diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/interleaved-masked-access.ll b/llvm/test/Transforms/LoopVectorize/RISCV/interleaved-masked-access.ll
new file mode 100644
index 000000000000..1b0feef3e666
--- /dev/null
+++ b/llvm/test/Transforms/LoopVectorize/RISCV/interleaved-masked-access.ll
@@ -0,0 +1,408 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --filter-out-after "^scalar.ph:" --version 2
+; RUN: opt -mtriple=riscv64-none-linux-gnu -S -passes=loop-vectorize,instcombine -mattr=+v -prefer-predicate-over-epilogue=scalar-epilogue %s 2>&1 | FileCheck %s -check-prefix=SCALAR_EPILOGUE
+; RUN: opt -mtriple=riscv64-none-linux-gnu -S -passes=loop-vectorize,instcombine -mattr=+v -prefer-predicate-over-epilogue=predicate-dont-vectorize %s 2>&1 | FileCheck %s -check-prefix=PREDICATED_TAIL_FOLDING
+; RUN: opt -mtriple=riscv64-none-linux-gnu -S -passes=loop-vectorize,instcombine -mattr=+v -prefer-predicate-over-epilogue=predicate-dont-vectorize -force-tail-folding-style=data-with-evl %s 2>&1 | FileCheck %s -check-prefix=PREDICATED_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) {
+; 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 i32 [[TMP0]], 3
+; SCALAR_EPILOGUE-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ugt i32 [[TMP1]], 1024
+; SCALAR_EPILOGUE-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; SCALAR_EPILOGUE:       vector.ph:
+; SCALAR_EPILOGUE-NEXT:    [[TMP2:%.*]] = call i32 @llvm.vscale.i32()
+; SCALAR_EPILOGUE-NEXT:    [[TMP3:%.*]] = shl i32 [[TMP2]], 3
+; SCALAR_EPILOGUE-NEXT:    [[N_MOD_VF:%.*]] = urem i32 1024, [[TMP3]]
+; SCALAR_EPILOGUE-NEXT:    [[N_VEC:%.*]] = sub nuw nsw i32 1024, [[N_MOD_VF]]
+; SCALAR_EPILOGUE-NEXT:    [[TMP4:%.*]] = call i32 @llvm.vscale.i32()
+; SCALAR_EPILOGUE-NEXT:    [[TMP5:%.*]] = shl i32 [[TMP4]], 3
+; SCALAR_EPILOGUE-NEXT:    [[TMP6:%.*]] = call <vscale x 8 x i32> @llvm.stepvector.nxv8i32()
+; SCALAR_EPILOGUE-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 8 x i32> poison, i32 [[CONV]], i64 0
+; SCALAR_EPILOGUE-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 8 x i32> [[BROADCAST_SPLATINSERT]], <vscale x 8 x i32> poison, <vscale x 8 x i32> zeroinitializer
+; SCALAR_EPILOGUE-NEXT:    [[DOTSPLATINSERT:%.*]] = insertelement <vscale x 8 x i32> poison, i32 [[TMP5]], i64 0
+; SCALAR_EPILOGUE-NEXT:    [[DOTSPLAT:%.*]] = shufflevector <vscale x 8 x i32> [[DOTSPLATINSERT]], <vscale x 8 x i32> poison, <vscale x 8 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 8 x i32> [ [[TMP6]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
+; SCALAR_EPILOGUE-NEXT:    [[TMP7:%.*]] = icmp ugt <vscale x 8 x i32> [[VEC_IND]], [[BROADCAST_SPLAT]]
+; SCALAR_EPILOGUE-NEXT:    [[TMP8:%.*]] = shl nuw nsw <vscale x 8 x i32> [[VEC_IND]], splat (i32 1)
+; SCALAR_EPILOGUE-NEXT:    [[TMP9:%.*]] = zext nneg <vscale x 8 x i32> [[TMP8]] to <vscale x 8 x i64>
+; SCALAR_EPILOGUE-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i8, ptr [[P]], <vscale x 8 x i64> [[TMP9]]
+; SCALAR_EPILOGUE-NEXT:    [[WIDE_MASKED_GATHER:%.*]] = call <vscale x 8 x i8> @llvm.masked.gather.nxv8i8.nxv8p0(<vscale x 8 x ptr> [[TMP10]], i32 1, <vscale x 8 x i1> [[TMP7]], <vscale x 8 x i8> poison)
+; SCALAR_EPILOGUE-NEXT:    [[TMP11:%.*]] = or disjoint <vscale x 8 x i32> [[TMP8]], splat (i32 1)
+; SCALAR_EPILOGUE-NEXT:    [[TMP12:%.*]] = zext nneg <vscale x 8 x i32> [[TMP11]] to <vscale x 8 x i64>
+; SCALAR_EPILOGUE-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i8, ptr [[P]], <vscale x 8 x i64> [[TMP12]]
+; SCALAR_EPILOGUE-NEXT:    [[WIDE_MASKED_GATHER1:%.*]] = call <vscale x 8 x i8> @llvm.masked.gather.nxv8i8.nxv8p0(<vscale x 8 x ptr> [[TMP13]], i32 1, <vscale x 8 x i1> [[TMP7]], <vscale x 8 x i8> poison)
+; SCALAR_EPILOGUE-NEXT:    [[TMP14:%.*]] = call <vscale x 8 x i8> @llvm.smax.nxv8i8(<vscale x 8 x i8> [[WIDE_MASKED_GATHER]], <vscale x 8 x i8> [[WIDE_MASKED_GATHER1]])
