llvm/test/Transforms/LoopVectorize/AArch64/reduction-recurrence-costs-sve.ll

; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 4
; RUN: opt -p loop-vectorize -S %s | FileCheck --check-prefixes=DEFAULT %s
; RUN: opt -p loop-vectorize -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue -S %s | FileCheck --check-prefixes=PRED %s

target datalayout = "e-m:o-i64:64-i128:128-n32:64-S128"
target triple = "arm64-apple-macosx14.0.0"

define i32 @chained_recurrences(i32 %x, i64 %y, ptr %src.1, i32 %z, ptr %src.2) #0 {
; DEFAULT-LABEL: define i32 @chained_recurrences(
; DEFAULT-SAME: i32 [[X:%.*]], i64 [[Y:%.*]], ptr [[SRC_1:%.*]], i32 [[Z:%.*]], ptr [[SRC_2:%.*]]) #[[ATTR0:[0-9]+]] {
; DEFAULT-NEXT:  entry:
; DEFAULT-NEXT:    [[TMP0:%.*]] = add i64 [[Y]], 1
; DEFAULT-NEXT:    [[TMP1:%.*]] = call i64 @llvm.vscale.i64()
; DEFAULT-NEXT:    [[TMP2:%.*]] = mul i64 [[TMP1]], 8
; DEFAULT-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP0]], [[TMP2]]
; DEFAULT-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; DEFAULT:       vector.ph:
; DEFAULT-NEXT:    [[TMP3:%.*]] = call i64 @llvm.vscale.i64()
; DEFAULT-NEXT:    [[TMP4:%.*]] = mul i64 [[TMP3]], 8
; DEFAULT-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[TMP0]], [[TMP4]]
; DEFAULT-NEXT:    [[N_VEC:%.*]] = sub i64 [[TMP0]], [[N_MOD_VF]]
; DEFAULT-NEXT:    [[TMP5:%.*]] = call i64 @llvm.vscale.i64()
; DEFAULT-NEXT:    [[TMP6:%.*]] = mul i64 [[TMP5]], 8
; DEFAULT-NEXT:    [[TMP7:%.*]] = call i32 @llvm.vscale.i32()
; DEFAULT-NEXT:    [[TMP8:%.*]] = mul i32 [[TMP7]], 4
; DEFAULT-NEXT:    [[TMP9:%.*]] = sub i32 [[TMP8]], 1
; DEFAULT-NEXT:    [[VECTOR_RECUR_INIT:%.*]] = insertelement <vscale x 4 x i32> poison, i32 0, i32 [[TMP9]]
; DEFAULT-NEXT:    [[TMP10:%.*]] = call i32 @llvm.vscale.i32()
; DEFAULT-NEXT:    [[TMP11:%.*]] = mul i32 [[TMP10]], 4
; DEFAULT-NEXT:    [[TMP12:%.*]] = sub i32 [[TMP11]], 1
; DEFAULT-NEXT:    [[VECTOR_RECUR_INIT1:%.*]] = insertelement <vscale x 4 x i32> poison, i32 0, i32 [[TMP12]]
; DEFAULT-NEXT:    [[BROADCAST_SPLATINSERT6:%.*]] = insertelement <vscale x 4 x i32> poison, i32 [[X]], i64 0
; DEFAULT-NEXT:    [[BROADCAST_SPLAT7:%.*]] = shufflevector <vscale x 4 x i32> [[BROADCAST_SPLATINSERT6]], <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer
; DEFAULT-NEXT:    [[BROADCAST_SPLATINSERT8:%.*]] = insertelement <vscale x 4 x i32> poison, i32 [[Z]], i64 0
; DEFAULT-NEXT:    [[BROADCAST_SPLAT9:%.*]] = shufflevector <vscale x 4 x i32> [[BROADCAST_SPLATINSERT8]], <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer
; DEFAULT-NEXT:    br label [[VECTOR_BODY:%.*]]
; DEFAULT:       vector.body:
; DEFAULT-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; DEFAULT-NEXT:    [[VECTOR_RECUR:%.*]] = phi <vscale x 4 x i32> [ [[VECTOR_RECUR_INIT]], [[VECTOR_PH]] ], [ [[BROADCAST_SPLAT5:%.*]], [[VECTOR_BODY]] ]
; DEFAULT-NEXT:    [[VECTOR_RECUR2:%.*]] = phi <vscale x 4 x i32> [ [[VECTOR_RECUR_INIT1]], [[VECTOR_PH]] ], [ [[TMP20:%.*]], [[VECTOR_BODY]] ]
; DEFAULT-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP57:%.*]], [[VECTOR_BODY]] ]
; DEFAULT-NEXT:    [[VEC_PHI3:%.*]] = phi <vscale x 4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP58:%.*]], [[VECTOR_BODY]] ]
; DEFAULT-NEXT:    [[TMP13:%.*]] = add i64 [[Y]], 1
; DEFAULT-NEXT:    [[TMP14:%.*]] = add i64 [[Y]], 1
; DEFAULT-NEXT:    [[TMP15:%.*]] = getelementptr i32, ptr [[SRC_1]], i64 [[TMP13]]
; DEFAULT-NEXT:    [[TMP16:%.*]] = getelementptr i32, ptr [[SRC_1]], i64 [[TMP14]]
; DEFAULT-NEXT:    [[TMP17:%.*]] = load i32, ptr [[TMP15]], align 4
; DEFAULT-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 4 x i32> poison, i32 [[TMP17]], i64 0
; DEFAULT-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 4 x i32> [[BROADCAST_SPLATINSERT]], <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer
; DEFAULT-NEXT:    [[TMP18:%.*]] = load i32, ptr [[TMP16]], align 4
; DEFAULT-NEXT:    [[BROADCAST_SPLATINSERT4:%.*]] = insertelement <vscale x 4 x i32> poison, i32 [[TMP18]], i64 0
; DEFAULT-NEXT:    [[BROADCAST_SPLAT5]] = shufflevector <vscale x 4 x i32> [[BROADCAST_SPLATINSERT4]], <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer
; DEFAULT-NEXT:    [[TMP19:%.*]] = call <vscale x 4 x i32> @llvm.vector.splice.nxv4i32(<vscale x 4 x i32> [[VECTOR_RECUR]], <vscale x 4 x i32> [[BROADCAST_SPLAT]], i32 -1)
; DEFAULT-NEXT:    [[TMP20]] = call <vscale x 4 x i32> @llvm.vector.splice.nxv4i32(<vscale x 4 x i32> [[BROADCAST_SPLAT]], <vscale x 4 x i32> [[BROADCAST_SPLAT5]], i32 -1)
; DEFAULT-NEXT:    [[TMP21:%.*]] = call <vscale x 4 x i32> @llvm.vector.splice.nxv4i32(<vscale x 4 x i32> [[VECTOR_RECUR2]], <vscale x 4 x i32> [[TMP19]], i32 -1)
