; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 2
; RUN: opt -mtriple=aarch64-none-linux-gnu -S -passes=loop-vectorize,instcombine -mattr=+sve -force-vector-width=16 -force-vector-interleave=1 -scalable-vectorization=on -prefer-predicate-over-epilogue=scalar-epilogue %s 2>&1 | FileCheck %s -check-prefix=SCALAR_TAIL_FOLDING
; RUN: opt -mtriple=aarch64-none-linux-gnu -S -passes=loop-vectorize,instcombine -mattr=+sve -force-vector-width=16 -force-vector-interleave=1 -scalable-vectorization=on -prefer-predicate-over-epilogue=predicate-dont-vectorize %s 2>&1 | FileCheck %s -check-prefix=PREDICATED_TAIL_FOLDING
target datalayout = "e-m:e-i64:64-i128:128-n32:64-S128"
; Tests are taken from llvm/test/Transforms/LoopVectorize/interleaved-accesses-masked-group.ll
;
; Expected to contain interleave2/deinterleave2 instructions
;
; void masked_strided1(const unsigned char* restrict p,
; unsigned char* restrict q,
; unsigned char guard) {
; for(ix=0; ix < 1024; ++ix) {
; if (ix > guard) {
; char left = p[2*ix];
; char right = p[2*ix + 1];
; char max = max(left, right);
; q[2*ix] = max;
; q[2*ix+1] = 0 - max;
; }
; }
;}
define dso_local void @masked_strided1(ptr noalias nocapture readonly %p, ptr noalias nocapture %q, i8 zeroext %guard) local_unnamed_addr #0 {
; SCALAR_TAIL_FOLDING-LABEL: define dso_local void @masked_strided1
; SCALAR_TAIL_FOLDING-SAME: (ptr noalias nocapture readonly [[P:%.*]], ptr noalias nocapture [[Q:%.*]], i8 zeroext [[GUARD:%.*]]) local_unnamed_addr #[[ATTR0:[0-9]+]] {
; SCALAR_TAIL_FOLDING-NEXT: entry:
; SCALAR_TAIL_FOLDING-NEXT: [[CONV:%.*]] = zext i8 [[GUARD]] to i32
; SCALAR_TAIL_FOLDING-NEXT: [[TMP0:%.*]] = call i32 @llvm.vscale.i32()
; SCALAR_TAIL_FOLDING-NEXT: [[TMP1:%.*]] = shl i32 [[TMP0]], 4
; SCALAR_TAIL_FOLDING-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ugt i32 [[TMP1]], 1024
; SCALAR_TAIL_FOLDING-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; SCALAR_TAIL_FOLDING: vector.ph:
; SCALAR_TAIL_FOLDING-NEXT: [[TMP2:%.*]] = call i32 @llvm.vscale.i32()
; SCALAR_TAIL_FOLDING-NEXT: [[TMP3:%.*]] = shl i32 [[TMP2]], 4
; SCALAR_TAIL_FOLDING-NEXT: [[N_MOD_VF:%.*]] = urem i32 1024, [[TMP3]]
; SCALAR_TAIL_FOLDING-NEXT: [[N_VEC:%.*]] = sub nuw nsw i32 1024, [[N_MOD_VF]]
; SCALAR_TAIL_FOLDING-NEXT: [[TMP19:%.*]] = call i32 @llvm.vscale.i32()
; SCALAR_TAIL_FOLDING-NEXT: [[TMP20:%.*]] = shl i32 [[TMP19]], 4
; SCALAR_TAIL_FOLDING-NEXT: [[TMP4:%.*]] = call <vscale x 16 x i32> @llvm.stepvector.nxv16i32()
; SCALAR_TAIL_FOLDING-NEXT: [[DOTSPLATINSERT:%.*]] = insertelement <vscale x 16 x i32> poison, i32 [[TMP20]], i64 0
; SCALAR_TAIL_FOLDING-NEXT: [[DOTSPLAT:%.*]] = shufflevector <vscale x 16 x i32> [[DOTSPLATINSERT]], <vscale x 16 x i32> poison, <vscale x 16 x i32> zeroinitializer