+; SCALAR_EPILOGUE-NEXT:    [[TMP15:%.*]] = zext nneg <vscale x 8 x i32> [[TMP8]] to <vscale x 8 x i64>
+; SCALAR_EPILOGUE-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i8, ptr [[Q]], <vscale x 8 x i64> [[TMP15]]
+; SCALAR_EPILOGUE-NEXT:    call void @llvm.masked.scatter.nxv8i8.nxv8p0(<vscale x 8 x i8> [[TMP14]], <vscale x 8 x ptr> [[TMP16]], i32 1, <vscale x 8 x i1> [[TMP7]])
+; SCALAR_EPILOGUE-NEXT:    [[TMP17:%.*]] = sub <vscale x 8 x i8> zeroinitializer, [[TMP14]]
+; SCALAR_EPILOGUE-NEXT:    [[TMP18:%.*]] = zext nneg <vscale x 8 x i32> [[TMP11]] to <vscale x 8 x i64>
+; SCALAR_EPILOGUE-NEXT:    [[TMP19:%.*]] = getelementptr inbounds i8, ptr [[Q]], <vscale x 8 x i64> [[TMP18]]
+; SCALAR_EPILOGUE-NEXT:    call void @llvm.masked.scatter.nxv8i8.nxv8p0(<vscale x 8 x i8> [[TMP17]], <vscale x 8 x ptr> [[TMP19]], i32 1, <vscale x 8 x i1> [[TMP7]])
+; SCALAR_EPILOGUE-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], [[TMP5]]
+; SCALAR_EPILOGUE-NEXT:    [[VEC_IND_NEXT]] = add <vscale x 8 x i32> [[VEC_IND]], [[DOTSPLAT]]
+; SCALAR_EPILOGUE-NEXT:    [[TMP20:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
+; SCALAR_EPILOGUE-NEXT:    br i1 [[TMP20]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
+; SCALAR_EPILOGUE:       middle.block:
+; SCALAR_EPILOGUE-NEXT:    [[CMP_N:%.*]] = icmp eq i32 [[N_MOD_VF]], 0
+; SCALAR_EPILOGUE-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; SCALAR_EPILOGUE:       scalar.ph:
+;
+; PREDICATED_TAIL_FOLDING-LABEL: define void @masked_strided_factor2
+; PREDICATED_TAIL_FOLDING-SAME: (ptr noalias readonly captures(none) [[P:%.*]], ptr noalias captures(none) [[Q:%.*]], i8 zeroext [[GUARD:%.*]]) #[[ATTR0:[0-9]+]] {
+; PREDICATED_TAIL_FOLDING-NEXT:  entry:
+; PREDICATED_TAIL_FOLDING-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; PREDICATED_TAIL_FOLDING:       vector.ph:
+; PREDICATED_TAIL_FOLDING-NEXT:    [[CONV:%.*]] = zext i8 [[GUARD]] to i32
+; PREDICATED_TAIL_FOLDING-NEXT:    [[TMP0:%.*]] = call i32 @llvm.vscale.i32()
+; PREDICATED_TAIL_FOLDING-NEXT:    [[TMP1:%.*]] = shl i32 [[TMP0]], 3
+; PREDICATED_TAIL_FOLDING-NEXT:    [[N_RND_UP:%.*]] = add i32 [[TMP1]], 1023
+; PREDICATED_TAIL_FOLDING-NEXT:    [[N_MOD_VF:%.*]] = urem i32 [[N_RND_UP]], [[TMP1]]
+; PREDICATED_TAIL_FOLDING-NEXT:    [[N_VEC:%.*]] = sub i32 [[N_RND_UP]], [[N_MOD_VF]]
+; PREDICATED_TAIL_FOLDING-NEXT:    [[TMP2:%.*]] = call i32 @llvm.vscale.i32()
+; PREDICATED_TAIL_FOLDING-NEXT:    [[TMP3:%.*]] = shl i32 [[TMP2]], 3
+; PREDICATED_TAIL_FOLDING-NEXT:    [[TMP4:%.*]] = call <vscale x 8 x i32> @llvm.stepvector.nxv8i32()
+; PREDICATED_TAIL_FOLDING-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 8 x i32> poison, i32 [[CONV]], i64 0
+; PREDICATED_TAIL_FOLDING-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 8 x i32> [[BROADCAST_SPLATINSERT]], <vscale x 8 x i32> poison, <vscale x 8 x i32> zeroinitializer
+; PREDICATED_TAIL_FOLDING-NEXT:    [[DOTSPLATINSERT:%.*]] = insertelement <vscale x 8 x i32> poison, i32 [[TMP3]], i64 0
+; PREDICATED_TAIL_FOLDING-NEXT:    [[DOTSPLAT:%.*]] = shufflevector <vscale x 8 x i32> [[DOTSPLATINSERT]], <vscale x 8 x i32> poison, <vscale x 8 x i32> zeroinitializer
+; PREDICATED_TAIL_FOLDING-NEXT:    br label [[VECTOR_BODY:%.*]]
+; PREDICATED_TAIL_FOLDING:       vector.body:
+; PREDICATED_TAIL_FOLDING-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; PREDICATED_TAIL_FOLDING-NEXT:    [[VEC_IND:%.*]] = phi <vscale x 8 x i32> [ [[TMP4]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
+; PREDICATED_TAIL_FOLDING-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = call <vscale x 8 x i1> @llvm.get.active.lane.mask.nxv8i1.i32(i32 [[INDEX]], i32 1024)
+; PREDICATED_TAIL_FOLDING-NEXT:    [[TMP5:%.*]] = icmp ugt <vscale x 8 x i32> [[VEC_IND]], [[BROADCAST_SPLAT]]
