-EVL-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 2 x i64> poison, i64 [[TMP0:%.*]], i64 0
; IF-EVL-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 2 x i64> [[BROADCAST_SPLATINSERT]], <vscale x 2 x i64> poison, <vscale x 2 x i32> zeroinitializer
; IF-EVL-NEXT:    [[TMP1:%.*]] = call <vscale x 2 x i64> @llvm.stepvector.nxv2i64()
; IF-EVL-NEXT:    br label [[VECTOR_BODY:%.*]]
; IF-EVL:       vector.body:
; IF-EVL-NEXT:    [[I_018:%.*]] = phi i64 [ 0, [[FOR_BODY]] ], [ [[CURRENT_ITERATION_NEXT:%.*]], [[VECTOR_BODY]] ]
; IF-EVL-NEXT:    [[VEC_IND:%.*]] = phi <vscale x 2 x i64> [ [[TMP1]], [[FOR_BODY]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
; IF-EVL-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 2 x i64> [ [[BROADCAST_SPLAT]], [[FOR_BODY]] ], [ [[TMP7:%.*]], [[VECTOR_BODY]] ]
; IF-EVL-NEXT:    [[VEC_PHI1:%.*]] = phi <vscale x 2 x i64> [ poison, [[FOR_BODY]] ], [ [[TMP8:%.*]], [[VECTOR_BODY]] ]
; IF-EVL-NEXT:    [[AVL:%.*]] = phi i64 [ [[N:%.*]], [[FOR_BODY]] ], [ [[AVL_NEXT:%.*]], [[VECTOR_BODY]] ]
; IF-EVL-NEXT:    [[TMP2:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[AVL]], i32 2, i1 true)
; IF-EVL-NEXT:    [[TMP3:%.*]] = zext i32 [[TMP2]] to i64
; IF-EVL-NEXT:    [[BROADCAST_SPLATINSERT2:%.*]] = insertelement <vscale x 2 x i64> poison, i64 [[TMP3]], i64 0
; IF-EVL-NEXT:    [[BROADCAST_SPLAT3:%.*]] = shufflevector <vscale x 2 x i64> [[BROADCAST_SPLATINSERT2]], <vscale x 2 x i64> poison, <vscale x 2 x i32> zeroinitializer
; IF-EVL-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds nuw [8 x i8], ptr [[ARR:%.*]], i64 [[I_018]]
; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 2 x i64> @llvm.vp.load.nxv2i64.p0(ptr align 8 [[ARRAYIDX2]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP2]])
; IF-EVL-NEXT:    [[TMP5:%.*]] = icmp slt <vscale x 2 x i64> [[VP_OP_LOAD]], [[VEC_PHI]]
; IF-EVL-NEXT:    [[TMP6:%.*]] = call <vscale x 2 x i64> @llvm.smin.nxv2i64(<vscale x 2 x i64> [[VP_OP_LOAD]], <vscale x 2 x i64> [[VEC_PHI]])
; IF-EVL-NEXT:    [[TMP7]] = call <vscale x 2 x i64> @llvm.vp.merge.nxv2i64(<vscale x 2 x i1> splat (i1 true), <vscale x 2 x i64> [[TMP6]], <vscale x 2 x i64> [[VEC_PHI]], i32 [[TMP2]])
; IF-EVL-NEXT:    [[TMP8]] = call <vscale x 2 x i64> @llvm.vp.merge.nxv2i64(<vscale x 2 x i1> [[TMP5]], <vscale x 2 x i64> [[VEC_IND]], <vscale x 2 x i64> [[VEC_PHI1]], i32 [[TMP2]])
; IF-EVL-NEXT:    [[CURRENT_ITERATION_NEXT]] = add i64 [[TMP3]], [[I_018]]
; IF-EVL-NEXT:    [[AVL_NEXT]] = sub nuw i64 [[AVL]], [[TMP3]]
; IF-EVL-NEXT:    [[VEC_IND_NEXT]] = add nuw nsw <vscale x 2 x i64> [[VEC_IND]], [[BROADCAST_SPLAT3]]
; IF-EVL-NEXT:    [[TMP9:%.*]] = icmp eq i64 [[AVL_NEXT]], 0
; IF-EVL-NEXT:    br i1 [[TMP9]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP24:![0-9]+]]
; IF-EVL:       middle.block:
; IF-EVL-NEXT:    [[TMP10:%.*]] = call i64 @llvm.vector.reduce.smin.nxv2i64(<vscale x 2 x i64> [[TMP7]])
; IF-EVL-NEXT:    [[BROADCAST_SPLATINSERT4:%.*]] = insertelement <vscale x 2 x i64> poison, i64 [[TMP10]], i64 0
; IF-EVL-NEXT:    [[BROADCAST_SPLAT5:%.*]] = shufflevector <vscale x 2 x i64> [[BROADCAST_SPLATINSERT4]], <vscale x 2 x i64> poison, <vscale x 2 x i32> zeroinitializer
; IF-EVL-NEXT:    [[TMP11:%.*]] = icmp eq <vscale x 2 x i64> [[TMP7]], [[BROADCAST_SPLAT5]]
; IF-EVL-NEXT:    [[TMP12:%.*]] = select <vscale x 2 x i1> [[TMP11]], <vscale x 2 x i64> [[TMP8]], <vscale x 2 x i64> splat (i64 -1)
; IF-EVL-NEXT:    [[TMP13:%.*]] = call i64 @llvm.vector.reduce.umin.nxv2i64(<vscale x 2 x i64> [[TMP12]])
; IF-EVL-NEXT:    [[TMP14:%.*]] = icmp eq i64 [[TMP10]], [[TMP0]]
; IF-EVL-NEXT:    [[SPEC_SELECT_LCSSA:%.*]] = select i1 [[TMP14]], i64 0, i64 [[TMP13]]
; IF-EVL-NEXT:    br label [[FOR_END:%.*]]
; IF-EVL:       for.end:
; IF-EVL-NEXT:    ret i64 [[SPEC_SELECT_LCSSA]]
;
; NO-VP-LABEL: @argmin_first_index(
; NO-VP-NEXT:  entry:
; NO-VP-NEXT:    [[TMP1:%.*]] = call i64 @llvm.vscale.i64()
; NO-VP-NEXT:    [[TMP2:%.*]] = shl nuw i64 [[TMP1]], 1
; NO-VP-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], [[TMP2]]
; NO-VP-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; NO-VP:       vector.ph:
; NO-VP-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP2]]
; NO-VP-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
; NO-VP-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 2 x i64> poison, i64 [[TMP0:%.*]], i64 0
; NO-VP-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 2 x i64> [[BROADCAST_SPLATINSERT]], <vscale x 2 x i64> poison, <vscale x 2 x i32> zeroinitializer
; NO-VP-NEXT:    [[TMP4:%.*]] = call <vscale x 2 x i64> @llvm.stepvector.nxv2i64()
; NO-VP-NEXT:    [[BROADCAST_SPLATINSERT1:%.*]] = insertelement <vscale x 2 x i64> poison, i64 [[TMP2]], i64 0
; NO-VP-NEXT:    [[BROADCAST_SPLAT2:%.*]] = shufflevector <vscale x 2 x i64> [[BROADCAST_SPLATINSERT1]], <vscale x 2 x i64> poison, <vscale x 2 x i32> zeroinitializer
