llvm/test/Transforms/LoopVectorize/RISCV/interleaved-masked-access.ll

; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --check-globals none --filter-out-after "^scalar.ph:" --version 6
; RUN: opt -mtriple=riscv64-none-linux-gnu -S -passes=loop-vectorize -mattr=+v -tail-folding-policy=dont-fold-tail %s 2>&1 | FileCheck %s -check-prefix=SCALAR_EPILOGUE
; RUN: opt -mtriple=riscv64-none-linux-gnu -S -passes=loop-vectorize -mattr=+v -tail-folding-policy=prefer-fold-tail -force-tail-folding-style=data %s 2>&1 | FileCheck %s -check-prefix=PREDICATED_DATA
; RUN: opt -mtriple=riscv64-none-linux-gnu -S -passes=loop-vectorize -mattr=+v -tail-folding-policy=prefer-fold-tail %s 2>&1 | FileCheck %s -check-prefix=PREDICATED_DATA-WITH-EVL

target datalayout = "e-m:e-i64:64-i128:128-n32:64-S128"

define void @masked_strided_factor2(ptr noalias nocapture readonly %p, ptr noalias nocapture %q, i8 zeroext %guard) vscale_range(2, 1024) {
; SCALAR_EPILOGUE-LABEL: define void @masked_strided_factor2(
; SCALAR_EPILOGUE-SAME: ptr noalias readonly captures(none) [[P:%.*]], ptr noalias captures(none) [[Q:%.*]], i8 zeroext [[GUARD:%.*]]) #[[ATTR0:[0-9]+]] {
; SCALAR_EPILOGUE-NEXT:  [[ENTRY:.*:]]
; SCALAR_EPILOGUE-NEXT:    [[CONV:%.*]] = zext i8 [[GUARD]] to i32
; SCALAR_EPILOGUE-NEXT:    [[TMP0:%.*]] = call i32 @llvm.vscale.i32()
; SCALAR_EPILOGUE-NEXT:    [[TMP1:%.*]] = shl nuw i32 [[TMP0]], 4
; SCALAR_EPILOGUE-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 1024, [[TMP1]]
; SCALAR_EPILOGUE-NEXT:    br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
; SCALAR_EPILOGUE:       [[VECTOR_PH]]:
; SCALAR_EPILOGUE-NEXT:    [[N_MOD_VF:%.*]] = urem i32 1024, [[TMP1]]
; SCALAR_EPILOGUE-NEXT:    [[N_VEC:%.*]] = sub i32 1024, [[N_MOD_VF]]
; SCALAR_EPILOGUE-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 16 x i32> poison, i32 [[CONV]], i64 0
; SCALAR_EPILOGUE-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 16 x i32> [[BROADCAST_SPLATINSERT]], <vscale x 16 x i32> poison, <vscale x 16 x i32> zeroinitializer
; SCALAR_EPILOGUE-NEXT:    [[TMP4:%.*]] = call <vscale x 16 x i32> @llvm.stepvector.nxv16i32()
; SCALAR_EPILOGUE-NEXT:    [[BROADCAST_SPLATINSERT1:%.*]] = insertelement <vscale x 16 x i32> poison, i32 [[TMP1]], i64 0
; SCALAR_EPILOGUE-NEXT:    [[BROADCAST_SPLAT2:%.*]] = shufflevector <vscale x 16 x i32> [[BROADCAST_SPLATINSERT1]], <vscale x 16 x i32> poison, <vscale x 16 x i32> zeroinitializer
; SCALAR_EPILOGUE-NEXT:    br label %[[VECTOR_BODY:.*]]
; SCALAR_EPILOGUE:       [[VECTOR_BODY]]:
; SCALAR_EPILOGUE-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
; SCALAR_EPILOGUE-NEXT:    [[VEC_IND:%.*]] = phi <vscale x 16 x i32> [ [[TMP4]], %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], %[[VECTOR_BODY]] ]
; SCALAR_EPILOGUE-NEXT:    [[TMP5:%.*]] = icmp ugt <vscale x 16 x i32> [[VEC_IND]], [[BROADCAST_SPLAT]]
; SCALAR_EPILOGUE-NEXT:    [[TMP6:%.*]] = shl i32 [[INDEX]], 1
; SCALAR_EPILOGUE-NEXT:    [[TMP7:%.*]] = getelementptr i8, ptr [[P]], i32 [[TMP6]]
; SCALAR_EPILOGUE-NEXT:    [[INTERLEAVED_MASK:%.*]] = call <vscale x 32 x i1> @llvm.vector.interleave2.nxv32i1(<vscale x 16 x i1> [[TMP5]], <vscale x 16 x i1> [[TMP5]])
; SCALAR_EPILOGUE-NEXT:    [[WIDE_MASKED_VEC:%.*]] = call <vscale x 32 x i8> @llvm.masked.load.nxv32i8.p0(ptr align 1 [[TMP7]], <vscale x 32 x i1> [[INTERLEAVED_MASK]], <vscale x 32 x i8> poison)
; SCALAR_EPILOGUE-NEXT:    [[STRIDED_VEC:%.*]] = call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.vector.deinterleave2.nxv32i8(<vscale x 32 x i8> [[WIDE_MASKED_VEC]])
; SCALAR_EPILOGUE-NEXT:    [[TMP8:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[STRIDED_VEC]], 0
; SCALAR_EPILOGUE-NEXT:    [[TMP9:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[STRIDED_VEC]], 1
; SCALAR_EPILOGUE-NEXT:    [[TMP10:%.*]] = icmp slt <vscale x 16 x i8> [[TMP8]], [[TMP9]]
; SCALAR_EPILOGUE-NEXT:    [[TMP11:%.*]] = select <vscale x 16 x i1> [[TMP10]], <vscale x 16 x i8> [[TMP9]], <vscale x 16 x i8> [[TMP8]]
; SCALAR_EPILOGUE-NEXT:    [[TMP12:%.*]] = getelementptr i8, ptr [[Q]], i32 [[TMP6]]
; SCALAR_EPILOGUE-NEXT:    [[TMP13:%.*]] = sub <vscale x 16 x i8> zeroinitializer, [[TMP11]]
; SCALAR_EPILOGUE-NEXT:    [[INTERLEAVED_VEC:%.*]] = call <vscale x 32 x i8> @llvm.vector.interleave2.nxv32i8(<vscale x 16 x i8> [[TMP11]], <vscale x 16 x i8> [[TMP13]])
; SCALAR_EPILOGUE-NEXT:    [[INTERLEAVED_MASK3:%.*]] = call <vscale x 32 x i1> @llvm.vector.interleave2.nxv32i1(<vscale x 16 x i1> [[TMP5]], <vscale x 16 x i1> [[TMP5]])
; SCALAR_EPILOGUE-NEXT:    call void @llvm.masked.store.nxv32i8.p0(<vscale x 32 x i8> [[INTERLEAVED_VEC]], ptr align 1 [[TMP12]], <vscale x 32 x i1> [[INTERLEAVED_MASK3]])
; SCALAR_EPILOGUE-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], [[TMP1]]
; SCALAR_EPILOGUE-NEXT:    [[VEC_IND_NEXT]] = add nuw nsw <vscale x 16 x i32> [[VEC_IND]], [[BROADCAST_SPLAT2]]
; SCALAR_EPILOGUE-NEXT:    [[TMP14:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