+; PREDICATED_TAIL_FOLDING-NEXT:    [[TMP6:%.*]] = select <vscale x 8 x i1> [[ACTIVE_LANE_MASK]], <vscale x 8 x i1> [[TMP5]], <vscale x 8 x i1> zeroinitializer
+; PREDICATED_TAIL_FOLDING-NEXT:    [[TMP7:%.*]] = shl nuw nsw <vscale x 8 x i32> [[VEC_IND]], splat (i32 1)
+; PREDICATED_TAIL_FOLDING-NEXT:    [[TMP8:%.*]] = zext nneg <vscale x 8 x i32> [[TMP7]] to <vscale x 8 x i64>
+; PREDICATED_TAIL_FOLDING-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i8, ptr [[P]], <vscale x 8 x i64> [[TMP8]]
+; PREDICATED_TAIL_FOLDING-NEXT:    [[WIDE_MASKED_GATHER:%.*]] = call <vscale x 8 x i8> @llvm.masked.gather.nxv8i8.nxv8p0(<vscale x 8 x ptr> [[TMP9]], i32 1, <vscale x 8 x i1> [[TMP6]], <vscale x 8 x i8> poison)
+; PREDICATED_TAIL_FOLDING-NEXT:    [[TMP10:%.*]] = or disjoint <vscale x 8 x i32> [[TMP7]], splat (i32 1)
+; PREDICATED_TAIL_FOLDING-NEXT:    [[TMP11:%.*]] = zext nneg <vscale x 8 x i32> [[TMP10]] to <vscale x 8 x i64>
+; PREDICATED_TAIL_FOLDING-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i8, ptr [[P]], <vscale x 8 x i64> [[TMP11]]
+; PREDICATED_TAIL_FOLDING-NEXT:    [[WIDE_MASKED_GATHER1:%.*]] = call <vscale x 8 x i8> @llvm.masked.gather.nxv8i8.nxv8p0(<vscale x 8 x ptr> [[TMP12]], i32 1, <vscale x 8 x i1> [[TMP6]], <vscale x 8 x i8> poison)
+; PREDICATED_TAIL_FOLDING-NEXT:    [[TMP13:%.*]] = call <vscale x 8 x i8> @llvm.smax.nxv8i8(<vscale x 8 x i8> [[WIDE_MASKED_GATHER]], <vscale x 8 x i8> [[WIDE_MASKED_GATHER1]])
+; PREDICATED_TAIL_FOLDING-NEXT:    [[TMP14:%.*]] = zext nneg <vscale x 8 x i32> [[TMP7]] to <vscale x 8 x i64>
+; PREDICATED_TAIL_FOLDING-NEXT:    [[TMP15:%.*]] = getelementptr inbounds i8, ptr [[Q]], <vscale x 8 x i64> [[TMP14]]
+; PREDICATED_TAIL_FOLDING-NEXT:    call void @llvm.masked.scatter.nxv8i8.nxv8p0(<vscale x 8 x i8> [[TMP13]], <vscale x 8 x ptr> [[TMP15]], i32 1, <vscale x 8 x i1> [[TMP6]])
+; PREDICATED_TAIL_FOLDING-NEXT:    [[TMP16:%.*]] = sub <vscale x 8 x i8> zeroinitializer, [[TMP13]]
+; PREDICATED_TAIL_FOLDING-NEXT:    [[TMP17:%.*]] = zext nneg <vscale x 8 x i32> [[TMP10]] to <vscale x 8 x i64>
+; PREDICATED_TAIL_FOLDING-NEXT:    [[TMP18:%.*]] = getelementptr inbounds i8, ptr [[Q]], <vscale x 8 x i64> [[TMP17]]
+; PREDICATED_TAIL_FOLDING-NEXT:    call void @llvm.masked.scatter.nxv8i8.nxv8p0(<vscale x 8 x i8> [[TMP16]], <vscale x 8 x ptr> [[TMP18]], i32 1, <vscale x 8 x i1> [[TMP6]])
+; PREDICATED_TAIL_FOLDING-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], [[TMP3]]
+; PREDICATED_TAIL_FOLDING-NEXT:    [[VEC_IND_NEXT]] = add <vscale x 8 x i32> [[VEC_IND]], [[DOTSPLAT]]
+; PREDICATED_TAIL_FOLDING-NEXT:    [[TMP19:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
+; PREDICATED_TAIL_FOLDING-NEXT:    br i1 [[TMP19]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
+; PREDICATED_TAIL_FOLDING:       middle.block:
+; PREDICATED_TAIL_FOLDING-NEXT:    br label [[FOR_END:%.*]]
+; PREDICATED_TAIL_FOLDING:       scalar.ph:
+;
+; PREDICATED_EVL-LABEL: define void @masked_strided_factor2
+; PREDICATED_EVL-SAME: (ptr noalias readonly captures(none) [[P:%.*]], ptr noalias captures(none) [[Q:%.*]], i8 zeroext [[GUARD:%.*]]) #[[ATTR0:[0-9]+]] {
+; PREDICATED_EVL-NEXT:  entry:
+; PREDICATED_EVL-NEXT:    [[CONV:%.*]] = zext i8 [[GUARD]] to i32
+; PREDICATED_EVL-NEXT:    br label [[FOR_BODY:%.*]]
+; PREDICATED_EVL:       for.body:
+; PREDICATED_EVL-NEXT:    [[IX_024:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[INC:%.*]], [[FOR_INC:%.*]] ]
+; PREDICATED_EVL-NEXT:    [[CMP1:%.*]] = icmp samesign ugt i32 [[IX_024]], [[CONV]]
+; PREDICATED_EVL-NEXT:    br i1 [[CMP1]], label [[IF_THEN:%.*]], label [[FOR_INC]]
+; PREDICATED_EVL:       if.then:
+; PREDICATED_EVL-NEXT:    [[MUL:%.*]] = shl nuw nsw i32 [[IX_024]], 1
+; PREDICATED_EVL-NEXT:    [[TMP0:%.*]] = zext nneg i32 [[MUL]] to i64