; NO-VP-NEXT:    br label [[VECTOR_BODY:%.*]]
; NO-VP:       vector.body:
; NO-VP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; NO-VP-NEXT:    [[VEC_IND:%.*]] = phi <vscale x 2 x i64> [ [[TMP4]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
; NO-VP-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 2 x i64> [ [[BROADCAST_SPLAT]], [[VECTOR_PH]] ], [ [[TMP8:%.*]], [[VECTOR_BODY]] ]
; NO-VP-NEXT:    [[VEC_PHI3:%.*]] = phi <vscale x 2 x i64> [ poison, [[VECTOR_PH]] ], [ [[TMP7:%.*]], [[VECTOR_BODY]] ]
; NO-VP-NEXT:    [[TMP5:%.*]] = getelementptr inbounds nuw [8 x i8], ptr [[ARR:%.*]], i64 [[INDEX]]
; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 2 x i64>, ptr [[TMP5]], align 8
; NO-VP-NEXT:    [[TMP6:%.*]] = icmp slt <vscale x 2 x i64> [[WIDE_LOAD]], [[VEC_PHI]]
; NO-VP-NEXT:    [[TMP7]] = select <vscale x 2 x i1> [[TMP6]], <vscale x 2 x i64> [[VEC_IND]], <vscale x 2 x i64> [[VEC_PHI3]]
; NO-VP-NEXT:    [[TMP8]] = call <vscale x 2 x i64> @llvm.smin.nxv2i64(<vscale x 2 x i64> [[WIDE_LOAD]], <vscale x 2 x i64> [[VEC_PHI]])
; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP2]]
; NO-VP-NEXT:    [[VEC_IND_NEXT]] = add nuw nsw <vscale x 2 x i64> [[VEC_IND]], [[BROADCAST_SPLAT2]]
; NO-VP-NEXT:    [[TMP9:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; NO-VP-NEXT:    br i1 [[TMP9]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP38:![0-9]+]]
; NO-VP:       middle.block:
; NO-VP-NEXT:    [[TMP10:%.*]] = call i64 @llvm.vector.reduce.smin.nxv2i64(<vscale x 2 x i64> [[TMP8]])
; NO-VP-NEXT:    [[BROADCAST_SPLATINSERT4:%.*]] = insertelement <vscale x 2 x i64> poison, i64 [[TMP10]], i64 0
; NO-VP-NEXT:    [[BROADCAST_SPLAT5:%.*]] = shufflevector <vscale x 2 x i64> [[BROADCAST_SPLATINSERT4]], <vscale x 2 x i64> poison, <vscale x 2 x i32> zeroinitializer
; NO-VP-NEXT:    [[TMP11:%.*]] = icmp eq <vscale x 2 x i64> [[TMP8]], [[BROADCAST_SPLAT5]]
; NO-VP-NEXT:    [[TMP12:%.*]] = select <vscale x 2 x i1> [[TMP11]], <vscale x 2 x i64> [[TMP7]], <vscale x 2 x i64> splat (i64 -1)
; NO-VP-NEXT:    [[TMP13:%.*]] = call i64 @llvm.vector.reduce.umin.nxv2i64(<vscale x 2 x i64> [[TMP12]])
; NO-VP-NEXT:    [[TMP14:%.*]] = icmp eq i64 [[TMP10]], [[TMP0]]
; NO-VP-NEXT:    [[TMP15:%.*]] = select i1 [[TMP14]], i64 0, i64 [[TMP13]]
; NO-VP-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
; NO-VP-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
; NO-VP:       scalar.ph:
; NO-VP-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_BODY_PREHEADER:%.*]] ]
; NO-VP-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i64 [ [[TMP10]], [[MIDDLE_BLOCK]] ], [ [[TMP0]], [[FOR_BODY_PREHEADER]] ]
; NO-VP-NEXT:    [[BC_MERGE_RDX6:%.*]] = phi i64 [ [[TMP15]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_BODY_PREHEADER]] ]
; NO-VP-NEXT:    br label [[FOR_BODY:%.*]]
; NO-VP:       for.body:
; NO-VP-NEXT:    [[I_018:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INC:%.*]], [[FOR_BODY]] ]
; NO-VP-NEXT:    [[MIN_017:%.*]] = phi i64 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[SPEC_SELECT14:%.*]], [[FOR_BODY]] ]
; NO-VP-NEXT:    [[MINLOC_016:%.*]] = phi i64 [ [[BC_MERGE_RDX6]], [[SCALAR_PH]] ], [ [[SPEC_SELECT:%.*]], [[FOR_BODY]] ]
; NO-VP-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds nuw [8 x i8], ptr [[ARR]], i64 [[I_018]]
; NO-VP-NEXT:    [[TMP16:%.*]] = load i64, ptr [[ARRAYIDX2]], align 8
; NO-VP-NEXT:    [[CMP3:%.*]] = icmp slt i64 [[TMP16]], [[MIN_017]]
; NO-VP-NEXT:    [[SPEC_SELECT]] = select i1 [[CMP3]], i64 [[I_018]], i64 [[MINLOC_016]]
; NO-VP-NEXT:    [[SPEC_SELECT14]] = tail call i64 @llvm.smin.i64(i64 [[TMP16]], i64 [[MIN_017]])
; NO-VP-NEXT:    [[INC]] = add nuw nsw i64 [[I_018]], 1
; NO-VP-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[INC]], [[N]]
; NO-VP-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP39:![0-9]+]]
; NO-VP:       for.end:
; NO-VP-NEXT:    [[SPEC_SELECT_LCSSA:%.*]] = phi i64 [ [[SPEC_SELECT]], [[FOR_BODY]] ], [ [[TMP15]], [[MIDDLE_BLOCK]] ]
; NO-VP-NEXT:    ret i64 [[SPEC_SELECT_LCSSA]]
;
entry:
  br label %for.body

for.body:
  %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]
  %rdx = phi i64 [ %start, %entry ], [ %min, %for.body ]
  %rdx2 = phi i64 [ 0, %entry ], [ %select, %for.body ]
  %arrayidx = getelementptr inbounds nuw [8 x i8], ptr %arr, i64 %iv
  %0 = load i64, ptr %arrayidx
  %cmp = icmp slt i64 %0, %rdx
  %select = select i1 %cmp, i64 %iv, i64 %rdx2
  %min = tail call i64 @llvm.smin.i64(i64 %0, i64 %rdx)
  %iv.next = add nuw nsw i64 %iv, 1
  %exitcond.not = icmp eq i64 %iv.next, %n
  br i1 %exitcond.not, label %for.end, label %for.body, !llvm.loop !0

for.end:
  ret i64 %select
}


define i64 @argmin_last_index(ptr %arr, i64 %n, i64 %start) {
; IF-EVL-LABEL: @argmin_last_index(
; IF-EVL-NEXT:  entry:
; IF-EVL-NEXT:    br label [[FOR_BODY:%.*]]
; IF-EVL:       vector.ph:
; IF-EVL-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 2 x i64> poison, i64 [[TMP0:%.*]], i64 0
; IF-EVL-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 2 x i64> [[BROADCAST_SPLATINSERT]], <vscale x 2 x i64> poison, <vscale x 2 x i32> zeroinitializer
; IF-EVL-NEXT:    [[TMP1:%.*]] = call <vscale x 2 x i64> @llvm.stepvector.nxv2i64()
; IF-EVL-NEXT:    br label [[VECTOR_BODY:%.*]]
; IF-EVL:       vector.body:
; IF-EVL-NEXT:    [[I_018:%.*]] = phi i64 [ 0, [[FOR_BODY]] ], [ [[CURRENT_ITERATION_NEXT:%.*]], [[VECTOR_BODY]] ]
; IF-EVL-NEXT:    [[VEC_IND:%.*]] = phi <vscale x 2 x i64> [ [[TMP1]], [[FOR_BODY]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
; IF-EVL-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 2 x i64> [ [[BROADCAST_SPLAT]], [[FOR_BODY]] ], [ [[TMP7:%.*]], [[VECTOR_BODY]] ]
; IF-EVL-NEXT:    [[VEC_PHI1:%.*]] = phi <vscale x 2 x i64> [ splat (i64 -9223372036854775808), [[FOR_BODY]] ], [ [[TMP8:%.*]], [[VECTOR_BODY]] ]
; IF-EVL-NEXT:    [[AVL:%.*]] = phi i64 [ [[N:%.*]], [[FOR_BODY]] ], [ [[AVL_NEXT:%.*]], [[VECTOR_BODY]] ]
; IF-EVL-NEXT:    [[TMP2:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[AVL]], i32 2, i1 true)
; IF-EVL-NEXT:    [[TMP3:%.*]] = zext i32 [[TMP2]] to i64
; IF-EVL-NEXT:    [[BROADCAST_SPLATINSERT2:%.*]] = insertelement <vscale x 2 x i64> poison, i64 [[TMP3]], i64 0
; IF-EVL-NEXT:    [[BROADCAST_SPLAT3:%.*]] = shufflevector <vscale x 2 x i64> [[BROADCAST_SPLATINSERT2]], <vscale x 2 x i64> poison, <vscale x 2 x i32> zeroinitializer
; IF-EVL-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds nuw [8 x i8], ptr [[ARR:%.*]], i64 [[I_018]]
; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 2 x i64> @llvm.vp.load.nxv2i64.p0(ptr align 8 [[ARRAYIDX2]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP2]])
; IF-EVL-NEXT:    [[TMP5:%.*]] = icmp sle <vscale x 2 x i64> [[VP_OP_LOAD]], [[VEC_PHI]]
; IF-EVL-NEXT:    [[TMP6:%.*]] = call <vscale x 2 x i64> @llvm.smin.nxv2i64(<vscale x 2 x i64> [[VP_OP_LOAD]], <vscale x 2 x i64> [[VEC_PHI]])
; IF-EVL-NEXT:    [[TMP7]] = call <vscale x 2 x i64> @llvm.vp.merge.nxv2i64(<vscale x 2 x i1> splat (i1 true), <vscale x 2 x i64> [[TMP6]], <vscale x 2 x i64> [[VEC_PHI]], i32 [[TMP2]])
; IF-EVL-NEXT:    [[TMP8]] = call <vscale x 2 x i64> @llvm.vp.merge.nxv2i64(<vscale x 2 x i1> [[TMP5]], <vscale x 2 x i64> [[VEC_IND]], <vscale x 2 x i64> [[VEC_PHI1]], i32 [[TMP2]])
; IF-EVL-NEXT:    [[CURRENT_ITERATION_NEXT]] = add i64 [[TMP3]], [[I_018]]
; IF-EVL-NEXT:    [[AVL_NEXT]] = sub nuw i64 [[AVL]], [[TMP3]]
; IF-EVL-NEXT:    [[VEC_IND_NEXT]] = add nuw nsw <vscale x 2 x i64> [[VEC_IND]], [[BROADCAST_SPLAT3]]
; IF-EVL-NEXT:    [[TMP9:%.*]] = icmp eq i64 [[AVL_NEXT]], 0
; IF-EVL-NEXT:    br i1 [[TMP9]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP25:![0-9]+]]
; IF-EVL:       middle.block:
; IF-EVL-NEXT:    [[TMP10:%.*]] = call i64 @llvm.vector.reduce.smin.nxv2i64(<vscale x 2 x i64> [[TMP7]])
; IF-EVL-NEXT:    [[BROADCAST_SPLATINSERT4:%.*]] = insertelement <vscale x 2 x i64> poison, i64 [[TMP10]], i64 0
; IF-EVL-NEXT:    [[BROADCAST_SPLAT5:%.*]] = shufflevector <vscale x 2 x i64> [[BROADCAST_SPLATINSERT4]], <vscale x 2 x i64> poison, <vscale x 2 x i32> zeroinitializer
; IF-EVL-NEXT:    [[TMP11:%.*]] = icmp eq <vscale x 2 x i64> [[TMP7]], [[BROADCAST_SPLAT5]]
; IF-EVL-NEXT:    [[TMP12:%.*]] = select <vscale x 2 x i1> [[TMP11]], <vscale x 2 x i64> [[TMP8]], <vscale x 2 x i64> splat (i64 -9223372036854775808)
; IF-EVL-NEXT:    [[TMP13:%.*]] = call i64 @llvm.vector.reduce.smax.nxv2i64(<vscale x 2 x i64> [[TMP12]])
; IF-EVL-NEXT:    [[TMP14:%.*]] = icmp ne i64 [[TMP13]], -9223372036854775808
; IF-EVL-NEXT:    [[SPEC_SELECT_LCSSA:%.*]] = select i1 [[TMP14]], i64 [[TMP13]], i64 0
; IF-EVL-NEXT:    br label [[FOR_END:%.*]]
; IF-EVL:       for.end:
; IF-EVL-NEXT:    ret i64 [[SPEC_SELECT_LCSSA]]
;
; NO-VP-LABEL: @argmin_last_index(
; NO-VP-NEXT:  entry:
; NO-VP-NEXT:    [[TMP1:%.*]] = call i64 @llvm.vscale.i64()
; NO-VP-NEXT:    [[TMP2:%.*]] = shl nuw i64 [[TMP1]], 1
; NO-VP-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], [[TMP2]]
; NO-VP-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; NO-VP:       vector.ph:
; NO-VP-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP2]]
; NO-VP-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
; NO-VP-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 2 x i64> poison, i64 [[TMP0:%.*]], i64 0
; NO-VP-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 2 x i64> [[BROADCAST_SPLATINSERT]], <vscale x 2 x i64> poison, <vscale x 2 x i32> zeroinitializer
; NO-VP-NEXT:    [[TMP4:%.*]] = call <vscale x 2 x i64> @llvm.stepvector.nxv2i64()
; NO-VP-NEXT:    [[BROADCAST_SPLATINSERT1:%.*]] = insertelement <vscale x 2 x i64> poison, i64 [[TMP2]], i64 0
; NO-VP-NEXT:    [[BROADCAST_SPLAT2:%.*]] = shufflevector <vscale x 2 x i64> [[BROADCAST_SPLATINSERT1]], <vscale x 2 x i64> poison, <vscale x 2 x i32> zeroinitializer