; SCALAR_TAIL_FOLDING-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 16 x i32> poison, i32 [[CONV]], i64 0
; SCALAR_TAIL_FOLDING-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 16 x i32> [[BROADCAST_SPLATINSERT]], <vscale x 16 x i32> poison, <vscale x 16 x i32> zeroinitializer
; SCALAR_TAIL_FOLDING-NEXT: br label [[VECTOR_BODY:%.*]]
; SCALAR_TAIL_FOLDING: vector.body:
; SCALAR_TAIL_FOLDING-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; SCALAR_TAIL_FOLDING-NEXT: [[VEC_IND:%.*]] = phi <vscale x 16 x i32> [ [[TMP4]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
; SCALAR_TAIL_FOLDING-NEXT: [[TMP7:%.*]] = icmp ugt <vscale x 16 x i32> [[VEC_IND]], [[BROADCAST_SPLAT]]
; SCALAR_TAIL_FOLDING-NEXT: [[TMP8:%.*]] = shl i32 [[INDEX]], 1
; SCALAR_TAIL_FOLDING-NEXT: [[TMP9:%.*]] = sext i32 [[TMP8]] to i64
; SCALAR_TAIL_FOLDING-NEXT: [[TMP10:%.*]] = getelementptr i8, ptr [[P]], i64 [[TMP9]]
; SCALAR_TAIL_FOLDING-NEXT: [[INTERLEAVED_MASK:%.*]] = call <vscale x 32 x i1> @llvm.vector.interleave2.nxv32i1(<vscale x 16 x i1> [[TMP7]], <vscale x 16 x i1> [[TMP7]])
; SCALAR_TAIL_FOLDING-NEXT: [[WIDE_MASKED_VEC:%.*]] = call <vscale x 32 x i8> @llvm.masked.load.nxv32i8.p0(ptr [[TMP10]], i32 1, <vscale x 32 x i1> [[INTERLEAVED_MASK]], <vscale x 32 x i8> poison)
; SCALAR_TAIL_FOLDING-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_TAIL_FOLDING-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[STRIDED_VEC]], 0
; SCALAR_TAIL_FOLDING-NEXT: [[TMP12:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[STRIDED_VEC]], 1
; SCALAR_TAIL_FOLDING-NEXT: [[TMP13:%.*]] = or disjoint i32 [[TMP8]], 1
; SCALAR_TAIL_FOLDING-NEXT: [[TMP14:%.*]] = call <vscale x 16 x i8> @llvm.smax.nxv16i8(<vscale x 16 x i8> [[TMP11]], <vscale x 16 x i8> [[TMP12]])
; SCALAR_TAIL_FOLDING-NEXT: [[TMP15:%.*]] = sub <vscale x 16 x i8> zeroinitializer, [[TMP14]]
; SCALAR_TAIL_FOLDING-NEXT: [[TMP16:%.*]] = sext i32 [[TMP13]] to i64
; SCALAR_TAIL_FOLDING-NEXT: [[TMP17:%.*]] = getelementptr i8, ptr [[Q]], i64 [[TMP16]]
; SCALAR_TAIL_FOLDING-NEXT: [[TMP18:%.*]] = getelementptr i8, ptr [[TMP17]], i64 -1
; SCALAR_TAIL_FOLDING-NEXT: [[INTERLEAVED_VEC:%.*]] = call <vscale x 32 x i8> @llvm.vector.interleave2.nxv32i8(<vscale x 16 x i8> [[TMP14]], <vscale x 16 x i8> [[TMP15]])
; SCALAR_TAIL_FOLDING-NEXT: [[INTERLEAVED_MASK1:%.*]] = call <vscale x 32 x i1> @llvm.vector.interleave2.nxv32i1(<vscale x 16 x i1> [[TMP7]], <vscale x 16 x i1> [[TMP7]])
; SCALAR_TAIL_FOLDING-NEXT: call void @llvm.masked.store.nxv32i8.p0(<vscale x 32 x i8> [[INTERLEAVED_VEC]], ptr [[TMP18]], i32 1, <vscale x 32 x i1> [[INTERLEAVED_MASK1]])
; SCALAR_TAIL_FOLDING-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], [[TMP20]]
; SCALAR_TAIL_FOLDING-NEXT: [[VEC_IND_NEXT]] = add <vscale x 16 x i32> [[VEC_IND]], [[DOTSPLAT]]
; SCALAR_TAIL_FOLDING-NEXT: [[TMP21:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