+; PREDICATED_EVL-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds nuw i8, ptr [[P]], i64 [[TMP0]]
+; PREDICATED_EVL-NEXT:    [[TMP1:%.*]] = load i8, ptr [[ARRAYIDX]], align 1
+; PREDICATED_EVL-NEXT:    [[ADD:%.*]] = or disjoint i32 [[MUL]], 1
+; PREDICATED_EVL-NEXT:    [[TMP2:%.*]] = zext nneg i32 [[ADD]] to i64
+; PREDICATED_EVL-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds nuw i8, ptr [[P]], i64 [[TMP2]]
+; PREDICATED_EVL-NEXT:    [[TMP3:%.*]] = load i8, ptr [[ARRAYIDX4]], align 1
+; PREDICATED_EVL-NEXT:    [[SPEC_SELECT_I:%.*]] = call i8 @llvm.smax.i8(i8 [[TMP1]], i8 [[TMP3]])
+; PREDICATED_EVL-NEXT:    [[TMP4:%.*]] = zext nneg i32 [[MUL]] to i64
+; PREDICATED_EVL-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds nuw i8, ptr [[Q]], i64 [[TMP4]]
+; PREDICATED_EVL-NEXT:    store i8 [[SPEC_SELECT_I]], ptr [[ARRAYIDX6]], align 1
+; PREDICATED_EVL-NEXT:    [[SUB:%.*]] = sub i8 0, [[SPEC_SELECT_I]]
+; PREDICATED_EVL-NEXT:    [[TMP5:%.*]] = zext nneg i32 [[ADD]] to i64
+; PREDICATED_EVL-NEXT:    [[ARRAYIDX11:%.*]] = getelementptr inbounds nuw i8, ptr [[Q]], i64 [[TMP5]]
+; PREDICATED_EVL-NEXT:    store i8 [[SUB]], ptr [[ARRAYIDX11]], align 1
+; PREDICATED_EVL-NEXT:    br label [[FOR_INC]]
+; PREDICATED_EVL:       for.inc:
+; PREDICATED_EVL-NEXT:    [[INC]] = add nuw nsw i32 [[IX_024]], 1
+; PREDICATED_EVL-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[INC]], 1024
+; PREDICATED_EVL-NEXT:    br i1 [[EXITCOND]], label [[FOR_END:%.*]], label [[FOR_BODY]]
+; PREDICATED_EVL:       for.end:
+; PREDICATED_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) {
+; 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 i32 [[TMP0]], 3
+; SCALAR_EPILOGUE-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ugt i32 [[TMP1]], 1024
+; SCALAR_EPILOGUE-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; SCALAR_EPILOGUE:       vector.ph:
+; SCALAR_EPILOGUE-NEXT:    [[TMP2:%.*]] = call i32 @llvm.vscale.i32()
+; SCALAR_EPILOGUE-NEXT:    [[TMP3:%.*]] = shl i32 [[TMP2]], 3
+; SCALAR_EPILOGUE-NEXT:    [[N_MOD_VF:%.*]] = urem i32 1024, [[TMP3]]
+; SCALAR_EPILOGUE-NEXT:    [[N_VEC:%.*]] = sub nuw nsw i32 1024, [[N_MOD_VF]]
+; SCALAR_EPILOGUE-NEXT:    [[TMP4:%.*]] = call i32 @llvm.vscale.i32()
+; SCALAR_EPILOGUE-NEXT:    [[TMP5:%.*]] = shl i32 [[TMP4]], 3
+; SCALAR_EPILOGUE-NEXT:    [[TMP6:%.*]] = call <vscale x 8 x i32> @llvm.stepvector.nxv8i32()
+; SCALAR_EPILOGUE-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 8 x i32> poison, i32 [[CONV]], i64 0
+; SCALAR_EPILOGUE-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 8 x i32> [[BROADCAST_SPLATINSERT]], <vscale x 8 x i32> poison, <vscale x 8 x i32> zeroinitializer
+; SCALAR_EPILOGUE-NEXT:    [[DOTSPLATINSERT:%.*]] = insertelement <vscale x 8 x i32> poison, i32 [[TMP5]], i64 0
+; SCALAR_EPILOGUE-NEXT:    [[DOTSPLAT:%.*]] = shufflevector <vscale x 8 x i32> [[DOTSPLATINSERT]], <vscale x 8 x i32> poison, <vscale x 8 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 8 x i32> [ [[TMP6]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
+; SCALAR_EPILOGUE-NEXT:    [[TMP7:%.*]] = icmp ugt <vscale x 8 x i32> [[VEC_IND]], [[BROADCAST_SPLAT]]
+; SCALAR_EPILOGUE-NEXT:    [[TMP8:%.*]] = shl nuw nsw <vscale x 8 x i32> [[VEC_IND]], splat (i32 2)
+; SCALAR_EPILOGUE-NEXT:    [[TMP9:%.*]] = or disjoint <vscale x 8 x i32> [[TMP8]], splat (i32 1)
+; SCALAR_EPILOGUE-NEXT:    [[TMP10:%.*]] = or disjoint <vscale x 8 x i32> [[TMP8]], splat (i32 2)
+; SCALAR_EPILOGUE-NEXT:    [[TMP11:%.*]] = or disjoint <vscale x 8 x i32> [[TMP8]], splat (i32 3)
+; SCALAR_EPILOGUE-NEXT:    [[TMP12:%.*]] = zext nneg <vscale x 8 x i32> [[TMP8]] to <vscale x 8 x i64>
+; SCALAR_EPILOGUE-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i8, ptr [[P]], <vscale x 8 x i64> [[TMP12]]