; DEFAULT-NEXT:    [[TMP22:%.*]] = call <vscale x 4 x i32> @llvm.vector.splice.nxv4i32(<vscale x 4 x i32> [[TMP19]], <vscale x 4 x i32> [[TMP20]], i32 -1)
; DEFAULT-NEXT:    [[TMP23:%.*]] = or <vscale x 4 x i32> [[TMP21]], [[BROADCAST_SPLAT7]]
; DEFAULT-NEXT:    [[TMP24:%.*]] = or <vscale x 4 x i32> [[TMP22]], [[BROADCAST_SPLAT7]]
; DEFAULT-NEXT:    [[TMP25:%.*]] = lshr <vscale x 4 x i32> [[BROADCAST_SPLAT7]], shufflevector (<vscale x 4 x i32> insertelement (<vscale x 4 x i32> poison, i32 1, i64 0), <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer)
; DEFAULT-NEXT:    [[TMP26:%.*]] = lshr <vscale x 4 x i32> [[BROADCAST_SPLAT7]], shufflevector (<vscale x 4 x i32> insertelement (<vscale x 4 x i32> poison, i32 1, i64 0), <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer)
; DEFAULT-NEXT:    [[TMP27:%.*]] = shl <vscale x 4 x i32> [[TMP23]], shufflevector (<vscale x 4 x i32> insertelement (<vscale x 4 x i32> poison, i32 1, i64 0), <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer)
; DEFAULT-NEXT:    [[TMP28:%.*]] = shl <vscale x 4 x i32> [[TMP24]], shufflevector (<vscale x 4 x i32> insertelement (<vscale x 4 x i32> poison, i32 1, i64 0), <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer)
; DEFAULT-NEXT:    [[TMP29:%.*]] = or <vscale x 4 x i32> [[TMP27]], shufflevector (<vscale x 4 x i32> insertelement (<vscale x 4 x i32> poison, i32 2, i64 0), <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer)
; DEFAULT-NEXT:    [[TMP30:%.*]] = or <vscale x 4 x i32> [[TMP28]], shufflevector (<vscale x 4 x i32> insertelement (<vscale x 4 x i32> poison, i32 2, i64 0), <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer)
; DEFAULT-NEXT:    [[TMP31:%.*]] = shl <vscale x 4 x i32> [[BROADCAST_SPLAT7]], shufflevector (<vscale x 4 x i32> insertelement (<vscale x 4 x i32> poison, i32 1, i64 0), <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer)
; DEFAULT-NEXT:    [[TMP32:%.*]] = shl <vscale x 4 x i32> [[BROADCAST_SPLAT7]], shufflevector (<vscale x 4 x i32> insertelement (<vscale x 4 x i32> poison, i32 1, i64 0), <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer)
; DEFAULT-NEXT:    [[TMP33:%.*]] = or <vscale x 4 x i32> [[TMP25]], [[TMP31]]
; DEFAULT-NEXT:    [[TMP34:%.*]] = or <vscale x 4 x i32> [[TMP26]], [[TMP32]]
; DEFAULT-NEXT:    [[TMP35:%.*]] = or <vscale x 4 x i32> [[TMP33]], [[TMP29]]
; DEFAULT-NEXT:    [[TMP36:%.*]] = or <vscale x 4 x i32> [[TMP34]], [[TMP30]]
; DEFAULT-NEXT:    [[TMP37:%.*]] = or <vscale x 4 x i32> [[TMP35]], [[BROADCAST_SPLAT7]]
; DEFAULT-NEXT:    [[TMP38:%.*]] = or <vscale x 4 x i32> [[TMP36]], [[BROADCAST_SPLAT7]]
; DEFAULT-NEXT:    [[TMP39:%.*]] = or <vscale x 4 x i32> [[BROADCAST_SPLAT9]], [[BROADCAST_SPLAT7]]
; DEFAULT-NEXT:    [[TMP40:%.*]] = or <vscale x 4 x i32> [[BROADCAST_SPLAT9]], [[BROADCAST_SPLAT7]]
; DEFAULT-NEXT:    [[TMP41:%.*]] = and <vscale x 4 x i32> [[TMP39]], shufflevector (<vscale x 4 x i32> insertelement (<vscale x 4 x i32> poison, i32 1, i64 0), <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer)
; DEFAULT-NEXT:    [[TMP42:%.*]] = and <vscale x 4 x i32> [[TMP40]], shufflevector (<vscale x 4 x i32> insertelement (<vscale x 4 x i32> poison, i32 1, i64 0), <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer)
; DEFAULT-NEXT:    [[TMP43:%.*]] = xor <vscale x 4 x i32> [[TMP41]], shufflevector (<vscale x 4 x i32> insertelement (<vscale x 4 x i32> poison, i32 1, i64 0), <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer)
; DEFAULT-NEXT:    [[TMP44:%.*]] = xor <vscale x 4 x i32> [[TMP42]], shufflevector (<vscale x 4 x i32> insertelement (<vscale x 4 x i32> poison, i32 1, i64 0), <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer)
; DEFAULT-NEXT:    [[TMP45:%.*]] = zext <vscale x 4 x i32> [[TMP43]] to <vscale x 4 x i64>
; DEFAULT-NEXT:    [[TMP46:%.*]] = zext <vscale x 4 x i32> [[TMP44]] to <vscale x 4 x i64>
; DEFAULT-NEXT:    [[TMP47:%.*]] = getelementptr i32, ptr [[SRC_2]], <vscale x 4 x i64> [[TMP45]]
; DEFAULT-NEXT:    [[TMP48:%.*]] = getelementptr i32, ptr [[SRC_2]], <vscale x 4 x i64> [[TMP46]]
; DEFAULT-NEXT:    [[WIDE_MASKED_GATHER:%.*]] = call <vscale x 4 x i32> @llvm.masked.gather.nxv4i32.nxv4p0(<vscale x 4 x ptr> [[TMP47]], i32 4, <vscale x 4 x i1> shufflevector (<vscale x 4 x i1> insertelement (<vscale x 4 x i1> poison, i1 true, i64 0), <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer), <vscale x 4 x i32> poison)
; DEFAULT-NEXT:    [[WIDE_MASKED_GATHER10:%.*]] = call <vscale x 4 x i32> @llvm.masked.gather.nxv4i32.nxv4p0(<vscale x 4 x ptr> [[TMP48]], i32 4, <vscale x 4 x i1> shufflevector (<vscale x 4 x i1> insertelement (<vscale x 4 x i1> poison, i1 true, i64 0), <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer), <vscale x 4 x i32> poison)