; SCALAR_TAIL_FOLDING-NEXT: br i1 [[TMP21]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
; SCALAR_TAIL_FOLDING: middle.block:
; SCALAR_TAIL_FOLDING-NEXT: [[CMP_N:%.*]] = icmp eq i32 [[N_MOD_VF]], 0
; SCALAR_TAIL_FOLDING-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
; SCALAR_TAIL_FOLDING: scalar.ph:
; SCALAR_TAIL_FOLDING-NEXT: [[BC_RESUME_VAL:%.*]] = phi i32 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
; SCALAR_TAIL_FOLDING-NEXT: br label [[FOR_BODY:%.*]]
; SCALAR_TAIL_FOLDING: for.body:
; SCALAR_TAIL_FOLDING-NEXT: [[IX_024:%.*]] = phi i32 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INC:%.*]], [[FOR_INC:%.*]] ]
; SCALAR_TAIL_FOLDING-NEXT: [[CMP1:%.*]] = icmp ugt i32 [[IX_024]], [[CONV]]
; SCALAR_TAIL_FOLDING-NEXT: br i1 [[CMP1]], label [[IF_THEN:%.*]], label [[FOR_INC]]
; SCALAR_TAIL_FOLDING: if.then:
; SCALAR_TAIL_FOLDING-NEXT: [[MUL:%.*]] = shl nuw nsw i32 [[IX_024]], 1
; SCALAR_TAIL_FOLDING-NEXT: [[TMP22:%.*]] = zext nneg i32 [[MUL]] to i64
; SCALAR_TAIL_FOLDING-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i8, ptr [[P]], i64 [[TMP22]]
; SCALAR_TAIL_FOLDING-NEXT: [[TMP23:%.*]] = load i8, ptr [[ARRAYIDX]], align 1
; SCALAR_TAIL_FOLDING-NEXT: [[ADD:%.*]] = or disjoint i32 [[MUL]], 1
; SCALAR_TAIL_FOLDING-NEXT: [[TMP24:%.*]] = zext nneg i32 [[ADD]] to i64
; SCALAR_TAIL_FOLDING-NEXT: [[ARRAYIDX4:%.*]] = getelementptr inbounds i8, ptr [[P]], i64 [[TMP24]]
; SCALAR_TAIL_FOLDING-NEXT: [[TMP25:%.*]] = load i8, ptr [[ARRAYIDX4]], align 1
; SCALAR_TAIL_FOLDING-NEXT: [[SPEC_SELECT_I:%.*]] = call i8 @llvm.smax.i8(i8 [[TMP23]], i8 [[TMP25]])
; SCALAR_TAIL_FOLDING-NEXT: [[TMP26:%.*]] = zext nneg i32 [[MUL]] to i64
; SCALAR_TAIL_FOLDING-NEXT: [[ARRAYIDX6:%.*]] = getelementptr inbounds i8, ptr [[Q]], i64 [[TMP26]]
; SCALAR_TAIL_FOLDING-NEXT: store i8 [[SPEC_SELECT_I]], ptr [[ARRAYIDX6]], align 1
; SCALAR_TAIL_FOLDING-NEXT: [[SUB:%.*]] = sub i8 0, [[SPEC_SELECT_I]]
; SCALAR_TAIL_FOLDING-NEXT: [[TMP27:%.*]] = zext nneg i32 [[ADD]] to i64
; SCALAR_TAIL_FOLDING-NEXT: [[ARRAYIDX11:%.*]] = getelementptr inbounds i8, ptr [[Q]], i64 [[TMP27]]
; SCALAR_TAIL_FOLDING-NEXT: store i8 [[SUB]], ptr [[ARRAYIDX11]], align 1
; SCALAR_TAIL_FOLDING-NEXT: br label [[FOR_INC]]
; SCALAR_TAIL_FOLDING: for.inc:
; SCALAR_TAIL_FOLDING-NEXT: [[INC]] = add nuw nsw i32 [[IX_024]], 1
; SCALAR_TAIL_FOLDING-NEXT: [[EXITCOND:%.*]] = icmp eq i32 [[INC]], 1024
; SCALAR_TAIL_FOLDING-NEXT: br i1 [[EXITCOND]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP3:![0-9]+]]
; SCALAR_TAIL_FOLDING: for.end:
; SCALAR_TAIL_FOLDING-NEXT: ret void
;
; PREDICATED_TAIL_FOLDING-LABEL: define dso_local void @masked_strided1
; PREDICATED_TAIL_FOLDING-SAME: (ptr noalias nocapture readonly [[P:%.*]], ptr noalias nocapture [[Q:%.*]], i8 zeroext [[GUARD:%.*]]) local_unnamed_addr #[[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: [[TMP19:%.*]] = call i32 @llvm.vscale.i32()