+; SCALAR_EPILOGUE-NEXT:    [[WIDE_MASKED_GATHER:%.*]] = call <vscale x 8 x i8> @llvm.masked.gather.nxv8i8.nxv8p0(<vscale x 8 x ptr> [[TMP13]], i32 1, <vscale x 8 x i1> [[TMP7]], <vscale x 8 x i8> poison)
+; SCALAR_EPILOGUE-NEXT:    [[TMP14:%.*]] = zext nneg <vscale x 8 x i32> [[TMP9]] to <vscale x 8 x i64>
+; SCALAR_EPILOGUE-NEXT:    [[TMP15:%.*]] = getelementptr inbounds i8, ptr [[P]], <vscale x 8 x i64> [[TMP14]]
+; SCALAR_EPILOGUE-NEXT:    [[WIDE_MASKED_GATHER1:%.*]] = call <vscale x 8 x i8> @llvm.masked.gather.nxv8i8.nxv8p0(<vscale x 8 x ptr> [[TMP15]], i32 1, <vscale x 8 x i1> [[TMP7]], <vscale x 8 x i8> poison)
+; SCALAR_EPILOGUE-NEXT:    [[TMP16:%.*]] = zext nneg <vscale x 8 x i32> [[TMP10]] to <vscale x 8 x i64>
+; SCALAR_EPILOGUE-NEXT:    [[TMP17:%.*]] = getelementptr inbounds i8, ptr [[P]], <vscale x 8 x i64> [[TMP16]]
+; SCALAR_EPILOGUE-NEXT:    [[WIDE_MASKED_GATHER2:%.*]] = call <vscale x 8 x i8> @llvm.masked.gather.nxv8i8.nxv8p0(<vscale x 8 x ptr> [[TMP17]], i32 1, <vscale x 8 x i1> [[TMP7]], <vscale x 8 x i8> poison)
+; SCALAR_EPILOGUE-NEXT:    [[TMP18:%.*]] = zext nneg <vscale x 8 x i32> [[TMP11]] to <vscale x 8 x i64>
+; SCALAR_EPILOGUE-NEXT:    [[TMP19:%.*]] = getelementptr inbounds i8, ptr [[P]], <vscale x 8 x i64> [[TMP18]]
+; SCALAR_EPILOGUE-NEXT:    [[WIDE_MASKED_GATHER3:%.*]] = call <vscale x 8 x i8> @llvm.masked.gather.nxv8i8.nxv8p0(<vscale x 8 x ptr> [[TMP19]], i32 1, <vscale x 8 x i1> [[TMP7]], <vscale x 8 x i8> poison)
+; SCALAR_EPILOGUE-NEXT:    [[TMP20:%.*]] = call <vscale x 8 x i8> @llvm.smax.nxv8i8(<vscale x 8 x i8> [[WIDE_MASKED_GATHER]], <vscale x 8 x i8> [[WIDE_MASKED_GATHER1]])
+; SCALAR_EPILOGUE-NEXT:    [[TMP21:%.*]] = sub <vscale x 8 x i8> zeroinitializer, [[TMP20]]
+; SCALAR_EPILOGUE-NEXT:    [[TMP22:%.*]] = call <vscale x 8 x i8> @llvm.smax.nxv8i8(<vscale x 8 x i8> [[WIDE_MASKED_GATHER2]], <vscale x 8 x i8> [[WIDE_MASKED_GATHER3]])
+; SCALAR_EPILOGUE-NEXT:    [[TMP23:%.*]] = sub <vscale x 8 x i8> zeroinitializer, [[TMP22]]
+; SCALAR_EPILOGUE-NEXT:    [[TMP24:%.*]] = zext nneg <vscale x 8 x i32> [[TMP8]] to <vscale x 8 x i64>
+; SCALAR_EPILOGUE-NEXT:    [[TMP25:%.*]] = getelementptr inbounds i8, ptr [[Q]], <vscale x 8 x i64> [[TMP24]]
+; SCALAR_EPILOGUE-NEXT:    call void @llvm.masked.scatter.nxv8i8.nxv8p0(<vscale x 8 x i8> [[TMP20]], <vscale x 8 x ptr> [[TMP25]], i32 1, <vscale x 8 x i1> [[TMP7]])
+; SCALAR_EPILOGUE-NEXT:    [[TMP26:%.*]] = zext nneg <vscale x 8 x i32> [[TMP9]] to <vscale x 8 x i64>
+; SCALAR_EPILOGUE-NEXT:    [[TMP27:%.*]] = getelementptr inbounds i8, ptr [[Q]], <vscale x 8 x i64> [[TMP26]]
+; SCALAR_EPILOGUE-NEXT:    call void @llvm.masked.scatter.nxv8i8.nxv8p0(<vscale x 8 x i8> [[TMP21]], <vscale x 8 x ptr> [[TMP27]], i32 1, <vscale x 8 x i1> [[TMP7]])
+; SCALAR_EPILOGUE-NEXT:    [[TMP28:%.*]] = zext nneg <vscale x 8 x i32> [[TMP10]] to <vscale x 8 x i64>
+; SCALAR_EPILOGUE-NEXT:    [[TMP29:%.*]] = getelementptr inbounds i8, ptr [[Q]], <vscale x 8 x i64> [[TMP28]]
+; SCALAR_EPILOGUE-NEXT:    call void @llvm.masked.scatter.nxv8i8.nxv8p0(<vscale x 8 x i8> [[TMP22]], <vscale x 8 x ptr> [[TMP29]], i32 1, <vscale x 8 x i1> [[TMP7]])
+; SCALAR_EPILOGUE-NEXT:    [[TMP30:%.*]] = zext nneg <vscale x 8 x i32> [[TMP11]] to <vscale x 8 x i64>
+; SCALAR_EPILOGUE-NEXT:    [[TMP31:%.*]] = getelementptr inbounds i8, ptr [[Q]], <vscale x 8 x i64> [[TMP30]]
+; SCALAR_EPILOGUE-NEXT:    call void @llvm.masked.scatter.nxv8i8.nxv8p0(<vscale x 8 x i8> [[TMP23]], <vscale x 8 x ptr> [[TMP31]], i32 1, <vscale x 8 x i1> [[TMP7]])
+; SCALAR_EPILOGUE-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], [[TMP5]]
+; SCALAR_EPILOGUE-NEXT:    [[VEC_IND_NEXT]] = add <vscale x 8 x i32> [[VEC_IND]], [[DOTSPLAT]]
+; SCALAR_EPILOGUE-NEXT:    [[TMP32:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
+; SCALAR_EPILOGUE-NEXT:    br i1 [[TMP32]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