; DEFAULT-NEXT:    [[TMP49:%.*]] = lshr <vscale x 4 x i32> [[TMP37]], shufflevector (<vscale x 4 x i32> insertelement (<vscale x 4 x i32> poison, i32 1, i64 0), <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer)
; DEFAULT-NEXT:    [[TMP50:%.*]] = lshr <vscale x 4 x i32> [[TMP38]], shufflevector (<vscale x 4 x i32> insertelement (<vscale x 4 x i32> poison, i32 1, i64 0), <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer)
; DEFAULT-NEXT:    [[TMP51:%.*]] = zext <vscale x 4 x i32> [[TMP49]] to <vscale x 4 x i64>
; DEFAULT-NEXT:    [[TMP52:%.*]] = zext <vscale x 4 x i32> [[TMP50]] to <vscale x 4 x i64>
; DEFAULT-NEXT:    [[TMP53:%.*]] = getelementptr i32, ptr [[SRC_2]], <vscale x 4 x i64> [[TMP51]]
; DEFAULT-NEXT:    [[TMP54:%.*]] = getelementptr i32, ptr [[SRC_2]], <vscale x 4 x i64> [[TMP52]]
; DEFAULT-NEXT:    [[WIDE_MASKED_GATHER11:%.*]] = call <vscale x 4 x i32> @llvm.masked.gather.nxv4i32.nxv4p0(<vscale x 4 x ptr> [[TMP53]], i32 4, <vscale x 4 x i1> shufflevector (<vscale x 4 x i1> insertelement (<vscale x 4 x i1> poison, i1 true, i64 0), <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer), <vscale x 4 x i32> poison)
; DEFAULT-NEXT:    [[WIDE_MASKED_GATHER12:%.*]] = call <vscale x 4 x i32> @llvm.masked.gather.nxv4i32.nxv4p0(<vscale x 4 x ptr> [[TMP54]], i32 4, <vscale x 4 x i1> shufflevector (<vscale x 4 x i1> insertelement (<vscale x 4 x i1> poison, i1 true, i64 0), <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer), <vscale x 4 x i32> poison)
; DEFAULT-NEXT:    [[TMP55:%.*]] = or <vscale x 4 x i32> [[WIDE_MASKED_GATHER]], [[VEC_PHI]]
; DEFAULT-NEXT:    [[TMP56:%.*]] = or <vscale x 4 x i32> [[WIDE_MASKED_GATHER10]], [[VEC_PHI3]]
; DEFAULT-NEXT:    [[TMP57]] = or <vscale x 4 x i32> [[TMP55]], [[WIDE_MASKED_GATHER11]]
; DEFAULT-NEXT:    [[TMP58]] = or <vscale x 4 x i32> [[TMP56]], [[WIDE_MASKED_GATHER12]]
; DEFAULT-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP6]]
; DEFAULT-NEXT:    [[TMP59:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; DEFAULT-NEXT:    br i1 [[TMP59]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
; DEFAULT:       middle.block:
; DEFAULT-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP0]], [[N_VEC]]
; DEFAULT-NEXT:    [[BIN_RDX:%.*]] = or <vscale x 4 x i32> [[TMP58]], [[TMP57]]
; DEFAULT-NEXT:    [[TMP60:%.*]] = call i32 @llvm.vector.reduce.or.nxv4i32(<vscale x 4 x i32> [[BIN_RDX]])
; DEFAULT-NEXT:    [[TMP64:%.*]] = call i32 @llvm.vscale.i32()
; DEFAULT-NEXT:    [[TMP65:%.*]] = mul i32 [[TMP64]], 4
; DEFAULT-NEXT:    [[TMP66:%.*]] = sub i32 [[TMP65]], 1
; DEFAULT-NEXT:    [[VECTOR_RECUR_EXTRACT13:%.*]] = extractelement <vscale x 4 x i32> [[TMP20]], i32 [[TMP66]]
; DEFAULT-NEXT:    br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]
; DEFAULT:       scalar.ph:
; DEFAULT-NEXT:    [[SCALAR_RECUR_INIT14:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[VECTOR_RECUR_EXTRACT13]], [[MIDDLE_BLOCK]] ]
; DEFAULT-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ 0, [[ENTRY]] ], [ [[TMP18]], [[MIDDLE_BLOCK]] ]
; DEFAULT-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
; DEFAULT-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ 0, [[ENTRY]] ], [ [[TMP60]], [[MIDDLE_BLOCK]] ]
; DEFAULT-NEXT:    br label [[LOOP:%.*]]
; DEFAULT:       loop:
; DEFAULT-NEXT:    [[SCALAR_RECUR:%.*]] = phi i32 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[TMP68:%.*]], [[LOOP]] ]
; DEFAULT-NEXT:    [[SCALAR_RECUR15:%.*]] = phi i32 [ [[SCALAR_RECUR_INIT14]], [[SCALAR_PH]] ], [ [[SCALAR_RECUR]], [[LOOP]] ]
; DEFAULT-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP]] ]
; DEFAULT-NEXT:    [[SUM_RED:%.*]] = phi i32 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[RED_2:%.*]], [[LOOP]] ]
; DEFAULT-NEXT:    [[TMP67:%.*]] = add i64 [[Y]], 1
; DEFAULT-NEXT:    [[GEP_1:%.*]] = getelementptr i32, ptr [[SRC_1]], i64 [[TMP67]]
; DEFAULT-NEXT:    [[TMP68]] = load i32, ptr [[GEP_1]], align 4
; DEFAULT-NEXT:    [[OR3:%.*]] = or i32 [[SCALAR_RECUR15]], [[X]]
; DEFAULT-NEXT:    [[IV_NEXT]] = add i64 [[IV]], 1
; DEFAULT-NEXT:    [[SHR:%.*]] = lshr i32 [[X]], 1
; DEFAULT-NEXT:    [[TMP69:%.*]] = shl i32 [[OR3]], 1
; DEFAULT-NEXT:    [[TMP70:%.*]] = or i32 [[TMP69]], 2
; DEFAULT-NEXT:    [[SHL19:%.*]] = shl i32 [[X]], 1
; DEFAULT-NEXT:    [[TMP71:%.*]] = or i32 [[SHR]], [[SHL19]]
; DEFAULT-NEXT:    [[TMP72:%.*]] = or i32 [[TMP71]], [[TMP70]]
; DEFAULT-NEXT:    [[TMP73:%.*]] = or i32 [[TMP72]], [[X]]
; DEFAULT-NEXT:    [[OR20:%.*]] = or i32 [[Z]], [[X]]
; DEFAULT-NEXT:    [[NOT:%.*]] = and i32 [[OR20]], 1