; PREDICATED_TAIL_FOLDING-NEXT: [[TMP20:%.*]] = shl i32 [[TMP19]], 4
; PREDICATED_TAIL_FOLDING-NEXT: [[TMP0:%.*]] = call i32 @llvm.vscale.i32()
; PREDICATED_TAIL_FOLDING-NEXT: [[TMP1:%.*]] = shl i32 [[TMP0]], 4
; PREDICATED_TAIL_FOLDING-NEXT: [[TMP2:%.*]] = call i32 @llvm.usub.sat.i32(i32 1024, i32 [[TMP1]])
; PREDICATED_TAIL_FOLDING-NEXT: [[ACTIVE_LANE_MASK_ENTRY:%.*]] = call <vscale x 16 x i1> @llvm.get.active.lane.mask.nxv16i1.i32(i32 0, i32 1024)
; PREDICATED_TAIL_FOLDING-NEXT: [[TMP3:%.*]] = call <vscale x 16 x i32> @llvm.stepvector.nxv16i32()
; PREDICATED_TAIL_FOLDING-NEXT: [[DOTSPLATINSERT:%.*]] = insertelement <vscale x 16 x i32> poison, i32 [[TMP20]], i64 0
; PREDICATED_TAIL_FOLDING-NEXT: [[DOTSPLAT:%.*]] = shufflevector <vscale x 16 x i32> [[DOTSPLATINSERT]], <vscale x 16 x i32> poison, <vscale x 16 x i32> zeroinitializer
; PREDICATED_TAIL_FOLDING-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 16 x i32> poison, i32 [[CONV]], i64 0
; PREDICATED_TAIL_FOLDING-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 16 x i32> [[BROADCAST_SPLATINSERT]], <vscale x 16 x i32> poison, <vscale x 16 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: [[ACTIVE_LANE_MASK:%.*]] = phi <vscale x 16 x i1> [ [[ACTIVE_LANE_MASK_ENTRY]], [[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT:%.*]], [[VECTOR_BODY]] ]
; PREDICATED_TAIL_FOLDING-NEXT: [[VEC_IND:%.*]] = phi <vscale x 16 x i32> [ [[TMP3]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
; PREDICATED_TAIL_FOLDING-NEXT: [[TMP6:%.*]] = icmp ugt <vscale x 16 x i32> [[VEC_IND]], [[BROADCAST_SPLAT]]
; PREDICATED_TAIL_FOLDING-NEXT: [[TMP10:%.*]] = select <vscale x 16 x i1> [[ACTIVE_LANE_MASK]], <vscale x 16 x i1> [[TMP6]], <vscale x 16 x i1> zeroinitializer
; PREDICATED_TAIL_FOLDING-NEXT: [[TMP7:%.*]] = shl i32 [[INDEX]], 1
; PREDICATED_TAIL_FOLDING-NEXT: [[TMP8:%.*]] = sext i32 [[TMP7]] to i64
; PREDICATED_TAIL_FOLDING-NEXT: [[TMP9:%.*]] = getelementptr i8, ptr [[P]], i64 [[TMP8]]
; PREDICATED_TAIL_FOLDING-NEXT: [[INTERLEAVED_MASK:%.*]] = call <vscale x 32 x i1> @llvm.vector.interleave2.nxv32i1(<vscale x 16 x i1> [[TMP10]], <vscale x 16 x i1> [[TMP10]])
; PREDICATED_TAIL_FOLDING-NEXT: [[WIDE_MASKED_VEC:%.*]] = call <vscale x 32 x i8> @llvm.masked.load.nxv32i8.p0(ptr [[TMP9]], i32 1, <vscale x 32 x i1> [[INTERLEAVED_MASK]], <vscale x 32 x i8> poison)
; PREDICATED_TAIL_FOLDING-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_TAIL_FOLDING-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[STRIDED_VEC]], 0
; PREDICATED_TAIL_FOLDING-NEXT: [[TMP12:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[STRIDED_VEC]], 1
; PREDICATED_TAIL_FOLDING-NEXT: [[TMP13:%.*]] = or disjoint i32 [[TMP7]], 1