; SCALAR_EPILOGUE-NEXT:    br i1 [[TMP14]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
; SCALAR_EPILOGUE:       [[MIDDLE_BLOCK]]:
; SCALAR_EPILOGUE-NEXT:    [[CMP_N:%.*]] = icmp eq i32 1024, [[N_VEC]]
; SCALAR_EPILOGUE-NEXT:    br i1 [[CMP_N]], [[FOR_END:label %.*]], label %[[SCALAR_PH]]
; SCALAR_EPILOGUE:       [[SCALAR_PH]]:
;
; PREDICATED_DATA-LABEL: define void @masked_strided_factor2(
; PREDICATED_DATA-SAME: ptr noalias readonly captures(none) [[P:%.*]], ptr noalias captures(none) [[Q:%.*]], i8 zeroext [[GUARD:%.*]]) #[[ATTR0:[0-9]+]] {
; PREDICATED_DATA-NEXT:  [[ENTRY:.*:]]
; PREDICATED_DATA-NEXT:    [[CONV:%.*]] = zext i8 [[GUARD]] to i32
; PREDICATED_DATA-NEXT:    br label %[[VECTOR_PH:.*]]
; PREDICATED_DATA:       [[VECTOR_PH]]:
; PREDICATED_DATA-NEXT:    [[TMP0:%.*]] = call i32 @llvm.vscale.i32()
; PREDICATED_DATA-NEXT:    [[TMP1:%.*]] = shl nuw i32 [[TMP0]], 4
; PREDICATED_DATA-NEXT:    [[TMP2:%.*]] = sub i32 [[TMP1]], 1
; PREDICATED_DATA-NEXT:    [[N_RND_UP:%.*]] = add i32 1024, [[TMP2]]
; PREDICATED_DATA-NEXT:    [[N_MOD_VF:%.*]] = urem i32 [[N_RND_UP]], [[TMP1]]
; PREDICATED_DATA-NEXT:    [[N_VEC:%.*]] = sub i32 [[N_RND_UP]], [[N_MOD_VF]]
; PREDICATED_DATA-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 16 x i32> poison, i32 [[CONV]], i64 0
; PREDICATED_DATA-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 16 x i32> [[BROADCAST_SPLATINSERT]], <vscale x 16 x i32> poison, <vscale x 16 x i32> zeroinitializer
; PREDICATED_DATA-NEXT:    [[TMP3:%.*]] = call <vscale x 16 x i32> @llvm.stepvector.nxv16i32()
; PREDICATED_DATA-NEXT:    [[BROADCAST_SPLATINSERT1:%.*]] = insertelement <vscale x 16 x i32> poison, i32 [[TMP1]], i64 0
; PREDICATED_DATA-NEXT:    [[BROADCAST_SPLAT2:%.*]] = shufflevector <vscale x 16 x i32> [[BROADCAST_SPLATINSERT1]], <vscale x 16 x i32> poison, <vscale x 16 x i32> zeroinitializer
; PREDICATED_DATA-NEXT:    br label %[[VECTOR_BODY:.*]]
; PREDICATED_DATA:       [[VECTOR_BODY]]:
; PREDICATED_DATA-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
; PREDICATED_DATA-NEXT:    [[VEC_IND:%.*]] = phi <vscale x 16 x i32> [ [[TMP3]], %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], %[[VECTOR_BODY]] ]
; PREDICATED_DATA-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = call <vscale x 16 x i1> @llvm.get.active.lane.mask.nxv16i1.i32(i32 [[INDEX]], i32 1024)
; PREDICATED_DATA-NEXT:    [[TMP4:%.*]] = icmp ugt <vscale x 16 x i32> [[VEC_IND]], [[BROADCAST_SPLAT]]
; PREDICATED_DATA-NEXT:    [[TMP5:%.*]] = select <vscale x 16 x i1> [[ACTIVE_LANE_MASK]], <vscale x 16 x i1> [[TMP4]], <vscale x 16 x i1> zeroinitializer
; PREDICATED_DATA-NEXT:    [[TMP6:%.*]] = shl i32 [[INDEX]], 1
; PREDICATED_DATA-NEXT:    [[TMP7:%.*]] = getelementptr i8, ptr [[P]], i32 [[TMP6]]
; PREDICATED_DATA-NEXT:    [[INTERLEAVED_MASK:%.*]] = call <vscale x 32 x i1> @llvm.vector.interleave2.nxv32i1(<vscale x 16 x i1> [[TMP5]], <vscale x 16 x i1> [[TMP5]])
; PREDICATED_DATA-NEXT:    [[WIDE_MASKED_VEC:%.*]] = call <vscale x 32 x i8> @llvm.masked.load.nxv32i8.p0(ptr align 1 [[TMP7]], <vscale x 32 x i1> [[INTERLEAVED_MASK]], <vscale x 32 x i8> poison)
; PREDICATED_DATA-NEXT:    [[STRIDED_VEC:%.*]] = call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.vector.deinterleave2.nxv32i8(<vscale x 32 x i8> [[WIDE_MASKED_VEC]])
; PREDICATED_DATA-NEXT:    [[TMP8:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[STRIDED_VEC]], 0
; PREDICATED_DATA-NEXT:    [[TMP9:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[STRIDED_VEC]], 1
; PREDICATED_DATA-NEXT:    [[TMP10:%.*]] = icmp slt <vscale x 16 x i8> [[TMP8]], [[TMP9]]
; PREDICATED_DATA-NEXT:    [[TMP11:%.*]] = select <vscale x 16 x i1> [[TMP10]], <vscale x 16 x i8> [[TMP9]], <vscale x 16 x i8> [[TMP8]]
; PREDICATED_DATA-NEXT:    [[TMP12:%.*]] = getelementptr i8, ptr [[Q]], i32 [[TMP6]]
; PREDICATED_DATA-NEXT:    [[TMP13:%.*]] = sub <vscale x 16 x i8> zeroinitializer, [[TMP11]]
; PREDICATED_DATA-NEXT:    [[INTERLEAVED_VEC:%.*]] = call <vscale x 32 x i8> @llvm.vector.interleave2.nxv32i8(<vscale x 16 x i8> [[TMP11]], <vscale x 16 x i8> [[TMP13]])
; PREDICATED_DATA-NEXT:    [[INTERLEAVED_MASK3:%.*]] = call <vscale x 32 x i1> @llvm.vector.interleave2.nxv32i1(<vscale x 16 x i1> [[TMP5]], <vscale x 16 x i1> [[TMP5]])
; PREDICATED_DATA-NEXT:    call void @llvm.masked.store.nxv32i8.p0(<vscale x 32 x i8> [[INTERLEAVED_VEC]], ptr align 1 [[TMP12]], <vscale x 32 x i1> [[INTERLEAVED_MASK3]])
; PREDICATED_DATA-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], [[TMP1]]
; PREDICATED_DATA-NEXT:    [[VEC_IND_NEXT]] = add nuw nsw <vscale x 16 x i32> [[VEC_IND]], [[BROADCAST_SPLAT2]]
; PREDICATED_DATA-NEXT:    [[TMP14:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
; PREDICATED_DATA-NEXT:    br i1 [[TMP14]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
; PREDICATED_DATA:       [[MIDDLE_BLOCK]]:
; PREDICATED_DATA-NEXT:    br label %[[FOR_END:.*]]
; PREDICATED_DATA:       [[FOR_END]]:
; PREDICATED_DATA-NEXT:    ret void
;