; NO-VP-NEXT:    br label [[VECTOR_BODY:%.*]]
; NO-VP:       vector.body:
; NO-VP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; NO-VP-NEXT:    [[VEC_IND:%.*]] = phi <vscale x 2 x i64> [ [[TMP4]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
; NO-VP-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 2 x i64> [ [[BROADCAST_SPLAT]], [[VECTOR_PH]] ], [ [[TMP8:%.*]], [[VECTOR_BODY]] ]
; NO-VP-NEXT:    [[VEC_PHI3:%.*]] = phi <vscale x 2 x i64> [ splat (i64 -9223372036854775808), [[VECTOR_PH]] ], [ [[TMP7:%.*]], [[VECTOR_BODY]] ]
; NO-VP-NEXT:    [[TMP5:%.*]] = getelementptr inbounds nuw [8 x i8], ptr [[ARR:%.*]], i64 [[INDEX]]
; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 2 x i64>, ptr [[TMP5]], align 8
; NO-VP-NEXT:    [[TMP6:%.*]] = icmp sle <vscale x 2 x i64> [[WIDE_LOAD]], [[VEC_PHI]]
; NO-VP-NEXT:    [[TMP7]] = select <vscale x 2 x i1> [[TMP6]], <vscale x 2 x i64> [[VEC_IND]], <vscale x 2 x i64> [[VEC_PHI3]]
; NO-VP-NEXT:    [[TMP8]] = call <vscale x 2 x i64> @llvm.smin.nxv2i64(<vscale x 2 x i64> [[WIDE_LOAD]], <vscale x 2 x i64> [[VEC_PHI]])
; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP2]]
; NO-VP-NEXT:    [[VEC_IND_NEXT]] = add nuw nsw <vscale x 2 x i64> [[VEC_IND]], [[BROADCAST_SPLAT2]]
; NO-VP-NEXT:    [[TMP9:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; NO-VP-NEXT:    br i1 [[TMP9]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP40:![0-9]+]]
; NO-VP:       middle.block:
; NO-VP-NEXT:    [[TMP10:%.*]] = call i64 @llvm.vector.reduce.smin.nxv2i64(<vscale x 2 x i64> [[TMP8]])
; NO-VP-NEXT:    [[BROADCAST_SPLATINSERT4:%.*]] = insertelement <vscale x 2 x i64> poison, i64 [[TMP10]], i64 0
; NO-VP-NEXT:    [[BROADCAST_SPLAT5:%.*]] = shufflevector <vscale x 2 x i64> [[BROADCAST_SPLATINSERT4]], <vscale x 2 x i64> poison, <vscale x 2 x i32> zeroinitializer
; NO-VP-NEXT:    [[TMP11:%.*]] = icmp eq <vscale x 2 x i64> [[TMP8]], [[BROADCAST_SPLAT5]]
; NO-VP-NEXT:    [[TMP12:%.*]] = select <vscale x 2 x i1> [[TMP11]], <vscale x 2 x i64> [[TMP7]], <vscale x 2 x i64> splat (i64 -9223372036854775808)
; NO-VP-NEXT:    [[TMP13:%.*]] = call i64 @llvm.vector.reduce.smax.nxv2i64(<vscale x 2 x i64> [[TMP12]])
; NO-VP-NEXT:    [[TMP14:%.*]] = icmp ne i64 [[TMP13]], -9223372036854775808
; NO-VP-NEXT:    [[TMP15:%.*]] = select i1 [[TMP14]], i64 [[TMP13]], i64 0
; NO-VP-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
; NO-VP-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
; NO-VP:       scalar.ph:
; NO-VP-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_BODY_PREHEADER:%.*]] ]
; NO-VP-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i64 [ [[TMP10]], [[MIDDLE_BLOCK]] ], [ [[TMP0]], [[FOR_BODY_PREHEADER]] ]
; NO-VP-NEXT:    [[BC_MERGE_RDX6:%.*]] = phi i64 [ [[TMP15]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_BODY_PREHEADER]] ]
; NO-VP-NEXT:    br label [[FOR_BODY:%.*]]
; NO-VP:       for.body:
; NO-VP-NEXT:    [[I_018:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INC:%.*]], [[FOR_BODY]] ]
; NO-VP-NEXT:    [[MIN_017:%.*]] = phi i64 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[SPEC_SELECT14:%.*]], [[FOR_BODY]] ]
; NO-VP-NEXT:    [[MINLOC_016:%.*]] = phi i64 [ [[BC_MERGE_RDX6]], [[SCALAR_PH]] ], [ [[SPEC_SELECT:%.*]], [[FOR_BODY]] ]
; NO-VP-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds nuw [8 x i8], ptr [[ARR]], i64 [[I_018]]
; NO-VP-NEXT:    [[TMP16:%.*]] = load i64, ptr [[ARRAYIDX2]], align 8
; NO-VP-NEXT:    [[CMP3:%.*]] = icmp sle i64 [[TMP16]], [[MIN_017]]
; NO-VP-NEXT:    [[SPEC_SELECT]] = select i1 [[CMP3]], i64 [[I_018]], i64 [[MINLOC_016]]
; NO-VP-NEXT:    [[SPEC_SELECT14]] = tail call i64 @llvm.smin.i64(i64 [[TMP16]], i64 [[MIN_017]])
; NO-VP-NEXT:    [[INC]] = add nuw nsw i64 [[I_018]], 1
; NO-VP-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[INC]], [[N]]
; NO-VP-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP41:![0-9]+]]
; NO-VP:       for.end:
; NO-VP-NEXT:    [[SPEC_SELECT_LCSSA:%.*]] = phi i64 [ [[SPEC_SELECT]], [[FOR_BODY]] ], [ [[TMP15]], [[MIDDLE_BLOCK]] ]
; NO-VP-NEXT:    ret i64 [[SPEC_SELECT_LCSSA]]
;
entry:
  br label %for.body

for.body:
  %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]
  %rdx = phi i64 [ %start, %entry ], [ %min, %for.body ]
  %rdx2 = phi i64 [ 0, %entry ], [ %select, %for.body ]
  %arrayidx = getelementptr inbounds nuw [8 x i8], ptr %arr, i64 %iv
  %0 = load i64, ptr %arrayidx
  %cmp = icmp sle i64 %0, %rdx ; tests `sle` instead of strict `slt` from previous test
  %select = select i1 %cmp, i64 %iv, i64 %rdx2
  %min = tail call i64 @llvm.smin.i64(i64 %0, i64 %rdx)
  %iv.next = add nuw nsw i64 %iv, 1
  %exitcond.not = icmp eq i64 %iv.next, %n
  br i1 %exitcond.not, label %for.end, label %for.body, !llvm.loop !0

for.end:
  ret i64 %select
}


define i64 @argmax_first_index(ptr %arr, i64 %n, i64 %start) {
; IF-EVL-LABEL: @argmax_first_index(