; PREDICATED_TAIL_FOLDING-NEXT: [[TMP14:%.*]] = call <vscale x 16 x i8> @llvm.smax.nxv16i8(<vscale x 16 x i8> [[TMP11]], <vscale x 16 x i8> [[TMP12]])
; PREDICATED_TAIL_FOLDING-NEXT: [[TMP15:%.*]] = sub <vscale x 16 x i8> zeroinitializer, [[TMP14]]
; PREDICATED_TAIL_FOLDING-NEXT: [[TMP16:%.*]] = sext i32 [[TMP13]] to i64
; PREDICATED_TAIL_FOLDING-NEXT: [[TMP17:%.*]] = getelementptr i8, ptr [[Q]], i64 [[TMP16]]
; PREDICATED_TAIL_FOLDING-NEXT: [[TMP18:%.*]] = getelementptr i8, ptr [[TMP17]], i64 -1
; PREDICATED_TAIL_FOLDING-NEXT: [[INTERLEAVED_VEC:%.*]] = call <vscale x 32 x i8> @llvm.vector.interleave2.nxv32i8(<vscale x 16 x i8> [[TMP14]], <vscale x 16 x i8> [[TMP15]])
; PREDICATED_TAIL_FOLDING-NEXT: [[INTERLEAVED_MASK1:%.*]] = call <vscale x 32 x i1> @llvm.vector.interleave2.nxv32i1(<vscale x 16 x i1> [[TMP10]], <vscale x 16 x i1> [[TMP10]])
; PREDICATED_TAIL_FOLDING-NEXT: call void @llvm.masked.store.nxv32i8.p0(<vscale x 32 x i8> [[INTERLEAVED_VEC]], ptr [[TMP18]], i32 1, <vscale x 32 x i1> [[INTERLEAVED_MASK1]])
; PREDICATED_TAIL_FOLDING-NEXT: [[INDEX_NEXT]] = add i32 [[INDEX]], [[TMP20]]
; PREDICATED_TAIL_FOLDING-NEXT: [[ACTIVE_LANE_MASK_NEXT]] = call <vscale x 16 x i1> @llvm.get.active.lane.mask.nxv16i1.i32(i32 [[INDEX]], i32 [[TMP2]])
; PREDICATED_TAIL_FOLDING-NEXT: [[VEC_IND_NEXT]] = add <vscale x 16 x i32> [[VEC_IND]], [[DOTSPLAT]]
; PREDICATED_TAIL_FOLDING-NEXT: [[TMP21:%.*]] = extractelement <vscale x 16 x i1> [[ACTIVE_LANE_MASK_NEXT]], i64 0
; PREDICATED_TAIL_FOLDING-NEXT: br i1 [[TMP21]], label [[VECTOR_BODY]], label [[MIDDLE_BLOCK:%.*]], !llvm.loop [[LOOP0:![0-9]+]]
; PREDICATED_TAIL_FOLDING: middle.block:
; PREDICATED_TAIL_FOLDING-NEXT: br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]]
; PREDICATED_TAIL_FOLDING: scalar.ph:
; PREDICATED_TAIL_FOLDING-NEXT: br label [[FOR_BODY:%.*]]
; PREDICATED_TAIL_FOLDING: for.body:
; PREDICATED_TAIL_FOLDING-NEXT: br i1 poison, label [[IF_THEN:%.*]], label [[FOR_INC:%.*]]
; PREDICATED_TAIL_FOLDING: if.then:
; PREDICATED_TAIL_FOLDING-NEXT: br label [[FOR_INC]]
; PREDICATED_TAIL_FOLDING: for.inc:
; PREDICATED_TAIL_FOLDING-NEXT: br i1 poison, label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP3:![0-9]+]]
; PREDICATED_TAIL_FOLDING: for.end:
; PREDICATED_TAIL_FOLDING-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
}
; Expected not to have interleave2/deinterleave2 instructions
;
; void masked_strided2(const unsigned char* restrict p,
; unsigned char* restrict q,
; unsigned char guard1,
; unsigned char guard2) {
; for(ix=0; ix < 1024; ++ix) {
; if (ix > guard1) {
; q[2*ix] = 1;
; }
; q[2*ix+1] = 2;
; }
;}
define dso_local void @masked_strided2(ptr noalias nocapture readnone %p, ptr noalias nocapture %q, i8 zeroext %guard) local_unnamed_addr #0 {