; PREDICATED_DATA-WITH-EVL-LABEL: define void @masked_strided_factor2(
; PREDICATED_DATA-WITH-EVL-SAME: ptr noalias readonly captures(none) [[P:%.*]], ptr noalias captures(none) [[Q:%.*]], i8 zeroext [[GUARD:%.*]]) #[[ATTR0:[0-9]+]] {
; PREDICATED_DATA-WITH-EVL-NEXT:  [[ENTRY:.*:]]
; PREDICATED_DATA-WITH-EVL-NEXT:    [[CONV:%.*]] = zext i8 [[GUARD]] to i32
; PREDICATED_DATA-WITH-EVL-NEXT:    br label %[[VECTOR_PH:.*]]
; PREDICATED_DATA-WITH-EVL:       [[VECTOR_PH]]:
; PREDICATED_DATA-WITH-EVL-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 16 x i32> poison, i32 [[CONV]], i64 0
; PREDICATED_DATA-WITH-EVL-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 16 x i32> [[BROADCAST_SPLATINSERT]], <vscale x 16 x i32> poison, <vscale x 16 x i32> zeroinitializer
; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP0:%.*]] = call <vscale x 16 x i32> @llvm.stepvector.nxv16i32()
; PREDICATED_DATA-WITH-EVL-NEXT:    br label %[[VECTOR_BODY:.*]]
; PREDICATED_DATA-WITH-EVL:       [[VECTOR_BODY]]:
; PREDICATED_DATA-WITH-EVL-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, %[[VECTOR_PH]] ], [ [[CURRENT_ITERATION_NEXT:%.*]], %[[VECTOR_BODY]] ]
; PREDICATED_DATA-WITH-EVL-NEXT:    [[VEC_IND:%.*]] = phi <vscale x 16 x i32> [ [[TMP0]], %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], %[[VECTOR_BODY]] ]
; PREDICATED_DATA-WITH-EVL-NEXT:    [[AVL:%.*]] = phi i32 [ 1024, %[[VECTOR_PH]] ], [ [[AVL_NEXT:%.*]], %[[VECTOR_BODY]] ]
; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP1:%.*]] = call i32 @llvm.experimental.get.vector.length.i32(i32 [[AVL]], i32 16, i1 true)
; PREDICATED_DATA-WITH-EVL-NEXT:    [[BROADCAST_SPLATINSERT1:%.*]] = insertelement <vscale x 16 x i32> poison, i32 [[TMP1]], i64 0
; PREDICATED_DATA-WITH-EVL-NEXT:    [[BROADCAST_SPLAT2:%.*]] = shufflevector <vscale x 16 x i32> [[BROADCAST_SPLATINSERT1]], <vscale x 16 x i32> poison, <vscale x 16 x i32> zeroinitializer
; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP2:%.*]] = icmp ugt <vscale x 16 x i32> [[VEC_IND]], [[BROADCAST_SPLAT]]
; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP3:%.*]] = shl i32 [[INDEX]], 1
; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP4:%.*]] = getelementptr i8, ptr [[P]], i32 [[TMP3]]
; PREDICATED_DATA-WITH-EVL-NEXT:    [[INTERLEAVE_EVL:%.*]] = mul nuw nsw i32 [[TMP1]], 2
; PREDICATED_DATA-WITH-EVL-NEXT:    [[INTERLEAVED_MASK:%.*]] = call <vscale x 32 x i1> @llvm.vector.interleave2.nxv32i1(<vscale x 16 x i1> [[TMP2]], <vscale x 16 x i1> [[TMP2]])
; PREDICATED_DATA-WITH-EVL-NEXT:    [[WIDE_VP_LOAD:%.*]] = call <vscale x 32 x i8> @llvm.vp.load.nxv32i8.p0(ptr align 1 [[TMP4]], <vscale x 32 x i1> [[INTERLEAVED_MASK]], i32 [[INTERLEAVE_EVL]])
; PREDICATED_DATA-WITH-EVL-NEXT:    [[STRIDED_VEC:%.*]] = call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.vector.deinterleave2.nxv32i8(<vscale x 32 x i8> [[WIDE_VP_LOAD]])
; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP5:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[STRIDED_VEC]], 0
; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[STRIDED_VEC]], 1
; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP7:%.*]] = icmp slt <vscale x 16 x i8> [[TMP5]], [[TMP6]]
; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP8:%.*]] = select <vscale x 16 x i1> [[TMP7]], <vscale x 16 x i8> [[TMP6]], <vscale x 16 x i8> [[TMP5]]
; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP9:%.*]] = getelementptr i8, ptr [[Q]], i32 [[TMP3]]
; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP10:%.*]] = sub <vscale x 16 x i8> zeroinitializer, [[TMP8]]
; PREDICATED_DATA-WITH-EVL-NEXT:    [[INTERLEAVE_EVL3:%.*]] = mul nuw nsw i32 [[TMP1]], 2
; PREDICATED_DATA-WITH-EVL-NEXT:    [[INTERLEAVED_MASK4:%.*]] = call <vscale x 32 x i1> @llvm.vector.interleave2.nxv32i1(<vscale x 16 x i1> [[TMP2]], <vscale x 16 x i1> [[TMP2]])
; PREDICATED_DATA-WITH-EVL-NEXT:    [[INTERLEAVED_VEC:%.*]] = call <vscale x 32 x i8> @llvm.vector.interleave2.nxv32i8(<vscale x 16 x i8> [[TMP8]], <vscale x 16 x i8> [[TMP10]])
; PREDICATED_DATA-WITH-EVL-NEXT:    call void @llvm.vp.store.nxv32i8.p0(<vscale x 32 x i8> [[INTERLEAVED_VEC]], ptr align 1 [[TMP9]], <vscale x 32 x i1> [[INTERLEAVED_MASK4]], i32 [[INTERLEAVE_EVL3]])
; PREDICATED_DATA-WITH-EVL-NEXT:    [[CURRENT_ITERATION_NEXT]] = add nuw i32 [[TMP1]], [[INDEX]]
; PREDICATED_DATA-WITH-EVL-NEXT:    [[AVL_NEXT]] = sub nuw i32 [[AVL]], [[TMP1]]
; PREDICATED_DATA-WITH-EVL-NEXT:    [[VEC_IND_NEXT]] = add nuw nsw <vscale x 16 x i32> [[VEC_IND]], [[BROADCAST_SPLAT2]]
; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP11:%.*]] = icmp eq i32 [[AVL_NEXT]], 0
; PREDICATED_DATA-WITH-EVL-NEXT:    br i1 [[TMP11]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
; PREDICATED_DATA-WITH-EVL:       [[MIDDLE_BLOCK]]:
; PREDICATED_DATA-WITH-EVL-NEXT:    br label %[[FOR_END:.*]]
; PREDICATED_DATA-WITH-EVL:       [[FOR_END]]:
; PREDICATED_DATA-WITH-EVL-NEXT:    ret void
;
entry:
  %conv = zext i8 %guard to i32
  br label %for.body