; DEFAULT-NEXT:    [[AND:%.*]] = xor i32 [[NOT]], 1
; DEFAULT-NEXT:    [[IDX_EXT_1:%.*]] = zext i32 [[AND]] to i64
; DEFAULT-NEXT:    [[GEP_2:%.*]] = getelementptr i32, ptr [[SRC_2]], i64 [[IDX_EXT_1]]
; DEFAULT-NEXT:    [[TMP74:%.*]] = load i32, ptr [[GEP_2]], align 4
; DEFAULT-NEXT:    [[SHR24:%.*]] = lshr i32 [[TMP73]], 1
; DEFAULT-NEXT:    [[IDX_EXT_2:%.*]] = zext i32 [[SHR24]] to i64
; DEFAULT-NEXT:    [[GEP_3:%.*]] = getelementptr i32, ptr [[SRC_2]], i64 [[IDX_EXT_2]]
; DEFAULT-NEXT:    [[TMP75:%.*]] = load i32, ptr [[GEP_3]], align 4
; DEFAULT-NEXT:    [[RED_1:%.*]] = or i32 [[TMP74]], [[SUM_RED]]
; DEFAULT-NEXT:    [[RED_2]] = or i32 [[RED_1]], [[TMP75]]
; DEFAULT-NEXT:    [[EC:%.*]] = icmp eq i64 [[IV]], [[Y]]
; DEFAULT-NEXT:    br i1 [[EC]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP3:![0-9]+]]
; DEFAULT:       exit:
; DEFAULT-NEXT:    [[RED_2_LCSSA:%.*]] = phi i32 [ [[RED_2]], [[LOOP]] ], [ [[TMP60]], [[MIDDLE_BLOCK]] ]
; DEFAULT-NEXT:    ret i32 [[RED_2_LCSSA]]
;
; PRED-LABEL: define i32 @chained_recurrences(
; PRED-SAME: i32 [[X:%.*]], i64 [[Y:%.*]], ptr [[SRC_1:%.*]], i32 [[Z:%.*]], ptr [[SRC_2:%.*]]) #[[ATTR0:[0-9]+]] {
; PRED-NEXT:  entry:
; PRED-NEXT:    [[TMP0:%.*]] = add i64 [[Y]], 1
; PRED-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; PRED:       vector.ph:
; PRED-NEXT:    [[TMP1:%.*]] = call i64 @llvm.vscale.i64()
; PRED-NEXT:    [[TMP2:%.*]] = mul i64 [[TMP1]], 4
; PRED-NEXT:    [[TMP5:%.*]] = sub i64 [[TMP2]], 1
; PRED-NEXT:    [[N_RND_UP:%.*]] = add i64 [[TMP0]], [[TMP5]]
; PRED-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N_RND_UP]], [[TMP2]]
; PRED-NEXT:    [[N_VEC:%.*]] = sub i64 [[N_RND_UP]], [[N_MOD_VF]]
; PRED-NEXT:    [[TMP6:%.*]] = call i64 @llvm.vscale.i64()
; PRED-NEXT:    [[TMP7:%.*]] = mul i64 [[TMP6]], 4
; PRED-NEXT:    [[TMP8:%.*]] = call i64 @llvm.vscale.i64()
; PRED-NEXT:    [[TMP9:%.*]] = mul i64 [[TMP8]], 4
; PRED-NEXT:    [[TMP10:%.*]] = sub i64 [[TMP0]], [[TMP9]]
; PRED-NEXT:    [[TMP11:%.*]] = icmp ugt i64 [[TMP0]], [[TMP9]]
; PRED-NEXT:    [[TMP12:%.*]] = select i1 [[TMP11]], i64 [[TMP10]], i64 0
; PRED-NEXT:    [[ACTIVE_LANE_MASK_ENTRY:%.*]] = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i64(i64 0, i64 [[TMP0]])
; PRED-NEXT:    [[TMP13:%.*]] = call i32 @llvm.vscale.i32()
; PRED-NEXT:    [[TMP14:%.*]] = mul i32 [[TMP13]], 4
; PRED-NEXT:    [[TMP15:%.*]] = sub i32 [[TMP14]], 1
; PRED-NEXT:    [[VECTOR_RECUR_INIT:%.*]] = insertelement <vscale x 4 x i32> poison, i32 0, i32 [[TMP15]]
; PRED-NEXT:    [[TMP16:%.*]] = call i32 @llvm.vscale.i32()
; PRED-NEXT:    [[TMP17:%.*]] = mul i32 [[TMP16]], 4
; PRED-NEXT:    [[TMP18:%.*]] = sub i32 [[TMP17]], 1
; PRED-NEXT:    [[VECTOR_RECUR_INIT1:%.*]] = insertelement <vscale x 4 x i32> poison, i32 0, i32 [[TMP18]]
; PRED-NEXT:    [[BROADCAST_SPLATINSERT3:%.*]] = insertelement <vscale x 4 x i32> poison, i32 [[X]], i64 0
; PRED-NEXT:    [[BROADCAST_SPLAT4:%.*]] = shufflevector <vscale x 4 x i32> [[BROADCAST_SPLATINSERT3]], <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer
; PRED-NEXT:    [[BROADCAST_SPLATINSERT5:%.*]] = insertelement <vscale x 4 x i32> poison, i32 [[Z]], i64 0
; PRED-NEXT:    [[BROADCAST_SPLAT6:%.*]] = shufflevector <vscale x 4 x i32> [[BROADCAST_SPLATINSERT5]], <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer
; PRED-NEXT:    br label [[VECTOR_BODY:%.*]]
; PRED:       vector.body:
; PRED-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; PRED-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = phi <vscale x 4 x i1> [ [[ACTIVE_LANE_MASK_ENTRY]], [[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT:%.*]], [[VECTOR_BODY]] ]
; PRED-NEXT:    [[VECTOR_RECUR:%.*]] = phi <vscale x 4 x i32> [ [[VECTOR_RECUR_INIT]], [[VECTOR_PH]] ], [ [[BROADCAST_SPLAT:%.*]], [[VECTOR_BODY]] ]
; PRED-NEXT:    [[VECTOR_RECUR2:%.*]] = phi <vscale x 4 x i32> [ [[VECTOR_RECUR_INIT1]], [[VECTOR_PH]] ], [ [[TMP22:%.*]], [[VECTOR_BODY]] ]
; PRED-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP42:%.*]], [[VECTOR_BODY]] ]
; PRED-NEXT:    [[TMP19:%.*]] = add i64 [[Y]], 1
; PRED-NEXT:    [[TMP20:%.*]] = getelementptr i32, ptr [[SRC_1]], i64 [[TMP19]]
; PRED-NEXT:    [[TMP21:%.*]] = load i32, ptr [[TMP20]], align 4
; PRED-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 4 x i32> poison, i32 [[TMP21]], i64 0
; PRED-NEXT:    [[BROADCAST_SPLAT]] = shufflevector <vscale x 4 x i32> [[BROADCAST_SPLATINSERT]], <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer
; PRED-NEXT:    [[TMP22]] = call <vscale x 4 x i32> @llvm.vector.splice.nxv4i32(<vscale x 4 x i32> [[VECTOR_RECUR]], <vscale x 4 x i32> [[BROADCAST_SPLAT]], i32 -1)