; SCALAR_TAIL_FOLDING-LABEL: define dso_local void @masked_strided2
; SCALAR_TAIL_FOLDING-SAME: (ptr noalias nocapture readnone [[P:%.*]], ptr noalias nocapture [[Q:%.*]], i8 zeroext [[GUARD:%.*]]) local_unnamed_addr #[[ATTR0]] {
; SCALAR_TAIL_FOLDING-NEXT: entry:
; SCALAR_TAIL_FOLDING-NEXT: [[CONV:%.*]] = zext i8 [[GUARD]] to i32
; SCALAR_TAIL_FOLDING-NEXT: [[TMP0:%.*]] = call i32 @llvm.vscale.i32()
; SCALAR_TAIL_FOLDING-NEXT: [[TMP1:%.*]] = shl i32 [[TMP0]], 4
; SCALAR_TAIL_FOLDING-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ugt i32 [[TMP1]], 1024
; SCALAR_TAIL_FOLDING-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; SCALAR_TAIL_FOLDING: vector.ph:
; SCALAR_TAIL_FOLDING-NEXT: [[TMP2:%.*]] = call i32 @llvm.vscale.i32()
; SCALAR_TAIL_FOLDING-NEXT: [[TMP3:%.*]] = shl i32 [[TMP2]], 4
; SCALAR_TAIL_FOLDING-NEXT: [[N_MOD_VF:%.*]] = urem i32 1024, [[TMP3]]
; SCALAR_TAIL_FOLDING-NEXT: [[N_VEC:%.*]] = sub nuw nsw i32 1024, [[N_MOD_VF]]
; SCALAR_TAIL_FOLDING-NEXT: [[TMP14:%.*]] = call i32 @llvm.vscale.i32()
; SCALAR_TAIL_FOLDING-NEXT: [[TMP15:%.*]] = shl i32 [[TMP14]], 4
; SCALAR_TAIL_FOLDING-NEXT: [[TMP4:%.*]] = call <vscale x 16 x i32> @llvm.stepvector.nxv16i32()
; SCALAR_TAIL_FOLDING-NEXT: [[DOTSPLATINSERT:%.*]] = insertelement <vscale x 16 x i32> poison, i32 [[TMP15]], i64 0
; SCALAR_TAIL_FOLDING-NEXT: [[DOTSPLAT:%.*]] = shufflevector <vscale x 16 x i32> [[DOTSPLATINSERT]], <vscale x 16 x i32> poison, <vscale x 16 x i32> zeroinitializer
; SCALAR_TAIL_FOLDING-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 16 x i32> poison, i32 [[CONV]], i64 0
; SCALAR_TAIL_FOLDING-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 16 x i32> [[BROADCAST_SPLATINSERT]], <vscale x 16 x i32> poison, <vscale x 16 x i32> zeroinitializer
; SCALAR_TAIL_FOLDING-NEXT: br label [[VECTOR_BODY:%.*]]
; SCALAR_TAIL_FOLDING: vector.body:
; SCALAR_TAIL_FOLDING-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; SCALAR_TAIL_FOLDING-NEXT: [[VEC_IND:%.*]] = phi <vscale x 16 x i32> [ [[TMP4]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
; SCALAR_TAIL_FOLDING-NEXT: [[TMP7:%.*]] = shl nuw nsw <vscale x 16 x i32> [[VEC_IND]], shufflevector (<vscale x 16 x i32> insertelement (<vscale x 16 x i32> poison, i32 1, i64 0), <vscale x 16 x i32> poison, <vscale x 16 x i32> zeroinitializer)
; SCALAR_TAIL_FOLDING-NEXT: [[TMP8:%.*]] = zext nneg <vscale x 16 x i32> [[TMP7]] to <vscale x 16 x i64>
; SCALAR_TAIL_FOLDING-NEXT: [[TMP9:%.*]] = getelementptr inbounds i8, ptr [[Q]], <vscale x 16 x i64> [[TMP8]]
; SCALAR_TAIL_FOLDING-NEXT: call void @llvm.masked.scatter.nxv16i8.nxv16p0(<vscale x 16 x i8> shufflevector (<vscale x 16 x i8> insertelement (<vscale x 16 x i8> poison, i8 1, i64 0), <vscale x 16 x i8> poison, <vscale x 16 x i32> zeroinitializer), <vscale x 16 x ptr> [[TMP9]], i32 1, <vscale x 16 x i1> shufflevector (<vscale x 16 x i1> insertelement (<vscale x 16 x i1> poison, i1 true, i64 0), <vscale x 16 x i1> poison, <vscale x 16 x i32> zeroinitializer))