for.body:
  %ix.024 = phi i32 [ 0, %entry ], [ %inc, %for.inc ]
  %cmp1 = icmp ugt i32 %ix.024, %conv
  br i1 %cmp1, label %if.then, label %for.inc

if.then:
  %mul = shl nuw nsw i32 %ix.024, 1
  %arrayidx = getelementptr inbounds i8, ptr %p, i32 %mul
  %0 = load i8, ptr %arrayidx, align 1
  %add = or disjoint i32 %mul, 1
  %arrayidx4 = getelementptr inbounds i8, ptr %p, i32 %add
  %1 = load i8, ptr %arrayidx4, align 1
  %cmp.i = icmp slt i8 %0, %1
  %spec.select.i = select i1 %cmp.i, i8 %1, i8 %0
  %arrayidx6 = getelementptr inbounds i8, ptr %q, i32 %mul
  store i8 %spec.select.i, ptr %arrayidx6, align 1
  %sub = sub i8 0, %spec.select.i
  %arrayidx11 = getelementptr inbounds i8, ptr %q, i32 %add
  store i8 %sub, ptr %arrayidx11, align 1
  br label %for.inc

for.inc:
  %inc = add nuw nsw i32 %ix.024, 1
  %exitcond = icmp eq i32 %inc, 1024
  br i1 %exitcond, label %for.end, label %for.body