; PRED-NEXT:    [[TMP23:%.*]] = call <vscale x 4 x i32> @llvm.vector.splice.nxv4i32(<vscale x 4 x i32> [[VECTOR_RECUR2]], <vscale x 4 x i32> [[TMP22]], i32 -1)
; PRED-NEXT:    [[TMP24:%.*]] = or <vscale x 4 x i32> [[TMP23]], [[BROADCAST_SPLAT4]]
; PRED-NEXT:    [[TMP25:%.*]] = lshr <vscale x 4 x i32> [[BROADCAST_SPLAT4]], shufflevector (<vscale x 4 x i32> insertelement (<vscale x 4 x i32> poison, i32 1, i64 0), <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer)
; PRED-NEXT:    [[TMP26:%.*]] = shl <vscale x 4 x i32> [[TMP24]], shufflevector (<vscale x 4 x i32> insertelement (<vscale x 4 x i32> poison, i32 1, i64 0), <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer)
; PRED-NEXT:    [[TMP27:%.*]] = or <vscale x 4 x i32> [[TMP26]], shufflevector (<vscale x 4 x i32> insertelement (<vscale x 4 x i32> poison, i32 2, i64 0), <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer)
; PRED-NEXT:    [[TMP28:%.*]] = shl <vscale x 4 x i32> [[BROADCAST_SPLAT4]], shufflevector (<vscale x 4 x i32> insertelement (<vscale x 4 x i32> poison, i32 1, i64 0), <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer)
; PRED-NEXT:    [[TMP29:%.*]] = or <vscale x 4 x i32> [[TMP25]], [[TMP28]]
; PRED-NEXT:    [[TMP30:%.*]] = or <vscale x 4 x i32> [[TMP29]], [[TMP27]]
; PRED-NEXT:    [[TMP31:%.*]] = or <vscale x 4 x i32> [[TMP30]], [[BROADCAST_SPLAT4]]
; PRED-NEXT:    [[TMP32:%.*]] = or <vscale x 4 x i32> [[BROADCAST_SPLAT6]], [[BROADCAST_SPLAT4]]
; PRED-NEXT:    [[TMP33:%.*]] = and <vscale x 4 x i32> [[TMP32]], shufflevector (<vscale x 4 x i32> insertelement (<vscale x 4 x i32> poison, i32 1, i64 0), <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer)
; PRED-NEXT:    [[TMP34:%.*]] = xor <vscale x 4 x i32> [[TMP33]], shufflevector (<vscale x 4 x i32> insertelement (<vscale x 4 x i32> poison, i32 1, i64 0), <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer)
; PRED-NEXT:    [[TMP35:%.*]] = zext <vscale x 4 x i32> [[TMP34]] to <vscale x 4 x i64>
; PRED-NEXT:    [[TMP36:%.*]] = getelementptr i32, ptr [[SRC_2]], <vscale x 4 x i64> [[TMP35]]
; PRED-NEXT:    [[WIDE_MASKED_GATHER:%.*]] = call <vscale x 4 x i32> @llvm.masked.gather.nxv4i32.nxv4p0(<vscale x 4 x ptr> [[TMP36]], i32 4, <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x i32> poison)
; PRED-NEXT:    [[TMP37:%.*]] = lshr <vscale x 4 x i32> [[TMP31]], shufflevector (<vscale x 4 x i32> insertelement (<vscale x 4 x i32> poison, i32 1, i64 0), <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer)
; PRED-NEXT:    [[TMP38:%.*]] = zext <vscale x 4 x i32> [[TMP37]] to <vscale x 4 x i64>
; PRED-NEXT:    [[TMP39:%.*]] = getelementptr i32, ptr [[SRC_2]], <vscale x 4 x i64> [[TMP38]]
; PRED-NEXT:    [[WIDE_MASKED_GATHER7:%.*]] = call <vscale x 4 x i32> @llvm.masked.gather.nxv4i32.nxv4p0(<vscale x 4 x ptr> [[TMP39]], i32 4, <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x i32> poison)
; PRED-NEXT:    [[TMP40:%.*]] = or <vscale x 4 x i32> [[WIDE_MASKED_GATHER]], [[VEC_PHI]]
; PRED-NEXT:    [[TMP41:%.*]] = or <vscale x 4 x i32> [[TMP40]], [[WIDE_MASKED_GATHER7]]
; PRED-NEXT:    [[TMP42]] = select <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x i32> [[TMP41]], <vscale x 4 x i32> [[VEC_PHI]]
; PRED-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP7]]
; PRED-NEXT:    [[ACTIVE_LANE_MASK_NEXT]] = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i64(i64 [[INDEX]], i64 [[TMP12]])
; PRED-NEXT:    [[TMP43:%.*]] = xor <vscale x 4 x i1> [[ACTIVE_LANE_MASK_NEXT]], shufflevector (<vscale x 4 x i1> insertelement (<vscale x 4 x i1> poison, i1 true, i64 0), <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer)
; PRED-NEXT:    [[TMP44:%.*]] = extractelement <vscale x 4 x i1> [[TMP43]], i32 0
; PRED-NEXT:    br i1 [[TMP44]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
; PRED:       middle.block:
; PRED-NEXT:    [[TMP45:%.*]] = call i32 @llvm.vector.reduce.or.nxv4i32(<vscale x 4 x i32> [[TMP42]])
; PRED-NEXT:    [[TMP49:%.*]] = call i32 @llvm.vscale.i32()
; PRED-NEXT:    [[TMP50:%.*]] = mul i32 [[TMP49]], 4
; PRED-NEXT:    [[TMP51:%.*]] = sub i32 [[TMP50]], 1
; PRED-NEXT:    [[VECTOR_RECUR_EXTRACT8:%.*]] = extractelement <vscale x 4 x i32> [[TMP22]], i32 [[TMP51]]
; PRED-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
; PRED:       scalar.ph:
; PRED-NEXT:    [[SCALAR_RECUR_INIT9:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[VECTOR_RECUR_EXTRACT8]], [[MIDDLE_BLOCK]] ]
; PRED-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ 0, [[ENTRY]] ], [ [[TMP21]], [[MIDDLE_BLOCK]] ]
; PRED-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
; PRED-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ 0, [[ENTRY]] ], [ [[TMP45]], [[MIDDLE_BLOCK]] ]