+; SCALAR_EPILOGUE:       middle.block:
+; SCALAR_EPILOGUE-NEXT:    [[CMP_N:%.*]] = icmp eq i32 [[N_MOD_VF]], 0
+; SCALAR_EPILOGUE-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; SCALAR_EPILOGUE:       scalar.ph:
+;
+; PREDICATED_TAIL_FOLDING-LABEL: define void @masked_strided_factor4
+; PREDICATED_TAIL_FOLDING-SAME: (ptr noalias readonly captures(none) [[P:%.*]], ptr noalias captures(none) [[Q:%.*]], i8 zeroext [[GUARD:%.*]]) #[[ATTR0]] {
+; PREDICATED_TAIL_FOLDING-NEXT:  entry:
+; PREDICATED_TAIL_FOLDING-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; PREDICATED_TAIL_FOLDING:       vector.ph:
+; PREDICATED_TAIL_FOLDING-NEXT:    [[CONV:%.*]] = zext i8 [[GUARD]] to i32
+; PREDICATED_TAIL_FOLDING-NEXT:    [[TMP0:%.*]] = call i32 @llvm.vscale.i32()
+; PREDICATED_TAIL_FOLDING-NEXT:    [[TMP1:%.*]] = shl i32 [[TMP0]], 3
+; PREDICATED_TAIL_FOLDING-NEXT:    [[N_RND_UP:%.*]] = add i32 [[TMP1]], 1023
+; PREDICATED_TAIL_FOLDING-NEXT:    [[N_MOD_VF:%.*]] = urem i32 [[N_RND_UP]], [[TMP1]]
+; PREDICATED_TAIL_FOLDING-NEXT:    [[N_VEC:%.*]] = sub i32 [[N_RND_UP]], [[N_MOD_VF]]
+; PREDICATED_TAIL_FOLDING-NEXT:    [[TMP2:%.*]] = call i32 @llvm.vscale.i32()
+; PREDICATED_TAIL_FOLDING-NEXT:    [[TMP3:%.*]] = shl i32 [[TMP2]], 3
+; PREDICATED_TAIL_FOLDING-NEXT:    [[TMP4:%.*]] = call <vscale x 8 x i32> @llvm.stepvector.nxv8i32()
+; PREDICATED_TAIL_FOLDING-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 8 x i32> poison, i32 [[CONV]], i64 0
+; PREDICATED_TAIL_FOLDING-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 8 x i32> [[BROADCAST_SPLATINSERT]], <vscale x 8 x i32> poison, <vscale x 8 x i32> zeroinitializer
+; PREDICATED_TAIL_FOLDING-NEXT:    [[DOTSPLATINSERT:%.*]] = insertelement <vscale x 8 x i32> poison, i32 [[TMP3]], i64 0
+; PREDICATED_TAIL_FOLDING-NEXT:    [[DOTSPLAT:%.*]] = shufflevector <vscale x 8 x i32> [[DOTSPLATINSERT]], <vscale x 8 x i32> poison, <vscale x 8 x i32> zeroinitializer
+; PREDICATED_TAIL_FOLDING-NEXT:    br label [[VECTOR_BODY:%.*]]
+; PREDICATED_TAIL_FOLDING:       vector.body:
+; PREDICATED_TAIL_FOLDING-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; PREDICATED_TAIL_FOLDING-NEXT:    [[VEC_IND:%.*]] = phi <vscale x 8 x i32> [ [[TMP4]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
+; PREDICATED_TAIL_FOLDING-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = call <vscale x 8 x i1> @llvm.get.active.lane.mask.nxv8i1.i32(i32 [[INDEX]], i32 1024)
+; PREDICATED_TAIL_FOLDING-NEXT:    [[TMP5:%.*]] = icmp ugt <vscale x 8 x i32> [[VEC_IND]], [[BROADCAST_SPLAT]]
+; PREDICATED_TAIL_FOLDING-NEXT:    [[TMP6:%.*]] = select <vscale x 8 x i1> [[ACTIVE_LANE_MASK]], <vscale x 8 x i1> [[TMP5]], <vscale x 8 x i1> zeroinitializer
+; PREDICATED_TAIL_FOLDING-NEXT:    [[TMP7:%.*]] = shl nuw nsw <vscale x 8 x i32> [[VEC_IND]], splat (i32 2)
+; PREDICATED_TAIL_FOLDING-NEXT:    [[TMP8:%.*]] = or disjoint <vscale x 8 x i32> [[TMP7]], splat (i32 1)
+; PREDICATED_TAIL_FOLDING-NEXT:    [[TMP9:%.*]] = or disjoint <vscale x 8 x i32> [[TMP7]], splat (i32 2)
+; PREDICATED_TAIL_FOLDING-NEXT:    [[TMP10:%.*]] = or disjoint <vscale x 8 x i32> [[TMP7]], splat (i32 3)
+; PREDICATED_TAIL_FOLDING-NEXT:    [[TMP11:%.*]] = zext nneg <vscale x 8 x i32> [[TMP7]] to <vscale x 8 x i64>
+; PREDICATED_TAIL_FOLDING-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i8, ptr [[P]], <vscale x 8 x i64> [[TMP11]]
+; PREDICATED_TAIL_FOLDING-NEXT:    [[WIDE_MASKED_GATHER:%.*]] = call <vscale x 8 x i8> @llvm.masked.gather.nxv8i8.nxv8p0(<vscale x 8 x ptr> [[TMP12]], i32 1, <vscale x 8 x i1> [[TMP6]], <vscale x 8 x i8> poison)
+; PREDICATED_TAIL_FOLDING-NEXT:    [[TMP13:%.*]] = zext nneg <vscale x 8 x i32> [[TMP8]] to <vscale x 8 x i64>
+; PREDICATED_TAIL_FOLDING-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i8, ptr [[P]], <vscale x 8 x i64> [[TMP13]]