for.end:
  ret void
}


define void @masked_strided_factor4(ptr noalias nocapture readonly %p, ptr noalias nocapture %q, i8 zeroext %guard) vscale_range(2, 1024) {
; SCALAR_EPILOGUE-LABEL: define void @masked_strided_factor4(
; SCALAR_EPILOGUE-SAME: ptr noalias readonly captures(none) [[P:%.*]], ptr noalias captures(none) [[Q:%.*]], i8 zeroext [[GUARD:%.*]]) #[[ATTR0]] {
; SCALAR_EPILOGUE-NEXT:  [[ENTRY:.*:]]
; SCALAR_EPILOGUE-NEXT:    [[CONV:%.*]] = zext i8 [[GUARD]] to i32
; SCALAR_EPILOGUE-NEXT:    [[TMP0:%.*]] = call i32 @llvm.vscale.i32()
; SCALAR_EPILOGUE-NEXT:    [[TMP1:%.*]] = shl nuw i32 [[TMP0]], 4
; SCALAR_EPILOGUE-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 1024, [[TMP1]]
; SCALAR_EPILOGUE-NEXT:    br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
; SCALAR_EPILOGUE:       [[VECTOR_PH]]:
; SCALAR_EPILOGUE-NEXT:    [[N_MOD_VF:%.*]] = urem i32 1024, [[TMP1]]
; SCALAR_EPILOGUE-NEXT:    [[N_VEC:%.*]] = sub i32 1024, [[N_MOD_VF]]
; SCALAR_EPILOGUE-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 16 x i32> poison, i32 [[CONV]], i64 0
; SCALAR_EPILOGUE-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 16 x i32> [[BROADCAST_SPLATINSERT]], <vscale x 16 x i32> poison, <vscale x 16 x i32> zeroinitializer
; SCALAR_EPILOGUE-NEXT:    [[TMP4:%.*]] = call <vscale x 16 x i32> @llvm.stepvector.nxv16i32()
; SCALAR_EPILOGUE-NEXT:    [[BROADCAST_SPLATINSERT1:%.*]] = insertelement <vscale x 16 x i32> poison, i32 [[TMP1]], i64 0
; SCALAR_EPILOGUE-NEXT:    [[BROADCAST_SPLAT2:%.*]] = shufflevector <vscale x 16 x i32> [[BROADCAST_SPLATINSERT1]], <vscale x 16 x i32> poison, <vscale x 16 x i32> zeroinitializer
; SCALAR_EPILOGUE-NEXT:    br label %[[VECTOR_BODY:.*]]
; SCALAR_EPILOGUE:       [[VECTOR_BODY]]:
; SCALAR_EPILOGUE-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
; SCALAR_EPILOGUE-NEXT:    [[VEC_IND:%.*]] = phi <vscale x 16 x i32> [ [[TMP4]], %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], %[[VECTOR_BODY]] ]
; SCALAR_EPILOGUE-NEXT:    [[TMP5:%.*]] = icmp ugt <vscale x 16 x i32> [[VEC_IND]], [[BROADCAST_SPLAT]]
; SCALAR_EPILOGUE-NEXT:    [[TMP6:%.*]] = shl i32 [[INDEX]], 2
; SCALAR_EPILOGUE-NEXT:    [[TMP7:%.*]] = getelementptr i8, ptr [[P]], i32 [[TMP6]]
; SCALAR_EPILOGUE-NEXT:    [[INTERLEAVED_MASK:%.*]] = call <vscale x 64 x i1> @llvm.vector.interleave4.nxv64i1(<vscale x 16 x i1> [[TMP5]], <vscale x 16 x i1> [[TMP5]], <vscale x 16 x i1> [[TMP5]], <vscale x 16 x i1> [[TMP5]])
; SCALAR_EPILOGUE-NEXT:    [[WIDE_MASKED_VEC:%.*]] = call <vscale x 64 x i8> @llvm.masked.load.nxv64i8.p0(ptr align 1 [[TMP7]], <vscale x 64 x i1> [[INTERLEAVED_MASK]], <vscale x 64 x i8> poison)
; SCALAR_EPILOGUE-NEXT:    [[STRIDED_VEC:%.*]] = call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.vector.deinterleave4.nxv64i8(<vscale x 64 x i8> [[WIDE_MASKED_VEC]])
; SCALAR_EPILOGUE-NEXT:    [[TMP8:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[STRIDED_VEC]], 0
; SCALAR_EPILOGUE-NEXT:    [[TMP9:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[STRIDED_VEC]], 1
; SCALAR_EPILOGUE-NEXT:    [[TMP10:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[STRIDED_VEC]], 2
; SCALAR_EPILOGUE-NEXT:    [[TMP11:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[STRIDED_VEC]], 3
; SCALAR_EPILOGUE-NEXT:    [[TMP12:%.*]] = icmp slt <vscale x 16 x i8> [[TMP8]], [[TMP9]]
; SCALAR_EPILOGUE-NEXT:    [[TMP13:%.*]] = select <vscale x 16 x i1> [[TMP12]], <vscale x 16 x i8> [[TMP9]], <vscale x 16 x i8> [[TMP8]]
; SCALAR_EPILOGUE-NEXT:    [[TMP14:%.*]] = sub <vscale x 16 x i8> zeroinitializer, [[TMP13]]
; SCALAR_EPILOGUE-NEXT:    [[TMP15:%.*]] = icmp slt <vscale x 16 x i8> [[TMP10]], [[TMP11]]
; SCALAR_EPILOGUE-NEXT:    [[TMP16:%.*]] = select <vscale x 16 x i1> [[TMP15]], <vscale x 16 x i8> [[TMP11]], <vscale x 16 x i8> [[TMP10]]
; SCALAR_EPILOGUE-NEXT:    [[TMP17:%.*]] = sub <vscale x 16 x i8> zeroinitializer, [[TMP16]]
; SCALAR_EPILOGUE-NEXT:    [[TMP18:%.*]] = getelementptr i8, ptr [[Q]], i32 [[TMP6]]
; SCALAR_EPILOGUE-NEXT:    [[INTERLEAVED_VEC:%.*]] = call <vscale x 64 x i8> @llvm.vector.interleave4.nxv64i8(<vscale x 16 x i8> [[TMP13]], <vscale x 16 x i8> [[TMP14]], <vscale x 16 x i8> [[TMP16]], <vscale x 16 x i8> [[TMP17]])
; SCALAR_EPILOGUE-NEXT:    [[INTERLEAVED_MASK3:%.*]] = call <vscale x 64 x i1> @llvm.vector.interleave4.nxv64i1(<vscale x 16 x i1> [[TMP5]], <vscale x 16 x i1> [[TMP5]], <vscale x 16 x i1> [[TMP5]], <vscale x 16 x i1> [[TMP5]])
; SCALAR_EPILOGUE-NEXT:    call void @llvm.masked.store.nxv64i8.p0(<vscale x 64 x i8> [[INTERLEAVED_VEC]], ptr align 1 [[TMP18]], <vscale x 64 x i1> [[INTERLEAVED_MASK3]])
; SCALAR_EPILOGUE-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], [[TMP1]]
; SCALAR_EPILOGUE-NEXT:    [[VEC_IND_NEXT]] = add nuw nsw <vscale x 16 x i32> [[VEC_IND]], [[BROADCAST_SPLAT2]]
; SCALAR_EPILOGUE-NEXT:    [[TMP19:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
; SCALAR_EPILOGUE-NEXT:    br i1 [[TMP19]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
; SCALAR_EPILOGUE:       [[MIDDLE_BLOCK]]:
; SCALAR_EPILOGUE-NEXT:    [[CMP_N:%.*]] = icmp eq i32 1024, [[N_VEC]]
; SCALAR_EPILOGUE-NEXT:    br i1 [[CMP_N]], [[FOR_END:label %.*]], label %[[SCALAR_PH]]
; SCALAR_EPILOGUE:       [[SCALAR_PH]]:
;
; PREDICATED_DATA-LABEL: define void @masked_strided_factor4(
; PREDICATED_DATA-SAME: ptr noalias readonly captures(none) [[P:%.*]], ptr noalias captures(none) [[Q:%.*]], i8 zeroext [[GUARD:%.*]]) #[[ATTR0]] {
; PREDICATED_DATA-NEXT:  [[ENTRY:.*:]]