; IF-EVL-NEXT:  entry:
; IF-EVL-NEXT:    br label [[FOR_BODY:%.*]]
; IF-EVL:       vector.ph:
; IF-EVL-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 2 x i64> poison, i64 [[TMP0:%.*]], i64 0
; IF-EVL-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 2 x i64> [[BROADCAST_SPLATINSERT]], <vscale x 2 x i64> poison, <vscale x 2 x i32> zeroinitializer
; IF-EVL-NEXT:    [[TMP1:%.*]] = call <vscale x 2 x i64> @llvm.stepvector.nxv2i64()
; IF-EVL-NEXT:    br label [[VECTOR_BODY:%.*]]
; IF-EVL:       vector.body:
; IF-EVL-NEXT:    [[I_018:%.*]] = phi i64 [ 0, [[FOR_BODY]] ], [ [[CURRENT_ITERATION_NEXT:%.*]], [[VECTOR_BODY]] ]
; IF-EVL-NEXT:    [[VEC_IND:%.*]] = phi <vscale x 2 x i64> [ [[TMP1]], [[FOR_BODY]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
; IF-EVL-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 2 x i64> [ [[BROADCAST_SPLAT]], [[FOR_BODY]] ], [ [[TMP7:%.*]], [[VECTOR_BODY]] ]
; IF-EVL-NEXT:    [[VEC_PHI1:%.*]] = phi <vscale x 2 x i64> [ poison, [[FOR_BODY]] ], [ [[TMP8:%.*]], [[VECTOR_BODY]] ]
; IF-EVL-NEXT:    [[AVL:%.*]] = phi i64 [ [[N:%.*]], [[FOR_BODY]] ], [ [[AVL_NEXT:%.*]], [[VECTOR_BODY]] ]
; IF-EVL-NEXT:    [[TMP2:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[AVL]], i32 2, i1 true)
; IF-EVL-NEXT:    [[TMP3:%.*]] = zext i32 [[TMP2]] to i64
; IF-EVL-NEXT:    [[BROADCAST_SPLATINSERT2:%.*]] = insertelement <vscale x 2 x i64> poison, i64 [[TMP3]], i64 0
; IF-EVL-NEXT:    [[BROADCAST_SPLAT3:%.*]] = shufflevector <vscale x 2 x i64> [[BROADCAST_SPLATINSERT2]], <vscale x 2 x i64> poison, <vscale x 2 x i32> zeroinitializer
; IF-EVL-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds nuw [8 x i8], ptr [[ARR:%.*]], i64 [[I_018]]
; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 2 x i64> @llvm.vp.load.nxv2i64.p0(ptr align 8 [[ARRAYIDX2]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP2]])
; IF-EVL-NEXT:    [[TMP5:%.*]] = icmp sgt <vscale x 2 x i64> [[VP_OP_LOAD]], [[VEC_PHI]]
; IF-EVL-NEXT:    [[TMP6:%.*]] = call <vscale x 2 x i64> @llvm.smax.nxv2i64(<vscale x 2 x i64> [[VP_OP_LOAD]], <vscale x 2 x i64> [[VEC_PHI]])
; IF-EVL-NEXT:    [[TMP7]] = call <vscale x 2 x i64> @llvm.vp.merge.nxv2i64(<vscale x 2 x i1> splat (i1 true), <vscale x 2 x i64> [[TMP6]], <vscale x 2 x i64> [[VEC_PHI]], i32 [[TMP2]])
; IF-EVL-NEXT:    [[TMP8]] = call <vscale x 2 x i64> @llvm.vp.merge.nxv2i64(<vscale x 2 x i1> [[TMP5]], <vscale x 2 x i64> [[VEC_IND]], <vscale x 2 x i64> [[VEC_PHI1]], i32 [[TMP2]])
; IF-EVL-NEXT:    [[CURRENT_ITERATION_NEXT]] = add i64 [[TMP3]], [[I_018]]
; IF-EVL-NEXT:    [[AVL_NEXT]] = sub nuw i64 [[AVL]], [[TMP3]]
; IF-EVL-NEXT:    [[VEC_IND_NEXT]] = add nuw nsw <vscale x 2 x i64> [[VEC_IND]], [[BROADCAST_SPLAT3]]
; IF-EVL-NEXT:    [[TMP9:%.*]] = icmp eq i64 [[AVL_NEXT]], 0
; IF-EVL-NEXT:    br i1 [[TMP9]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP26:![0-9]+]]
; IF-EVL:       middle.block:
; IF-EVL-NEXT:    [[TMP10:%.*]] = call i64 @llvm.vector.reduce.smax.nxv2i64(<vscale x 2 x i64> [[TMP7]])
; IF-EVL-NEXT:    [[BROADCAST_SPLATINSERT4:%.*]] = insertelement <vscale x 2 x i64> poison, i64 [[TMP10]], i64 0
; IF-EVL-NEXT:    [[BROADCAST_SPLAT5:%.*]] = shufflevector <vscale x 2 x i64> [[BROADCAST_SPLATINSERT4]], <vscale x 2 x i64> poison, <vscale x 2 x i32> zeroinitializer
; IF-EVL-NEXT:    [[TMP11:%.*]] = icmp eq <vscale x 2 x i64> [[TMP7]], [[BROADCAST_SPLAT5]]
; IF-EVL-NEXT:    [[TMP12:%.*]] = select <vscale x 2 x i1> [[TMP11]], <vscale x 2 x i64> [[TMP8]], <vscale x 2 x i64> splat (i64 -1)
; IF-EVL-NEXT:    [[TMP13:%.*]] = call i64 @llvm.vector.reduce.umin.nxv2i64(<vscale x 2 x i64> [[TMP12]])
; IF-EVL-NEXT:    [[TMP14:%.*]] = icmp eq i64 [[TMP10]], [[TMP0]]
; IF-EVL-NEXT:    [[SPEC_SELECT_LCSSA:%.*]] = select i1 [[TMP14]], i64 0, i64 [[TMP13]]
; IF-EVL-NEXT:    br label [[FOR_END:%.*]]
; IF-EVL:       for.end:
; IF-EVL-NEXT:    ret i64 [[SPEC_SELECT_LCSSA]]
;
; NO-VP-LABEL: @argmax_first_index(
; NO-VP-NEXT:  entry:
; NO-VP-NEXT:    [[TMP1:%.*]] = call i64 @llvm.vscale.i64()
; NO-VP-NEXT:    [[TMP2:%.*]] = shl nuw i64 [[TMP1]], 1
; NO-VP-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], [[TMP2]]
; NO-VP-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; NO-VP:       vector.ph:
; NO-VP-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP2]]
; NO-VP-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
; NO-VP-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 2 x i64> poison, i64 [[TMP0:%.*]], i64 0
; NO-VP-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 2 x i64> [[BROADCAST_SPLATINSERT]], <vscale x 2 x i64> poison, <vscale x 2 x i32> zeroinitializer
; NO-VP-NEXT:    [[TMP4:%.*]] = call <vscale x 2 x i64> @llvm.stepvector.nxv2i64()
; NO-VP-NEXT:    [[BROADCAST_SPLATINSERT1:%.*]] = insertelement <vscale x 2 x i64> poison, i64 [[TMP2]], i64 0