+; PREDICATED_TAIL_FOLDING-NEXT:    [[WIDE_MASKED_GATHER1:%.*]] = call <vscale x 8 x i8> @llvm.masked.gather.nxv8i8.nxv8p0(<vscale x 8 x ptr> [[TMP14]], i32 1, <vscale x 8 x i1> [[TMP6]], <vscale x 8 x i8> poison)
+; PREDICATED_TAIL_FOLDING-NEXT:    [[TMP15:%.*]] = zext nneg <vscale x 8 x i32> [[TMP9]] to <vscale x 8 x i64>
+; PREDICATED_TAIL_FOLDING-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i8, ptr [[P]], <vscale x 8 x i64> [[TMP15]]
+; PREDICATED_TAIL_FOLDING-NEXT:    [[WIDE_MASKED_GATHER2:%.*]] = call <vscale x 8 x i8> @llvm.masked.gather.nxv8i8.nxv8p0(<vscale x 8 x ptr> [[TMP16]], i32 1, <vscale x 8 x i1> [[TMP6]], <vscale x 8 x i8> poison)
+; PREDICATED_TAIL_FOLDING-NEXT:    [[TMP17:%.*]] = zext nneg <vscale x 8 x i32> [[TMP10]] to <vscale x 8 x i64>
+; PREDICATED_TAIL_FOLDING-NEXT:    [[TMP18:%.*]] = getelementptr inbounds i8, ptr [[P]], <vscale x 8 x i64> [[TMP17]]
+; PREDICATED_TAIL_FOLDING-NEXT:    [[WIDE_MASKED_GATHER3:%.*]] = call <vscale x 8 x i8> @llvm.masked.gather.nxv8i8.nxv8p0(<vscale x 8 x ptr> [[TMP18]], i32 1, <vscale x 8 x i1> [[TMP6]], <vscale x 8 x i8> poison)
+; PREDICATED_TAIL_FOLDING-NEXT:    [[TMP19:%.*]] = call <vscale x 8 x i8> @llvm.smax.nxv8i8(<vscale x 8 x i8> [[WIDE_MASKED_GATHER]], <vscale x 8 x i8> [[WIDE_MASKED_GATHER1]])
+; PREDICATED_TAIL_FOLDING-NEXT:    [[TMP20:%.*]] = sub <vscale x 8 x i8> zeroinitializer, [[TMP19]]
+; PREDICATED_TAIL_FOLDING-NEXT:    [[TMP21:%.*]] = call <vscale x 8 x i8> @llvm.smax.nxv8i8(<vscale x 8 x i8> [[WIDE_MASKED_GATHER2]], <vscale x 8 x i8> [[WIDE_MASKED_GATHER3]])
+; PREDICATED_TAIL_FOLDING-NEXT:    [[TMP22:%.*]] = sub <vscale x 8 x i8> zeroinitializer, [[TMP21]]
+; PREDICATED_TAIL_FOLDING-NEXT:    [[TMP23:%.*]] = zext nneg <vscale x 8 x i32> [[TMP7]] to <vscale x 8 x i64>
+; PREDICATED_TAIL_FOLDING-NEXT:    [[TMP24:%.*]] = getelementptr inbounds i8, ptr [[Q]], <vscale x 8 x i64> [[TMP23]]
+; PREDICATED_TAIL_FOLDING-NEXT:    call void @llvm.masked.scatter.nxv8i8.nxv8p0(<vscale x 8 x i8> [[TMP19]], <vscale x 8 x ptr> [[TMP24]], i32 1, <vscale x 8 x i1> [[TMP6]])
+; PREDICATED_TAIL_FOLDING-NEXT:    [[TMP25:%.*]] = zext nneg <vscale x 8 x i32> [[TMP8]] to <vscale x 8 x i64>
+; PREDICATED_TAIL_FOLDING-NEXT:    [[TMP26:%.*]] = getelementptr inbounds i8, ptr [[Q]], <vscale x 8 x i64> [[TMP25]]
+; PREDICATED_TAIL_FOLDING-NEXT:    call void @llvm.masked.scatter.nxv8i8.nxv8p0(<vscale x 8 x i8> [[TMP20]], <vscale x 8 x ptr> [[TMP26]], i32 1, <vscale x 8 x i1> [[TMP6]])
+; PREDICATED_TAIL_FOLDING-NEXT:    [[TMP27:%.*]] = zext nneg <vscale x 8 x i32> [[TMP9]] to <vscale x 8 x i64>
+; PREDICATED_TAIL_FOLDING-NEXT:    [[TMP28:%.*]] = getelementptr inbounds i8, ptr [[Q]], <vscale x 8 x i64> [[TMP27]]
+; PREDICATED_TAIL_FOLDING-NEXT:    call void @llvm.masked.scatter.nxv8i8.nxv8p0(<vscale x 8 x i8> [[TMP21]], <vscale x 8 x ptr> [[TMP28]], i32 1, <vscale x 8 x i1> [[TMP6]])
+; PREDICATED_TAIL_FOLDING-NEXT:    [[TMP29:%.*]] = zext nneg <vscale x 8 x i32> [[TMP10]] to <vscale x 8 x i64>
+; PREDICATED_TAIL_FOLDING-NEXT:    [[TMP30:%.*]] = getelementptr inbounds i8, ptr [[Q]], <vscale x 8 x i64> [[TMP29]]
+; PREDICATED_TAIL_FOLDING-NEXT:    call void @llvm.masked.scatter.nxv8i8.nxv8p0(<vscale x 8 x i8> [[TMP22]], <vscale x 8 x ptr> [[TMP30]], i32 1, <vscale x 8 x i1> [[TMP6]])
+; PREDICATED_TAIL_FOLDING-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], [[TMP3]]
+; PREDICATED_TAIL_FOLDING-NEXT:    [[VEC_IND_NEXT]] = add <vscale x 8 x i32> [[VEC_IND]], [[DOTSPLAT]]
+; PREDICATED_TAIL_FOLDING-NEXT:    [[TMP31:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
+; PREDICATED_TAIL_FOLDING-NEXT:    br i1 [[TMP31]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
+; PREDICATED_TAIL_FOLDING:       middle.block:
+; PREDICATED_TAIL_FOLDING-NEXT:    br label [[FOR_END:%.*]]
+; PREDICATED_TAIL_FOLDING:       scalar.ph:
+;
+; PREDICATED_EVL-LABEL: define void @masked_strided_factor4
+; PREDICATED_EVL-SAME: (ptr noalias readonly captures(none) [[P:%.*]], ptr noalias captures(none) [[Q:%.*]], i8 zeroext [[GUARD:%.*]]) #[[ATTR0]] {
+; PREDICATED_EVL-NEXT:  entry:
+; PREDICATED_EVL-NEXT:    [[CONV:%.*]] = zext i8 [[GUARD]] to i32
+; PREDICATED_EVL-NEXT:    br label [[FOR_BODY:%.*]]
+; PREDICATED_EVL:       for.body:
+; PREDICATED_EVL-NEXT:    [[IX_024:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[INC:%.*]], [[FOR_INC:%.*]] ]
+; PREDICATED_EVL-NEXT:    [[CMP1:%.*]] = icmp samesign ugt i32 [[IX_024]], [[CONV]]
+; PREDICATED_EVL-NEXT:    br i1 [[CMP1]], label [[IF_THEN:%.*]], label [[FOR_INC]]
+; PREDICATED_EVL:       if.then:
+; PREDICATED_EVL-NEXT:    [[IDX0:%.*]] = shl nuw nsw i32 [[IX_024]], 2
+; PREDICATED_EVL-NEXT:    [[IDX1:%.*]] = or disjoint i32 [[IDX0]], 1
+; PREDICATED_EVL-NEXT:    [[IDX2:%.*]] = or disjoint i32 [[IDX0]], 2
+; PREDICATED_EVL-NEXT:    [[IDX3:%.*]] = or disjoint i32 [[IDX0]], 3
+; PREDICATED_EVL-NEXT:    [[TMP0:%.*]] = zext nneg i32 [[IDX0]] to i64
+; PREDICATED_EVL-NEXT:    [[ARRAY1IDX0:%.*]] = getelementptr inbounds nuw i8, ptr [[P]], i64 [[TMP0]]
+; PREDICATED_EVL-NEXT:    [[TMP1:%.*]] = load i8, ptr [[ARRAY1IDX0]], align 1
+; PREDICATED_EVL-NEXT:    [[TMP2:%.*]] = zext nneg i32 [[IDX1]] to i64
+; PREDICATED_EVL-NEXT:    [[ARRAY1IDX1:%.*]] = getelementptr inbounds nuw i8, ptr [[P]], i64 [[TMP2]]
+; PREDICATED_EVL-NEXT:    [[TMP3:%.*]] = load i8, ptr [[ARRAY1IDX1]], align 1
+; PREDICATED_EVL-NEXT:    [[TMP4:%.*]] = zext nneg i32 [[IDX2]] to i64
+; PREDICATED_EVL-NEXT:    [[ARRAY1IDX2:%.*]] = getelementptr inbounds nuw i8, ptr [[P]], i64 [[TMP4]]
+; PREDICATED_EVL-NEXT:    [[TMP5:%.*]] = load i8, ptr [[ARRAY1IDX2]], align 1
+; PREDICATED_EVL-NEXT:    [[TMP6:%.*]] = zext nneg i32 [[IDX3]] to i64
+; PREDICATED_EVL-NEXT:    [[ARRAY1IDX3:%.*]] = getelementptr inbounds nuw i8, ptr [[P]], i64 [[TMP6]]
+; PREDICATED_EVL-NEXT:    [[TMP7:%.*]] = load i8, ptr [[ARRAY1IDX3]], align 1
+; PREDICATED_EVL-NEXT:    [[SPEC_SELECT_I1:%.*]] = call i8 @llvm.smax.i8(i8 [[TMP1]], i8 [[TMP3]])
+; PREDICATED_EVL-NEXT:    [[SUB1:%.*]] = sub i8 0, [[SPEC_SELECT_I1]]
+; PREDICATED_EVL-NEXT:    [[SPEC_SELECT_I2:%.*]] = call i8 @llvm.smax.i8(i8 [[TMP5]], i8 [[TMP7]])
+; PREDICATED_EVL-NEXT:    [[SUB2:%.*]] = sub i8 0, [[SPEC_SELECT_I2]]
+; PREDICATED_EVL-NEXT:    [[TMP8:%.*]] = zext nneg i32 [[IDX0]] to i64
+; PREDICATED_EVL-NEXT:    [[ARRAY3IDX0:%.*]] = getelementptr inbounds nuw i8, ptr [[Q]], i64 [[TMP8]]
+; PREDICATED_EVL-NEXT:    store i8 [[SPEC_SELECT_I1]], ptr [[ARRAY3IDX0]], align 1
+; PREDICATED_EVL-NEXT:    [[TMP9:%.*]] = zext nneg i32 [[IDX1]] to i64
+; PREDICATED_EVL-NEXT:    [[ARRAY3IDX1:%.*]] = getelementptr inbounds nuw i8, ptr [[Q]], i64 [[TMP9]]
+; PREDICATED_EVL-NEXT:    store i8 [[SUB1]], ptr [[ARRAY3IDX1]], align 1
+; PREDICATED_EVL-NEXT:    [[TMP10:%.*]] = zext nneg i32 [[IDX2]] to i64
+; PREDICATED_EVL-NEXT:    [[ARRAY3IDX2:%.*]] = getelementptr inbounds nuw i8, ptr [[Q]], i64 [[TMP10]]
+; PREDICATED_EVL-NEXT:    store i8 [[SPEC_SELECT_I2]], ptr [[ARRAY3IDX2]], align 1
+; PREDICATED_EVL-NEXT:    [[TMP11:%.*]] = zext nneg i32 [[IDX3]] to i64
+; PREDICATED_EVL-NEXT:    [[ARRAY3IDX3:%.*]] = getelementptr inbounds nuw i8, ptr [[Q]], i64 [[TMP11]]
+; PREDICATED_EVL-NEXT:    store i8 [[SUB2]], ptr [[ARRAY3IDX3]], align 1
+; PREDICATED_EVL-NEXT:    br label [[FOR_INC]]
+; PREDICATED_EVL:       for.inc:
+; PREDICATED_EVL-NEXT:    [[INC]] = add nuw nsw i32 [[IX_024]], 1
+; PREDICATED_EVL-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[INC]], 1024
+; PREDICATED_EVL-NEXT:    br i1 [[EXITCOND]], label [[FOR_END:%.*]], label [[FOR_BODY]]
+; PREDICATED_EVL:       for.end:
+; PREDICATED_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
+}