; PREDICATED_DATA-NEXT:    [[CONV:%.*]] = zext i8 [[GUARD]] to i32
; PREDICATED_DATA-NEXT:    br label %[[VECTOR_PH:.*]]
; PREDICATED_DATA:       [[VECTOR_PH]]:
; PREDICATED_DATA-NEXT:    [[TMP0:%.*]] = call i32 @llvm.vscale.i32()
; PREDICATED_DATA-NEXT:    [[TMP1:%.*]] = shl nuw i32 [[TMP0]], 4
; PREDICATED_DATA-NEXT:    [[TMP2:%.*]] = sub i32 [[TMP1]], 1
; PREDICATED_DATA-NEXT:    [[N_RND_UP:%.*]] = add i32 1024, [[TMP2]]
; PREDICATED_DATA-NEXT:    [[N_MOD_VF:%.*]] = urem i32 [[N_RND_UP]], [[TMP1]]
; PREDICATED_DATA-NEXT:    [[N_VEC:%.*]] = sub i32 [[N_RND_UP]], [[N_MOD_VF]]
; PREDICATED_DATA-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 16 x i32> poison, i32 [[CONV]], i64 0
; PREDICATED_DATA-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 16 x i32> [[BROADCAST_SPLATINSERT]], <vscale x 16 x i32> poison, <vscale x 16 x i32> zeroinitializer
; PREDICATED_DATA-NEXT:    [[TMP3:%.*]] = call <vscale x 16 x i32> @llvm.stepvector.nxv16i32()
; PREDICATED_DATA-NEXT:    [[BROADCAST_SPLATINSERT1:%.*]] = insertelement <vscale x 16 x i32> poison, i32 [[TMP1]], i64 0
; PREDICATED_DATA-NEXT:    [[BROADCAST_SPLAT2:%.*]] = shufflevector <vscale x 16 x i32> [[BROADCAST_SPLATINSERT1]], <vscale x 16 x i32> poison, <vscale x 16 x i32> zeroinitializer
; PREDICATED_DATA-NEXT:    br label %[[VECTOR_BODY:.*]]
; PREDICATED_DATA:       [[VECTOR_BODY]]:
; PREDICATED_DATA-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
; PREDICATED_DATA-NEXT:    [[VEC_IND:%.*]] = phi <vscale x 16 x i32> [ [[TMP3]], %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], %[[VECTOR_BODY]] ]
; PREDICATED_DATA-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = call <vscale x 16 x i1> @llvm.get.active.lane.mask.nxv16i1.i32(i32 [[INDEX]], i32 1024)
; PREDICATED_DATA-NEXT:    [[TMP4:%.*]] = icmp ugt <vscale x 16 x i32> [[VEC_IND]], [[BROADCAST_SPLAT]]
; PREDICATED_DATA-NEXT:    [[TMP5:%.*]] = select <vscale x 16 x i1> [[ACTIVE_LANE_MASK]], <vscale x 16 x i1> [[TMP4]], <vscale x 16 x i1> zeroinitializer
; PREDICATED_DATA-NEXT:    [[TMP6:%.*]] = shl i32 [[INDEX]], 2
; PREDICATED_DATA-NEXT:    [[TMP7:%.*]] = getelementptr i8, ptr [[P]], i32 [[TMP6]]
; PREDICATED_DATA-NEXT:    [[INTERLEAVED_MASK:%.*]] = call <vscale x 64 x i1> @llvm.vector.interleave4.nxv64i1(<vscale x 16 x i1> [[TMP5]], <vscale x 16 x i1> [[TMP5]], <vscale x 16 x i1> [[TMP5]], <vscale x 16 x i1> [[TMP5]])
; PREDICATED_DATA-NEXT:    [[WIDE_MASKED_VEC:%.*]] = call <vscale x 64 x i8> @llvm.masked.load.nxv64i8.p0(ptr align 1 [[TMP7]], <vscale x 64 x i1> [[INTERLEAVED_MASK]], <vscale x 64 x i8> poison)
; PREDICATED_DATA-NEXT:    [[STRIDED_VEC:%.*]] = call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.vector.deinterleave4.nxv64i8(<vscale x 64 x i8> [[WIDE_MASKED_VEC]])
; PREDICATED_DATA-NEXT:    [[TMP8:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[STRIDED_VEC]], 0
; PREDICATED_DATA-NEXT:    [[TMP9:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[STRIDED_VEC]], 1
; PREDICATED_DATA-NEXT:    [[TMP10:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[STRIDED_VEC]], 2
; PREDICATED_DATA-NEXT:    [[TMP11:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[STRIDED_VEC]], 3
; PREDICATED_DATA-NEXT:    [[TMP12:%.*]] = icmp slt <vscale x 16 x i8> [[TMP8]], [[TMP9]]
; PREDICATED_DATA-NEXT:    [[TMP13:%.*]] = select <vscale x 16 x i1> [[TMP12]], <vscale x 16 x i8> [[TMP9]], <vscale x 16 x i8> [[TMP8]]
; PREDICATED_DATA-NEXT:    [[TMP14:%.*]] = sub <vscale x 16 x i8> zeroinitializer, [[TMP13]]
; PREDICATED_DATA-NEXT:    [[TMP15:%.*]] = icmp slt <vscale x 16 x i8> [[TMP10]], [[TMP11]]
; PREDICATED_DATA-NEXT:    [[TMP16:%.*]] = select <vscale x 16 x i1> [[TMP15]], <vscale x 16 x i8> [[TMP11]], <vscale x 16 x i8> [[TMP10]]
; PREDICATED_DATA-NEXT:    [[TMP17:%.*]] = sub <vscale x 16 x i8> zeroinitializer, [[TMP16]]
; PREDICATED_DATA-NEXT:    [[TMP18:%.*]] = getelementptr i8, ptr [[Q]], i32 [[TMP6]]
; PREDICATED_DATA-NEXT:    [[INTERLEAVED_VEC:%.*]] = call <vscale x 64 x i8> @llvm.vector.interleave4.nxv64i8(<vscale x 16 x i8> [[TMP13]], <vscale x 16 x i8> [[TMP14]], <vscale x 16 x i8> [[TMP16]], <vscale x 16 x i8> [[TMP17]])
; PREDICATED_DATA-NEXT:    [[INTERLEAVED_MASK3:%.*]] = call <vscale x 64 x i1> @llvm.vector.interleave4.nxv64i1(<vscale x 16 x i1> [[TMP5]], <vscale x 16 x i1> [[TMP5]], <vscale x 16 x i1> [[TMP5]], <vscale x 16 x i1> [[TMP5]])
; PREDICATED_DATA-NEXT:    call void @llvm.masked.store.nxv64i8.p0(<vscale x 64 x i8> [[INTERLEAVED_VEC]], ptr align 1 [[TMP18]], <vscale x 64 x i1> [[INTERLEAVED_MASK3]])
; PREDICATED_DATA-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], [[TMP1]]
; PREDICATED_DATA-NEXT:    [[VEC_IND_NEXT]] = add nuw nsw <vscale x 16 x i32> [[VEC_IND]], [[BROADCAST_SPLAT2]]
; PREDICATED_DATA-NEXT:    [[TMP19:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
; PREDICATED_DATA-NEXT:    br i1 [[TMP19]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP3:![0-9]+]]
; PREDICATED_DATA:       [[MIDDLE_BLOCK]]:
; PREDICATED_DATA-NEXT:    br label %[[FOR_END:.*]]
; PREDICATED_DATA:       [[FOR_END]]:
; PREDICATED_DATA-NEXT:    ret void
;
; PREDICATED_DATA-WITH-EVL-LABEL: define void @masked_strided_factor4(
; PREDICATED_DATA-WITH-EVL-SAME: ptr noalias readonly captures(none) [[P:%.*]], ptr noalias captures(none) [[Q:%.*]], i8 zeroext [[GUARD:%.*]]) #[[ATTR0]] {
; PREDICATED_DATA-WITH-EVL-NEXT:  [[ENTRY:.*:]]
; PREDICATED_DATA-WITH-EVL-NEXT:    [[CONV:%.*]] = zext i8 [[GUARD]] to i32