; SCALAR_TAIL_FOLDING-NEXT: [[TMP10:%.*]] = icmp ugt <vscale x 16 x i32> [[VEC_IND]], [[BROADCAST_SPLAT]]
; SCALAR_TAIL_FOLDING-NEXT: [[TMP11:%.*]] = or disjoint <vscale x 16 x i32> [[TMP7]], shufflevector (<vscale x 16 x i32> insertelement (<vscale x 16 x i32> poison, i32 1, i64 0), <vscale x 16 x i32> poison, <vscale x 16 x i32> zeroinitializer)
; SCALAR_TAIL_FOLDING-NEXT: [[TMP12:%.*]] = zext nneg <vscale x 16 x i32> [[TMP11]] to <vscale x 16 x i64>
; SCALAR_TAIL_FOLDING-NEXT: [[TMP13:%.*]] = getelementptr inbounds i8, ptr [[Q]], <vscale x 16 x i64> [[TMP12]]
; SCALAR_TAIL_FOLDING-NEXT: call void @llvm.masked.scatter.nxv16i8.nxv16p0(<vscale x 16 x i8> shufflevector (<vscale x 16 x i8> insertelement (<vscale x 16 x i8> poison, i8 2, i64 0), <vscale x 16 x i8> poison, <vscale x 16 x i32> zeroinitializer), <vscale x 16 x ptr> [[TMP13]], i32 1, <vscale x 16 x i1> [[TMP10]])
; SCALAR_TAIL_FOLDING-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], [[TMP15]]
; SCALAR_TAIL_FOLDING-NEXT: [[VEC_IND_NEXT]] = add <vscale x 16 x i32> [[VEC_IND]], [[DOTSPLAT]]
; SCALAR_TAIL_FOLDING-NEXT: [[TMP16:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
; SCALAR_TAIL_FOLDING-NEXT: br i1 [[TMP16]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
; SCALAR_TAIL_FOLDING: middle.block:
; SCALAR_TAIL_FOLDING-NEXT: [[CMP_N:%.*]] = icmp eq i32 [[N_MOD_VF]], 0
; SCALAR_TAIL_FOLDING-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
; SCALAR_TAIL_FOLDING: scalar.ph:
; SCALAR_TAIL_FOLDING-NEXT: [[BC_RESUME_VAL:%.*]] = phi i32 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
; SCALAR_TAIL_FOLDING-NEXT: br label [[FOR_BODY:%.*]]
; SCALAR_TAIL_FOLDING: for.body:
; SCALAR_TAIL_FOLDING-NEXT: [[IX_012:%.*]] = phi i32 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INC:%.*]], [[FOR_INC:%.*]] ]
; SCALAR_TAIL_FOLDING-NEXT: [[MUL:%.*]] = shl nuw nsw i32 [[IX_012]], 1
; SCALAR_TAIL_FOLDING-NEXT: [[TMP17:%.*]] = zext nneg i32 [[MUL]] to i64
; SCALAR_TAIL_FOLDING-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i8, ptr [[Q]], i64 [[TMP17]]
; SCALAR_TAIL_FOLDING-NEXT: store i8 1, ptr [[ARRAYIDX]], align 1
; SCALAR_TAIL_FOLDING-NEXT: [[CMP1:%.*]] = icmp ugt i32 [[IX_012]], [[CONV]]
; SCALAR_TAIL_FOLDING-NEXT: br i1 [[CMP1]], label [[IF_THEN:%.*]], label [[FOR_INC]]
; SCALAR_TAIL_FOLDING: if.then:
; SCALAR_TAIL_FOLDING-NEXT: [[ADD:%.*]] = or disjoint i32 [[MUL]], 1
; SCALAR_TAIL_FOLDING-NEXT: [[TMP18:%.*]] = zext nneg i32 [[ADD]] to i64
; SCALAR_TAIL_FOLDING-NEXT: [[ARRAYIDX3:%.*]] = getelementptr inbounds i8, ptr [[Q]], i64 [[TMP18]]
; SCALAR_TAIL_FOLDING-NEXT: store i8 2, ptr [[ARRAYIDX3]], align 1
; SCALAR_TAIL_FOLDING-NEXT: br label [[FOR_INC]]