; PRED-NEXT:    br label [[LOOP:%.*]]
; PRED:       loop:
; PRED-NEXT:    [[SCALAR_RECUR:%.*]] = phi i32 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[TMP53:%.*]], [[LOOP]] ]
; PRED-NEXT:    [[SCALAR_RECUR10:%.*]] = phi i32 [ [[SCALAR_RECUR_INIT9]], [[SCALAR_PH]] ], [ [[SCALAR_RECUR]], [[LOOP]] ]
; PRED-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP]] ]
; PRED-NEXT:    [[SUM_RED:%.*]] = phi i32 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[RED_2:%.*]], [[LOOP]] ]
; PRED-NEXT:    [[TMP52:%.*]] = add i64 [[Y]], 1
; PRED-NEXT:    [[GEP_1:%.*]] = getelementptr i32, ptr [[SRC_1]], i64 [[TMP52]]
; PRED-NEXT:    [[TMP53]] = load i32, ptr [[GEP_1]], align 4
; PRED-NEXT:    [[OR3:%.*]] = or i32 [[SCALAR_RECUR10]], [[X]]
; PRED-NEXT:    [[IV_NEXT]] = add i64 [[IV]], 1
; PRED-NEXT:    [[SHR:%.*]] = lshr i32 [[X]], 1
; PRED-NEXT:    [[TMP54:%.*]] = shl i32 [[OR3]], 1
; PRED-NEXT:    [[TMP55:%.*]] = or i32 [[TMP54]], 2
; PRED-NEXT:    [[SHL19:%.*]] = shl i32 [[X]], 1
; PRED-NEXT:    [[TMP56:%.*]] = or i32 [[SHR]], [[SHL19]]
; PRED-NEXT:    [[TMP57:%.*]] = or i32 [[TMP56]], [[TMP55]]
; PRED-NEXT:    [[TMP58:%.*]] = or i32 [[TMP57]], [[X]]
; PRED-NEXT:    [[OR20:%.*]] = or i32 [[Z]], [[X]]
; PRED-NEXT:    [[NOT:%.*]] = and i32 [[OR20]], 1
; PRED-NEXT:    [[AND:%.*]] = xor i32 [[NOT]], 1
; PRED-NEXT:    [[IDX_EXT_1:%.*]] = zext i32 [[AND]] to i64
; PRED-NEXT:    [[GEP_2:%.*]] = getelementptr i32, ptr [[SRC_2]], i64 [[IDX_EXT_1]]
; PRED-NEXT:    [[TMP59:%.*]] = load i32, ptr [[GEP_2]], align 4
; PRED-NEXT:    [[SHR24:%.*]] = lshr i32 [[TMP58]], 1
; PRED-NEXT:    [[IDX_EXT_2:%.*]] = zext i32 [[SHR24]] to i64
; PRED-NEXT:    [[GEP_3:%.*]] = getelementptr i32, ptr [[SRC_2]], i64 [[IDX_EXT_2]]
; PRED-NEXT:    [[TMP60:%.*]] = load i32, ptr [[GEP_3]], align 4
; PRED-NEXT:    [[RED_1:%.*]] = or i32 [[TMP59]], [[SUM_RED]]
; PRED-NEXT:    [[RED_2]] = or i32 [[RED_1]], [[TMP60]]
; PRED-NEXT:    [[EC:%.*]] = icmp eq i64 [[IV]], [[Y]]
; PRED-NEXT:    br i1 [[EC]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP3:![0-9]+]]
; PRED:       exit:
; PRED-NEXT:    [[RED_2_LCSSA:%.*]] = phi i32 [ [[RED_2]], [[LOOP]] ], [ [[TMP45]], [[MIDDLE_BLOCK]] ]
; PRED-NEXT:    ret i32 [[RED_2_LCSSA]]
;
entry:
  br label %loop

loop:
  %2 = phi i32 [ 0, %entry ], [ %5, %loop ]
  %3 = phi i32 [ 0, %entry ], [ %2, %loop ]
  %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]
  %sum.red = phi i32 [ 0, %entry ], [ %red.2, %loop ]
  %4 = add i64 %y, 1
  %gep.1 = getelementptr i32, ptr %src.1, i64 %4
  %5 = load i32, ptr %gep.1, align 4
  %or3 = or i32 %3, %x
  %iv.next = add i64 %iv, 1
  %shr = lshr i32 %x, 1
  %6 = shl i32 %or3, 1
  %7 = or i32 %6, 2
  %shl19 = shl i32 %x, 1
  %8 = or i32 %shr, %shl19
  %9 = or i32 %8, %7
  %10 = or i32 %9, %x
  %or20 = or i32 %z, %x
  %not = and i32 %or20, 1
  %and = xor i32 %not, 1
  %idx.ext.1 = zext i32 %and to i64
  %gep.2 = getelementptr i32, ptr %src.2, i64 %idx.ext.1
  %11 = load i32, ptr %gep.2, align 4
  %shr24 = lshr i32 %10, 1
  %idx.ext.2 = zext i32 %shr24 to i64
  %gep.3 = getelementptr i32, ptr %src.2, i64 %idx.ext.2
  %12 = load i32, ptr %gep.3, align 4
  %red.1 = or i32 %11, %sum.red
  %red.2 = or i32 %red.1, %12
  %ec = icmp eq i64 %iv, %y
  br i1 %ec, label %exit, label %loop

exit:
  ret i32 %red.2
}

define i16 @reduce_udiv(ptr %src, i16 %x, i64 %N) #0 {
; DEFAULT-LABEL: define i16 @reduce_udiv(
; DEFAULT-SAME: ptr [[SRC:%.*]], i16 [[X:%.*]], i64 [[N:%.*]]) #[[ATTR0]] {
; DEFAULT-NEXT:  entry:
; DEFAULT-NEXT:    [[TMP0:%.*]] = add i64 [[N]], 1
; DEFAULT-NEXT:    [[TMP1:%.*]] = call i64 @llvm.vscale.i64()
; DEFAULT-NEXT:    [[TMP2:%.*]] = mul i64 [[TMP1]], 8
; DEFAULT-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP0]], [[TMP2]]
; DEFAULT-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; DEFAULT:       vector.ph:
; DEFAULT-NEXT:    [[TMP3:%.*]] = call i64 @llvm.vscale.i64()
; DEFAULT-NEXT:    [[TMP4:%.*]] = mul i64 [[TMP3]], 8
; DEFAULT-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[TMP0]], [[TMP4]]
; DEFAULT-NEXT:    [[N_VEC:%.*]] = sub i64 [[TMP0]], [[N_MOD_VF]]
; DEFAULT-NEXT:    [[TMP5:%.*]] = call i64 @llvm.vscale.i64()
; DEFAULT-NEXT:    [[TMP6:%.*]] = mul i64 [[TMP5]], 8
; DEFAULT-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 4 x i16> poison, i16 [[X]], i64 0
; DEFAULT-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 4 x i16> [[BROADCAST_SPLATINSERT]], <vscale x 4 x i16> poison, <vscale x 4 x i32> zeroinitializer
; DEFAULT-NEXT:    br label [[VECTOR_BODY:%.*]]
; DEFAULT:       vector.body:
; DEFAULT-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; DEFAULT-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x i16> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP21:%.*]], [[VECTOR_BODY]] ]
; DEFAULT-NEXT:    [[VEC_PHI1:%.*]] = phi <vscale x 4 x i16> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP22:%.*]], [[VECTOR_BODY]] ]
; DEFAULT-NEXT:    [[TMP7:%.*]] = add i64 [[INDEX]], 0
; DEFAULT-NEXT:    [[TMP8:%.*]] = call i64 @llvm.vscale.i64()
; DEFAULT-NEXT:    [[TMP9:%.*]] = mul i64 [[TMP8]], 4
; DEFAULT-NEXT:    [[TMP10:%.*]] = add i64 [[TMP9]], 0
; DEFAULT-NEXT:    [[TMP11:%.*]] = mul i64 [[TMP10]], 1
; DEFAULT-NEXT:    [[TMP12:%.*]] = add i64 [[INDEX]], [[TMP11]]
; DEFAULT-NEXT:    [[TMP13:%.*]] = getelementptr i16, ptr [[SRC]], i64 [[TMP7]]
; DEFAULT-NEXT:    [[TMP14:%.*]] = getelementptr i16, ptr [[SRC]], i64 [[TMP12]]
; DEFAULT-NEXT:    [[TMP15:%.*]] = getelementptr i16, ptr [[TMP13]], i32 0
; DEFAULT-NEXT:    [[TMP16:%.*]] = call i64 @llvm.vscale.i64()
; DEFAULT-NEXT:    [[TMP17:%.*]] = mul i64 [[TMP16]], 4
; DEFAULT-NEXT:    [[TMP18:%.*]] = getelementptr i16, ptr [[TMP13]], i64 [[TMP17]]
; DEFAULT-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i16>, ptr [[TMP15]], align 2
; DEFAULT-NEXT:    [[WIDE_LOAD2:%.*]] = load <vscale x 4 x i16>, ptr [[TMP18]], align 2
; DEFAULT-NEXT:    [[TMP19:%.*]] = udiv <vscale x 4 x i16> [[WIDE_LOAD]], [[BROADCAST_SPLAT]]
; DEFAULT-NEXT:    [[TMP20:%.*]] = udiv <vscale x 4 x i16> [[WIDE_LOAD2]], [[BROADCAST_SPLAT]]
; DEFAULT-NEXT:    [[TMP21]] = or <vscale x 4 x i16> [[TMP19]], [[VEC_PHI]]
; DEFAULT-NEXT:    [[TMP22]] = or <vscale x 4 x i16> [[TMP20]], [[VEC_PHI1]]
; DEFAULT-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP6]]
; DEFAULT-NEXT:    [[TMP23:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; DEFAULT-NEXT:    br i1 [[TMP23]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
; DEFAULT:       middle.block:
; DEFAULT-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP0]], [[N_VEC]]
; DEFAULT-NEXT:    [[BIN_RDX:%.*]] = or <vscale x 4 x i16> [[TMP22]], [[TMP21]]
; DEFAULT-NEXT:    [[TMP24:%.*]] = call i16 @llvm.vector.reduce.or.nxv4i16(<vscale x 4 x i16> [[BIN_RDX]])
; DEFAULT-NEXT:    br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]
; DEFAULT:       scalar.ph:
; DEFAULT-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
; DEFAULT-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i16 [ 0, [[ENTRY]] ], [ [[TMP24]], [[MIDDLE_BLOCK]] ]
; DEFAULT-NEXT:    br label [[LOOP:%.*]]
; DEFAULT:       loop:
; DEFAULT-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP]] ]
; DEFAULT-NEXT:    [[RED:%.*]] = phi i16 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[RED_NEXT:%.*]], [[LOOP]] ]
; DEFAULT-NEXT:    [[GEP:%.*]] = getelementptr i16, ptr [[SRC]], i64 [[IV]]
; DEFAULT-NEXT:    [[L:%.*]] = load i16, ptr [[GEP]], align 2
; DEFAULT-NEXT:    [[DIV:%.*]] = udiv i16 [[L]], [[X]]
; DEFAULT-NEXT:    [[RED_NEXT]] = or i16 [[DIV]], [[RED]]
; DEFAULT-NEXT:    [[IV_NEXT]] = add i64 [[IV]], 1
; DEFAULT-NEXT:    [[EC:%.*]] = icmp eq i64 [[IV]], [[N]]
; DEFAULT-NEXT:    br i1 [[EC]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP5:![0-9]+]]
; DEFAULT:       exit:
; DEFAULT-NEXT:    [[RED_NEXT_LCSSA:%.*]] = phi i16 [ [[RED_NEXT]], [[LOOP]] ], [ [[TMP24]], [[MIDDLE_BLOCK]] ]
; DEFAULT-NEXT:    ret i16 [[RED_NEXT_LCSSA]]
;
; PRED-LABEL: define i16 @reduce_udiv(
; PRED-SAME: ptr [[SRC:%.*]], i16 [[X:%.*]], i64 [[N:%.*]]) #[[ATTR0]] {
; PRED-NEXT:  entry:
; PRED-NEXT:    [[TMP0:%.*]] = add i64 [[N]], 1
; PRED-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; PRED:       vector.ph:
; PRED-NEXT:    [[TMP1:%.*]] = call i64 @llvm.vscale.i64()
; PRED-NEXT:    [[TMP2:%.*]] = mul i64 [[TMP1]], 8
; PRED-NEXT:    [[TMP5:%.*]] = sub i64 [[TMP2]], 1
; PRED-NEXT:    [[N_RND_UP:%.*]] = add i64 [[TMP0]], [[TMP5]]
; PRED-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N_RND_UP]], [[TMP2]]
; PRED-NEXT:    [[N_VEC:%.*]] = sub i64 [[N_RND_UP]], [[N_MOD_VF]]
; PRED-NEXT:    [[TMP6:%.*]] = call i64 @llvm.vscale.i64()
; PRED-NEXT:    [[TMP7:%.*]] = mul i64 [[TMP6]], 8
; PRED-NEXT:    [[TMP8:%.*]] = call i64 @llvm.vscale.i64()
; PRED-NEXT:    [[TMP9:%.*]] = mul i64 [[TMP8]], 8
; PRED-NEXT:    [[TMP10:%.*]] = sub i64 [[TMP0]], [[TMP9]]
; PRED-NEXT:    [[TMP11:%.*]] = icmp ugt i64 [[TMP0]], [[TMP9]]
; PRED-NEXT:    [[TMP12:%.*]] = select i1 [[TMP11]], i64 [[TMP10]], i64 0
; PRED-NEXT:    [[ACTIVE_LANE_MASK_ENTRY:%.*]] = call <vscale x 8 x i1> @llvm.get.active.lane.mask.nxv8i1.i64(i64 0, i64 [[TMP0]])
; PRED-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 8 x i16> poison, i16 [[X]], i64 0