; PREDICATED_DATA-WITH-EVL-NEXT:    br label %[[VECTOR_PH:.*]]
; PREDICATED_DATA-WITH-EVL:       [[VECTOR_PH]]:
; PREDICATED_DATA-WITH-EVL-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 16 x i32> poison, i32 [[CONV]], i64 0
; PREDICATED_DATA-WITH-EVL-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 16 x i32> [[BROADCAST_SPLATINSERT]], <vscale x 16 x i32> poison, <vscale x 16 x i32> zeroinitializer
; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP0:%.*]] = call <vscale x 16 x i32> @llvm.stepvector.nxv16i32()
; PREDICATED_DATA-WITH-EVL-NEXT:    br label %[[VECTOR_BODY:.*]]
; PREDICATED_DATA-WITH-EVL:       [[VECTOR_BODY]]:
; PREDICATED_DATA-WITH-EVL-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, %[[VECTOR_PH]] ], [ [[CURRENT_ITERATION_NEXT:%.*]], %[[VECTOR_BODY]] ]
; PREDICATED_DATA-WITH-EVL-NEXT:    [[VEC_IND:%.*]] = phi <vscale x 16 x i32> [ [[TMP0]], %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], %[[VECTOR_BODY]] ]
; PREDICATED_DATA-WITH-EVL-NEXT:    [[AVL:%.*]] = phi i32 [ 1024, %[[VECTOR_PH]] ], [ [[AVL_NEXT:%.*]], %[[VECTOR_BODY]] ]
; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP1:%.*]] = call i32 @llvm.experimental.get.vector.length.i32(i32 [[AVL]], i32 16, i1 true)
; PREDICATED_DATA-WITH-EVL-NEXT:    [[BROADCAST_SPLATINSERT1:%.*]] = insertelement <vscale x 16 x i32> poison, i32 [[TMP1]], i64 0
; PREDICATED_DATA-WITH-EVL-NEXT:    [[BROADCAST_SPLAT2:%.*]] = shufflevector <vscale x 16 x i32> [[BROADCAST_SPLATINSERT1]], <vscale x 16 x i32> poison, <vscale x 16 x i32> zeroinitializer
; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP2:%.*]] = icmp ugt <vscale x 16 x i32> [[VEC_IND]], [[BROADCAST_SPLAT]]
; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP3:%.*]] = shl i32 [[INDEX]], 2
; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP4:%.*]] = getelementptr i8, ptr [[P]], i32 [[TMP3]]
; PREDICATED_DATA-WITH-EVL-NEXT:    [[INTERLEAVE_EVL:%.*]] = mul nuw nsw i32 [[TMP1]], 4
; PREDICATED_DATA-WITH-EVL-NEXT:    [[INTERLEAVED_MASK:%.*]] = call <vscale x 64 x i1> @llvm.vector.interleave4.nxv64i1(<vscale x 16 x i1> [[TMP2]], <vscale x 16 x i1> [[TMP2]], <vscale x 16 x i1> [[TMP2]], <vscale x 16 x i1> [[TMP2]])
; PREDICATED_DATA-WITH-EVL-NEXT:    [[WIDE_VP_LOAD:%.*]] = call <vscale x 64 x i8> @llvm.vp.load.nxv64i8.p0(ptr align 1 [[TMP4]], <vscale x 64 x i1> [[INTERLEAVED_MASK]], i32 [[INTERLEAVE_EVL]])
; PREDICATED_DATA-WITH-EVL-NEXT:    [[STRIDED_VEC:%.*]] = call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.vector.deinterleave4.nxv64i8(<vscale x 64 x i8> [[WIDE_VP_LOAD]])
; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP5:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[STRIDED_VEC]], 0
; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[STRIDED_VEC]], 1
; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP7:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[STRIDED_VEC]], 2
; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP8:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[STRIDED_VEC]], 3
; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP9:%.*]] = icmp slt <vscale x 16 x i8> [[TMP5]], [[TMP6]]
; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP10:%.*]] = select <vscale x 16 x i1> [[TMP9]], <vscale x 16 x i8> [[TMP6]], <vscale x 16 x i8> [[TMP5]]
; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP11:%.*]] = sub <vscale x 16 x i8> zeroinitializer, [[TMP10]]
; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP12:%.*]] = icmp slt <vscale x 16 x i8> [[TMP7]], [[TMP8]]
; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP13:%.*]] = select <vscale x 16 x i1> [[TMP12]], <vscale x 16 x i8> [[TMP8]], <vscale x 16 x i8> [[TMP7]]
; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP14:%.*]] = sub <vscale x 16 x i8> zeroinitializer, [[TMP13]]
; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP15:%.*]] = getelementptr i8, ptr [[Q]], i32 [[TMP3]]
; PREDICATED_DATA-WITH-EVL-NEXT:    [[INTERLEAVE_EVL3:%.*]] = mul nuw nsw i32 [[TMP1]], 4
; PREDICATED_DATA-WITH-EVL-NEXT:    [[INTERLEAVED_MASK4:%.*]] = call <vscale x 64 x i1> @llvm.vector.interleave4.nxv64i1(<vscale x 16 x i1> [[TMP2]], <vscale x 16 x i1> [[TMP2]], <vscale x 16 x i1> [[TMP2]], <vscale x 16 x i1> [[TMP2]])
; PREDICATED_DATA-WITH-EVL-NEXT:    [[INTERLEAVED_VEC:%.*]] = call <vscale x 64 x i8> @llvm.vector.interleave4.nxv64i8(<vscale x 16 x i8> [[TMP10]], <vscale x 16 x i8> [[TMP11]], <vscale x 16 x i8> [[TMP13]], <vscale x 16 x i8> [[TMP14]])
; PREDICATED_DATA-WITH-EVL-NEXT:    call void @llvm.vp.store.nxv64i8.p0(<vscale x 64 x i8> [[INTERLEAVED_VEC]], ptr align 1 [[TMP15]], <vscale x 64 x i1> [[INTERLEAVED_MASK4]], i32 [[INTERLEAVE_EVL3]])
; PREDICATED_DATA-WITH-EVL-NEXT:    [[CURRENT_ITERATION_NEXT]] = add nuw i32 [[TMP1]], [[INDEX]]
; PREDICATED_DATA-WITH-EVL-NEXT:    [[AVL_NEXT]] = sub nuw i32 [[AVL]], [[TMP1]]
; PREDICATED_DATA-WITH-EVL-NEXT:    [[VEC_IND_NEXT]] = add nuw nsw <vscale x 16 x i32> [[VEC_IND]], [[BROADCAST_SPLAT2]]
; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP16:%.*]] = icmp eq i32 [[AVL_NEXT]], 0
; PREDICATED_DATA-WITH-EVL-NEXT:    br i1 [[TMP16]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP3:![0-9]+]]
; PREDICATED_DATA-WITH-EVL:       [[MIDDLE_BLOCK]]:
; PREDICATED_DATA-WITH-EVL-NEXT:    br label %[[FOR_END:.*]]
; PREDICATED_DATA-WITH-EVL:       [[FOR_END]]:
; PREDICATED_DATA-WITH-EVL-NEXT:    ret void
;
entry:
  %conv = zext i8 %guard to i32
  br label %for.body