; NO-VP-NEXT:    [[BROADCAST_SPLAT2:%.*]] = shufflevector <vscale x 2 x i64> [[BROADCAST_SPLATINSERT1]], <vscale x 2 x i64> poison, <vscale x 2 x i32> zeroinitializer
; NO-VP-NEXT:    br label [[VECTOR_BODY:%.*]]
; NO-VP:       vector.body:
; NO-VP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; NO-VP-NEXT:    [[VEC_IND:%.*]] = phi <vscale x 2 x i64> [ [[TMP4]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
; NO-VP-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 2 x i64> [ [[BROADCAST_SPLAT]], [[VECTOR_PH]] ], [ [[TMP8:%.*]], [[VECTOR_BODY]] ]
; NO-VP-NEXT:    [[VEC_PHI3:%.*]] = phi <vscale x 2 x i64> [ poison, [[VECTOR_PH]] ], [ [[TMP7:%.*]], [[VECTOR_BODY]] ]
; NO-VP-NEXT:    [[TMP5:%.*]] = getelementptr inbounds nuw [8 x i8], ptr [[ARR:%.*]], i64 [[INDEX]]
; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 2 x i64>, ptr [[TMP5]], align 8
; NO-VP-NEXT:    [[TMP6:%.*]] = icmp sgt <vscale x 2 x i64> [[WIDE_LOAD]], [[VEC_PHI]]
; NO-VP-NEXT:    [[TMP7]] = select <vscale x 2 x i1> [[TMP6]], <vscale x 2 x i64> [[VEC_IND]], <vscale x 2 x i64> [[VEC_PHI3]]
; NO-VP-NEXT:    [[TMP8]] = call <vscale x 2 x i64> @llvm.smax.nxv2i64(<vscale x 2 x i64> [[WIDE_LOAD]], <vscale x 2 x i64> [[VEC_PHI]])
; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP2]]
; NO-VP-NEXT:    [[VEC_IND_NEXT]] = add nuw nsw <vscale x 2 x i64> [[VEC_IND]], [[BROADCAST_SPLAT2]]
; NO-VP-NEXT:    [[TMP9:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; NO-VP-NEXT:    br i1 [[TMP9]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP42:![0-9]+]]
; NO-VP:       middle.block:
; NO-VP-NEXT:    [[TMP10:%.*]] = call i64 @llvm.vector.reduce.smax.nxv2i64(<vscale x 2 x i64> [[TMP8]])
; NO-VP-NEXT:    [[BROADCAST_SPLATINSERT4:%.*]] = insertelement <vscale x 2 x i64> poison, i64 [[TMP10]], i64 0
; NO-VP-NEXT:    [[BROADCAST_SPLAT5:%.*]] = shufflevector <vscale x 2 x i64> [[BROADCAST_SPLATINSERT4]], <vscale x 2 x i64> poison, <vscale x 2 x i32> zeroinitializer
; NO-VP-NEXT:    [[TMP11:%.*]] = icmp eq <vscale x 2 x i64> [[TMP8]], [[BROADCAST_SPLAT5]]
; NO-VP-NEXT:    [[TMP12:%.*]] = select <vscale x 2 x i1> [[TMP11]], <vscale x 2 x i64> [[TMP7]], <vscale x 2 x i64> splat (i64 -1)
; NO-VP-NEXT:    [[TMP13:%.*]] = call i64 @llvm.vector.reduce.umin.nxv2i64(<vscale x 2 x i64> [[TMP12]])
; NO-VP-NEXT:    [[TMP14:%.*]] = icmp eq i64 [[TMP10]], [[TMP0]]
; NO-VP-NEXT:    [[TMP15:%.*]] = select i1 [[TMP14]], i64 0, i64 [[TMP13]]
; NO-VP-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
; NO-VP-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
; NO-VP:       scalar.ph:
; NO-VP-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_BODY_PREHEADER:%.*]] ]
; NO-VP-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i64 [ [[TMP10]], [[MIDDLE_BLOCK]] ], [ [[TMP0]], [[FOR_BODY_PREHEADER]] ]
; NO-VP-NEXT:    [[BC_MERGE_RDX6:%.*]] = phi i64 [ [[TMP15]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_BODY_PREHEADER]] ]
; NO-VP-NEXT:    br label [[FOR_BODY:%.*]]
; NO-VP:       for.body:
; NO-VP-NEXT:    [[I_018:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INC:%.*]], [[FOR_BODY]] ]
; NO-VP-NEXT:    [[MAX_017:%.*]] = phi i64 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[SPEC_SELECT14:%.*]], [[FOR_BODY]] ]
; NO-VP-NEXT:    [[MAX_016:%.*]] = phi i64 [ [[BC_MERGE_RDX6]], [[SCALAR_PH]] ], [ [[SPEC_SELECT:%.*]], [[FOR_BODY]] ]
; NO-VP-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds nuw [8 x i8], ptr [[ARR]], i64 [[I_018]]
; NO-VP-NEXT:    [[TMP16:%.*]] = load i64, ptr [[ARRAYIDX2]], align 8
; NO-VP-NEXT:    [[CMP3:%.*]] = icmp sgt i64 [[TMP16]], [[MAX_017]]
; NO-VP-NEXT:    [[SPEC_SELECT]] = select i1 [[CMP3]], i64 [[I_018]], i64 [[MAX_016]]
; NO-VP-NEXT:    [[SPEC_SELECT14]] = tail call i64 @llvm.smax.i64(i64 [[TMP16]], i64 [[MAX_017]])
; NO-VP-NEXT:    [[INC]] = add nuw nsw i64 [[I_018]], 1
; NO-VP-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[INC]], [[N]]
; NO-VP-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP43:![0-9]+]]
; NO-VP:       for.end:
; NO-VP-NEXT:    [[SPEC_SELECT_LCSSA:%.*]] = phi i64 [ [[SPEC_SELECT]], [[FOR_BODY]] ], [ [[TMP15]], [[MIDDLE_BLOCK]] ]
; NO-VP-NEXT:    ret i64 [[SPEC_SELECT_LCSSA]]
;
entry:
  br label %for.body

for.body:
  %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]
  %rdx = phi i64 [ %start, %entry ], [ %max, %for.body ]
  %rdx2 = phi i64 [ 0, %entry ], [ %select, %for.body ]
  %arrayidx = getelementptr inbounds nuw [8 x i8], ptr %arr, i64 %iv
  %0 = load i64, ptr %arrayidx
  %cmp = icmp sgt i64 %0, %rdx
  %select = select i1 %cmp, i64 %iv, i64 %rdx2
  %max = tail call i64 @llvm.smax.i64(i64 %0, i64 %rdx)
  %iv.next = add nuw nsw i64 %iv, 1
  %exitcond.not = icmp eq i64 %iv.next, %n
  br i1 %exitcond.not, label %for.end, label %for.body, !llvm.loop !0

for.end:
  ret i64 %select
}


define i64 @argmax_last_index(ptr %arr, i64 %n, i64 %start) {
; IF-EVL-LABEL: @argmax_last_index(
; IF-EVL-NEXT:  entry:
; IF-EVL-NEXT:    br label [[FOR_BODY:%.*]]
; IF-EVL:       vector.ph:
; IF-EVL-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 2 x i64> poison, i64 [[TMP0:%.*]], i64 0
; IF-EVL-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 2 x i64> [[BROADCAST_SPLATINSERT]], <vscale x 2 x i64> poison, <vscale x 2 x i32> zeroinitializer
; IF-EVL-NEXT:    [[TMP1:%.*]] = call <vscale x 2 x i64> @llvm.stepvector.nxv2i64()
; IF-EVL-NEXT:    br label [[VECTOR_BODY:%.*]]
; IF-EVL:       vector.body:
; IF-EVL-NEXT:    [[I_018:%.*]] = phi i64 [ 0, [[FOR_BODY]] ], [ [[CURRENT_ITERATION_NEXT:%.*]], [[VECTOR_BODY]] ]
; IF-EVL-NEXT:    [[VEC_IND:%.*]] = phi <vscale x 2 x i64> [ [[TMP1]], [[FOR_BODY]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
; IF-EVL-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 2 x i64> [ [[BROADCAST_SPLAT]], [[FOR_BODY]] ], [ [[TMP7:%.*]], [[VECTOR_BODY]] ]
; IF-EVL-NEXT:    [[VEC_PHI1:%.*]] = phi <vscale x 2 x i64> [ splat (i64 -9223372036854775808), [[FOR_BODY]] ], [ [[TMP8:%.*]], [[VECTOR_BODY]] ]
; IF-EVL-NEXT:    [[AVL:%.*]] = phi i64 [ [[N:%.*]], [[FOR_BODY]] ], [ [[AVL_NEXT:%.*]], [[VECTOR_BODY]] ]
; IF-EVL-NEXT:    [[TMP2:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[AVL]], i32 2, i1 true)
; IF-EVL-NEXT:    [[TMP3:%.*]] = zext i32 [[TMP2]] to i64
; IF-EVL-NEXT:    [[BROADCAST_SPLATINSERT2:%.*]] = insertelement <vscale x 2 x i64> poison, i64 [[TMP3]], i64 0
; IF-EVL-NEXT:    [[BROADCAST_SPLAT3:%.*]] = shufflevector <vscale x 2 x i64> [[BROADCAST_SPLATINSERT2]], <vscale x 2 x i64> poison, <vscale x 2 x i32> zeroinitializer
; IF-EVL-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds nuw [8 x i8], ptr [[ARR:%.*]], i64 [[I_018]]
; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 2 x i64> @llvm.vp.load.nxv2i64.p0(ptr align 8 [[ARRAYIDX2]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP2]])
; IF-EVL-NEXT:    [[TMP5:%.*]] = icmp sge <vscale x 2 x i64> [[VP_OP_LOAD]], [[VEC_PHI]]
; IF-EVL-NEXT:    [[TMP6:%.*]] = call <vscale x 2 x i64> @llvm.smax.nxv2i64(<vscale x 2 x i64> [[VP_OP_LOAD]], <vscale x 2 x i64> [[VEC_PHI]])
; IF-EVL-NEXT:    [[TMP7]] = call <vscale x 2 x i64> @llvm.vp.merge.nxv2i64(<vscale x 2 x i1> splat (i1 true), <vscale x 2 x i64> [[TMP6]], <vscale x 2 x i64> [[VEC_PHI]], i32 [[TMP2]])
; IF-EVL-NEXT:    [[TMP8]] = call <vscale x 2 x i64> @llvm.vp.merge.nxv2i64(<vscale x 2 x i1> [[TMP5]], <vscale x 2 x i64> [[VEC_IND]], <vscale x 2 x i64> [[VEC_PHI1]], i32 [[TMP2]])
; IF-EVL-NEXT:    [[CURRENT_ITERATION_NEXT]] = add i64 [[TMP3]], [[I_018]]
; IF-EVL-NEXT:    [[AVL_NEXT]] = sub nuw i64 [[AVL]], [[TMP3]]
; IF-EVL-NEXT:    [[VEC_IND_NEXT]] = add nuw nsw <vscale x 2 x i64> [[VEC_IND]], [[BROADCAST_SPLAT3]]
; IF-EVL-NEXT:    [[TMP9:%.*]] = icmp eq i64 [[AVL_NEXT]], 0
; IF-EVL-NEXT:    br i1 [[TMP9]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP27:![0-9]+]]
; IF-EVL:       middle.block:
; IF-EVL-NEXT:    [[TMP10:%.*]] = call i64 @llvm.vector.reduce.smax.nxv2i64(<vscale x 2 x i64> [[TMP7]])
; IF-EVL-NEXT:    [[BROADCAST_SPLATINSERT4:%.*]] = insertelement <vscale x 2 x i64> poison, i64 [[TMP10]], i64 0
; IF-EVL-NEXT:    [[BROADCAST_SPLAT5:%.*]] = shufflevector <vscale x 2 x i64> [[BROADCAST_SPLATINSERT4]], <vscale x 2 x i64> poison, <vscale x 2 x i32> zeroinitializer
; IF-EVL-NEXT:    [[TMP11:%.*]] = icmp eq <vscale x 2 x i64> [[TMP7]], [[BROADCAST_SPLAT5]]
; IF-EVL-NEXT:    [[TMP12:%.*]] = select <vscale x 2 x i1> [[TMP11]], <vscale x 2 x i64> [[TMP8]], <vscale x 2 x i64> splat (i64 -9223372036854775808)
; IF-EVL-NEXT:    [[TMP13:%.*]] = call i64 @llvm.vector.reduce.smax.nxv2i64(<vscale x 2 x i64> [[TMP12]])
; IF-EVL-NEXT:    [[TMP14:%.*]] = icmp ne i64 [[TMP13]], -9223372036854775808
; IF-EVL-NEXT:    [[SPEC_SELECT_LCSSA:%.*]] = select i1 [[TMP14]], i64 [[TMP13]], i64 0
; IF-EVL-NEXT:    br label [[FOR_END:%.*]]
; IF-EVL:       for.end:
; IF-EVL-NEXT:    ret i64 [[SPEC_SELECT_LCSSA]]
;
; NO-VP-LABEL: @argmax_last_index(
; NO-VP-NEXT:  entry:
; NO-VP-NEXT:    [[TMP1:%.*]] = call i64 @llvm.vscale.i64()
; NO-VP-NEXT:    [[TMP2:%.*]] = shl nuw i64 [[TMP1]], 1
; NO-VP-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], [[TMP2]]
; NO-VP-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; NO-VP:       vector.ph:
; NO-VP-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP2]]
; NO-VP-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
; NO-VP-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 2 x i64> poison, i64 [[TMP0:%.*]], i64 0
; NO-VP-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 2 x i64> [[BROADCAST_SPLATINSERT]], <vscale x 2 x i64> poison, <vscale x 2 x i32> zeroinitializer
; NO-VP-NEXT:    [[TMP4:%.*]] = call <vscale x 2 x i64> @llvm.stepvector.nxv2i64()
; NO-VP-NEXT:    [[BROADCAST_SPLATINSERT1:%.*]] = insertelement <vscale x 2 x i64> poison, i64 [[TMP2]], i64 0