; PRED-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 8 x i16> [[BROADCAST_SPLATINSERT]], <vscale x 8 x i16> poison, <vscale x 8 x i32> zeroinitializer
; PRED-NEXT:    br label [[VECTOR_BODY:%.*]]
; PRED:       vector.body:
; PRED-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; PRED-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = phi <vscale x 8 x i1> [ [[ACTIVE_LANE_MASK_ENTRY]], [[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT:%.*]], [[VECTOR_BODY]] ]
; PRED-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 8 x i16> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP19:%.*]], [[VECTOR_BODY]] ]
; PRED-NEXT:    [[TMP13:%.*]] = add i64 [[INDEX]], 0
; PRED-NEXT:    [[TMP14:%.*]] = getelementptr i16, ptr [[SRC]], i64 [[TMP13]]
; PRED-NEXT:    [[TMP15:%.*]] = getelementptr i16, ptr [[TMP14]], i32 0
; PRED-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 8 x i16> @llvm.masked.load.nxv8i16.p0(ptr [[TMP15]], i32 2, <vscale x 8 x i1> [[ACTIVE_LANE_MASK]], <vscale x 8 x i16> poison)
; PRED-NEXT:    [[TMP16:%.*]] = select <vscale x 8 x i1> [[ACTIVE_LANE_MASK]], <vscale x 8 x i16> [[BROADCAST_SPLAT]], <vscale x 8 x i16> shufflevector (<vscale x 8 x i16> insertelement (<vscale x 8 x i16> poison, i16 1, i64 0), <vscale x 8 x i16> poison, <vscale x 8 x i32> zeroinitializer)
; PRED-NEXT:    [[TMP17:%.*]] = udiv <vscale x 8 x i16> [[WIDE_MASKED_LOAD]], [[TMP16]]
; PRED-NEXT:    [[TMP18:%.*]] = or <vscale x 8 x i16> [[TMP17]], [[VEC_PHI]]
; PRED-NEXT:    [[TMP19]] = select <vscale x 8 x i1> [[ACTIVE_LANE_MASK]], <vscale x 8 x i16> [[TMP18]], <vscale x 8 x i16> [[VEC_PHI]]
; PRED-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP7]]
; PRED-NEXT:    [[ACTIVE_LANE_MASK_NEXT]] = call <vscale x 8 x i1> @llvm.get.active.lane.mask.nxv8i1.i64(i64 [[INDEX]], i64 [[TMP12]])
; PRED-NEXT:    [[TMP20:%.*]] = xor <vscale x 8 x i1> [[ACTIVE_LANE_MASK_NEXT]], shufflevector (<vscale x 8 x i1> insertelement (<vscale x 8 x i1> poison, i1 true, i64 0), <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer)
; PRED-NEXT:    [[TMP21:%.*]] = extractelement <vscale x 8 x i1> [[TMP20]], i32 0
; PRED-NEXT:    br i1 [[TMP21]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
; PRED:       middle.block:
; PRED-NEXT:    [[TMP22:%.*]] = call i16 @llvm.vector.reduce.or.nxv8i16(<vscale x 8 x i16> [[TMP19]])
; PRED-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
; PRED:       scalar.ph:
; PRED-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
; PRED-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i16 [ 0, [[ENTRY]] ], [ [[TMP22]], [[MIDDLE_BLOCK]] ]
; PRED-NEXT:    br label [[LOOP:%.*]]
; PRED:       loop:
; PRED-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP]] ]
; PRED-NEXT:    [[RED:%.*]] = phi i16 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[RED_NEXT:%.*]], [[LOOP]] ]
; PRED-NEXT:    [[GEP:%.*]] = getelementptr i16, ptr [[SRC]], i64 [[IV]]
; PRED-NEXT:    [[L:%.*]] = load i16, ptr [[GEP]], align 2
; PRED-NEXT:    [[DIV:%.*]] = udiv i16 [[L]], [[X]]
; PRED-NEXT:    [[RED_NEXT]] = or i16 [[DIV]], [[RED]]
; PRED-NEXT:    [[IV_NEXT]] = add i64 [[IV]], 1
; PRED-NEXT:    [[EC:%.*]] = icmp eq i64 [[IV]], [[N]]
; PRED-NEXT:    br i1 [[EC]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP5:![0-9]+]]
; PRED:       exit:
; PRED-NEXT:    [[RED_NEXT_LCSSA:%.*]] = phi i16 [ [[RED_NEXT]], [[LOOP]] ], [ [[TMP22]], [[MIDDLE_BLOCK]] ]
; PRED-NEXT:    ret i16 [[RED_NEXT_LCSSA]]
;
entry:
  br label %loop

loop:
  %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]
  %red = phi i16 [ 0, %entry ], [ %red.next, %loop ]
  %gep = getelementptr i16, ptr %src, i64 %iv
  %l = load i16, ptr %gep, align 2
  %div = udiv i16 %l, %x
  %red.next = or i16 %div, %red
  %iv.next = add i64 %iv, 1
  %ec = icmp eq i64 %iv, %N
  br i1 %ec, label %exit, label %loop

exit:
  ret i16 %red.next
}

attributes #0 = { "target-features"="+sve" }
;.
; DEFAULT: [[LOOP0]] = distinct !{[[LOOP0]], [[META1:![0-9]+]], [[META2:![0-9]+]]}
; DEFAULT: [[META1]] = !{!"llvm.loop.isvectorized", i32 1}
; DEFAULT: [[META2]] = !{!"llvm.loop.unroll.runtime.disable"}
; DEFAULT: [[LOOP3]] = distinct !{[[LOOP3]], [[META2]], [[META1]]}
; DEFAULT: [[LOOP4]] = distinct !{[[LOOP4]], [[META1]], [[META2]]}
; DEFAULT: [[LOOP5]] = distinct !{[[LOOP5]], [[META2]], [[META1]]}
;.
; PRED: [[LOOP0]] = distinct !{[[LOOP0]], [[META1:![0-9]+]], [[META2:![0-9]+]]}
; PRED: [[META1]] = !{!"llvm.loop.isvectorized", i32 1}
; PRED: [[META2]] = !{!"llvm.loop.unroll.runtime.disable"}
; PRED: [[LOOP3]] = distinct !{[[LOOP3]], [[META2]], [[META1]]}
; PRED: [[LOOP4]] = distinct !{[[LOOP4]], [[META1]], [[META2]]}
; PRED: [[LOOP5]] = distinct !{[[LOOP5]], [[META2]], [[META1]]}
;.