for.body:
  %ix.024 = phi i32 [ 0, %entry ], [ %inc, %for.inc ]
  %cmp1 = icmp ugt i32 %ix.024, %conv
  br i1 %cmp1, label %if.then, label %for.inc

if.then:
  %idx0 = shl nuw nsw i32 %ix.024, 2
  %idx1 = add i32 %idx0, 1
  %idx2 = add i32 %idx0, 2
  %idx3 = add i32 %idx0, 3

  %array1idx0 = getelementptr inbounds i8, ptr %p, i32 %idx0
  %0 = load i8, ptr %array1idx0, align 1
  %array1idx1 = getelementptr inbounds i8, ptr %p, i32 %idx1
  %1 = load i8, ptr %array1idx1, align 1
  %array1idx2 = getelementptr inbounds i8, ptr %p, i32 %idx2
  %2 = load i8, ptr %array1idx2, align 1
  %array1idx3 = getelementptr inbounds i8, ptr %p, i32 %idx3
  %3 = load i8, ptr %array1idx3, align 1

  %cmp.i1 = icmp slt i8 %0, %1
  %spec.select.i1 = select i1 %cmp.i1, i8 %1, i8 %0
  %sub1 = sub i8 0, %spec.select.i1
  %cmp.i2 = icmp slt i8 %2, %3
  %spec.select.i2 = select i1 %cmp.i2, i8 %3, i8 %2
  %sub2 = sub i8 0, %spec.select.i2

  %array3idx0 = getelementptr inbounds i8, ptr %q, i32 %idx0
  store i8 %spec.select.i1, ptr %array3idx0, align 1
  %array3idx1 = getelementptr inbounds i8, ptr %q, i32 %idx1
  store i8 %sub1, ptr %array3idx1, align 1
  %array3idx2 = getelementptr inbounds i8, ptr %q, i32 %idx2
  store i8 %spec.select.i2, ptr %array3idx2, align 1
  %array3idx3 = getelementptr inbounds i8, ptr %q, i32 %idx3
  store i8 %sub2, ptr %array3idx3, align 1

  br label %for.inc

for.inc:
  %inc = add nuw nsw i32 %ix.024, 1
  %exitcond = icmp eq i32 %inc, 1024
  br i1 %exitcond, label %for.end, label %for.body

for.end:
  ret void
}