; NO-VP-NEXT:    [[BROADCAST_SPLAT2:%.*]] = shufflevector <vscale x 2 x i64> [[BROADCAST_SPLATINSERT1]], <vscale x 2 x i64> poison, <vscale x 2 x i32> zeroinitializer
; NO-VP-NEXT:    br label [[VECTOR_BODY:%.*]]
; NO-VP:       vector.body:
; NO-VP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; NO-VP-NEXT:    [[VEC_IND:%.*]] = phi <vscale x 2 x i64> [ [[TMP4]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
; NO-VP-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 2 x i64> [ [[BROADCAST_SPLAT]], [[VECTOR_PH]] ], [ [[TMP8:%.*]], [[VECTOR_BODY]] ]
; NO-VP-NEXT:    [[VEC_PHI3:%.*]] = phi <vscale x 2 x i64> [ splat (i64 -9223372036854775808), [[VECTOR_PH]] ], [ [[TMP7:%.*]], [[VECTOR_BODY]] ]
; NO-VP-NEXT:    [[TMP5:%.*]] = getelementptr inbounds nuw [8 x i8], ptr [[ARR:%.*]], i64 [[INDEX]]
; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 2 x i64>, ptr [[TMP5]], align 8
; NO-VP-NEXT:    [[TMP6:%.*]] = icmp sge <vscale x 2 x i64> [[WIDE_LOAD]], [[VEC_PHI]]
; NO-VP-NEXT:    [[TMP7]] = select <vscale x 2 x i1> [[TMP6]], <vscale x 2 x i64> [[VEC_IND]], <vscale x 2 x i64> [[VEC_PHI3]]
; NO-VP-NEXT:    [[TMP8]] = call <vscale x 2 x i64> @llvm.smax.nxv2i64(<vscale x 2 x i64> [[WIDE_LOAD]], <vscale x 2 x i64> [[VEC_PHI]])
; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP2]]
; NO-VP-NEXT:    [[VEC_IND_NEXT]] = add nuw nsw <vscale x 2 x i64> [[VEC_IND]], [[BROADCAST_SPLAT2]]
; NO-VP-NEXT:    [[TMP9:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; NO-VP-NEXT:    br i1 [[TMP9]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP44:![0-9]+]]
; NO-VP:       middle.block:
; NO-VP-NEXT:    [[TMP10:%.*]] = call i64 @llvm.vector.reduce.smax.nxv2i64(<vscale x 2 x i64> [[TMP8]])
; NO-VP-NEXT:    [[BROADCAST_SPLATINSERT4:%.*]] = insertelement <vscale x 2 x i64> poison, i64 [[TMP10]], i64 0
; NO-VP-NEXT:    [[BROADCAST_SPLAT5:%.*]] = shufflevector <vscale x 2 x i64> [[BROADCAST_SPLATINSERT4]], <vscale x 2 x i64> poison, <vscale x 2 x i32> zeroinitializer
; NO-VP-NEXT:    [[TMP11:%.*]] = icmp eq <vscale x 2 x i64> [[TMP8]], [[BROADCAST_SPLAT5]]
; NO-VP-NEXT:    [[TMP12:%.*]] = select <vscale x 2 x i1> [[TMP11]], <vscale x 2 x i64> [[TMP7]], <vscale x 2 x i64> splat (i64 -9223372036854775808)
; NO-VP-NEXT:    [[TMP13:%.*]] = call i64 @llvm.vector.reduce.smax.nxv2i64(<vscale x 2 x i64> [[TMP12]])
; NO-VP-NEXT:    [[TMP14:%.*]] = icmp ne i64 [[TMP13]], -9223372036854775808
; NO-VP-NEXT:    [[TMP15:%.*]] = select i1 [[TMP14]], i64 [[TMP13]], i64 0
; NO-VP-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
; NO-VP-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
; NO-VP:       scalar.ph:
; NO-VP-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_BODY_PREHEADER:%.*]] ]
; NO-VP-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i64 [ [[TMP10]], [[MIDDLE_BLOCK]] ], [ [[TMP0]], [[FOR_BODY_PREHEADER]] ]
; NO-VP-NEXT:    [[BC_MERGE_RDX6:%.*]] = phi i64 [ [[TMP15]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_BODY_PREHEADER]] ]
; NO-VP-NEXT:    br label [[FOR_BODY:%.*]]
; NO-VP:       for.body:
; NO-VP-NEXT:    [[I_018:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INC:%.*]], [[FOR_BODY]] ]
; NO-VP-NEXT:    [[MAX_017:%.*]] = phi i64 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[SPEC_SELECT14:%.*]], [[FOR_BODY]] ]
; NO-VP-NEXT:    [[MAX_016:%.*]] = phi i64 [ [[BC_MERGE_RDX6]], [[SCALAR_PH]] ], [ [[SPEC_SELECT:%.*]], [[FOR_BODY]] ]
; NO-VP-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds nuw [8 x i8], ptr [[ARR]], i64 [[I_018]]
; NO-VP-NEXT:    [[TMP16:%.*]] = load i64, ptr [[ARRAYIDX2]], align 8
; NO-VP-NEXT:    [[CMP3:%.*]] = icmp sge i64 [[TMP16]], [[MAX_017]]
; NO-VP-NEXT:    [[SPEC_SELECT]] = select i1 [[CMP3]], i64 [[I_018]], i64 [[MAX_016]]
; NO-VP-NEXT:    [[SPEC_SELECT14]] = tail call i64 @llvm.smax.i64(i64 [[TMP16]], i64 [[MAX_017]])
; NO-VP-NEXT:    [[INC]] = add nuw nsw i64 [[I_018]], 1
; NO-VP-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[INC]], [[N]]
; NO-VP-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP45:![0-9]+]]
; NO-VP:       for.end:
; NO-VP-NEXT:    [[SPEC_SELECT_LCSSA:%.*]] = phi i64 [ [[SPEC_SELECT]], [[FOR_BODY]] ], [ [[TMP15]], [[MIDDLE_BLOCK]] ]
; NO-VP-NEXT:    ret i64 [[SPEC_SELECT_LCSSA]]
;
entry:
  br label %for.body

for.body:
  %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]
  %rdx = phi i64 [ %start, %entry ], [ %max, %for.body ]
  %rdx2 = phi i64 [ 0, %entry ], [ %select, %for.body ]
  %arrayidx = getelementptr inbounds nuw [8 x i8], ptr %arr, i64 %iv
  %0 = load i64, ptr %arrayidx
  %cmp = icmp sge i64 %0, %rdx
  %select = select i1 %cmp, i64 %iv, i64 %rdx2
  %max = tail call i64 @llvm.smax.i64(i64 %0, i64 %rdx)
  %iv.next = add nuw nsw i64 %iv, 1
  %exitcond.not = icmp eq i64 %iv.next, %n
  br i1 %exitcond.not, label %for.end, label %for.body, !llvm.loop !0

for.end:
  ret i64 %select
}


!0 = distinct !{!0, !1}
!1 = !{!"llvm.loop.vectorize.enable"}