; SCALAR_TAIL_FOLDING: for.inc:
; SCALAR_TAIL_FOLDING-NEXT: [[INC]] = add nuw nsw i32 [[IX_012]], 1
; SCALAR_TAIL_FOLDING-NEXT: [[EXITCOND:%.*]] = icmp eq i32 [[INC]], 1024
; SCALAR_TAIL_FOLDING-NEXT: br i1 [[EXITCOND]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]]
; SCALAR_TAIL_FOLDING: for.end:
; SCALAR_TAIL_FOLDING-NEXT: ret void
;
; PREDICATED_TAIL_FOLDING-LABEL: define dso_local void @masked_strided2
; PREDICATED_TAIL_FOLDING-SAME: (ptr noalias nocapture readnone [[P:%.*]], ptr noalias nocapture [[Q:%.*]], i8 zeroext [[GUARD:%.*]]) local_unnamed_addr #[[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: [[TMP14:%.*]] = call i32 @llvm.vscale.i32()
; PREDICATED_TAIL_FOLDING-NEXT: [[TMP15:%.*]] = shl i32 [[TMP14]], 4
; PREDICATED_TAIL_FOLDING-NEXT: [[TMP0:%.*]] = call i32 @llvm.vscale.i32()
; PREDICATED_TAIL_FOLDING-NEXT: [[TMP1:%.*]] = shl i32 [[TMP0]], 4
; PREDICATED_TAIL_FOLDING-NEXT: [[TMP2:%.*]] = call i32 @llvm.usub.sat.i32(i32 1024, i32 [[TMP1]])
; PREDICATED_TAIL_FOLDING-NEXT: [[ACTIVE_LANE_MASK_ENTRY:%.*]] = call <vscale x 16 x i1> @llvm.get.active.lane.mask.nxv16i1.i32(i32 0, i32 1024)
; PREDICATED_TAIL_FOLDING-NEXT: [[TMP3:%.*]] = call <vscale x 16 x i32> @llvm.stepvector.nxv16i32()
; PREDICATED_TAIL_FOLDING-NEXT: [[DOTSPLATINSERT:%.*]] = insertelement <vscale x 16 x i32> poison, i32 [[TMP15]], i64 0
; PREDICATED_TAIL_FOLDING-NEXT: [[DOTSPLAT:%.*]] = shufflevector <vscale x 16 x i32> [[DOTSPLATINSERT]], <vscale x 16 x i32> poison, <vscale x 16 x i32> zeroinitializer
; PREDICATED_TAIL_FOLDING-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 16 x i32> poison, i32 [[CONV]], i64 0
; PREDICATED_TAIL_FOLDING-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 16 x i32> [[BROADCAST_SPLATINSERT]], <vscale x 16 x i32> poison, <vscale x 16 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: [[ACTIVE_LANE_MASK:%.*]] = phi <vscale x 16 x i1> [ [[ACTIVE_LANE_MASK_ENTRY]], [[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT:%.*]], [[VECTOR_BODY]] ]
; PREDICATED_TAIL_FOLDING-NEXT: [[VEC_IND:%.*]] = phi <vscale x 16 x i32> [ [[TMP3]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
; PREDICATED_TAIL_FOLDING-NEXT: [[TMP6:%.*]] = shl nuw nsw <vscale x 16 x i32> [[VEC_IND]], shufflevector (<vscale x 16 x i32> insertelement (<vscale x 16 x i32> poison, i32 1, i64 0), <vscale x 16 x i32> poison, <vscale x 16 x i32> zeroinitializer)
; PREDICATED_TAIL_FOLDING-NEXT: [[TMP7:%.*]] = zext nneg <vscale x 16 x i32> [[TMP6]] to <vscale x 16 x i64>
; PREDICATED_TAIL_FOLDING-NEXT: [[TMP8:%.*]] = getelementptr inbounds i8, ptr [[Q]], <vscale x 16 x i64> [[TMP7]]
; PREDICATED_TAIL_FOLDING-NEXT: call void @llvm.masked.scatter.nxv16i8.nxv16p0(<vscale x 16 x i8> shufflevector (<vscale x 16 x i8> insertelement (<vscale x 16 x i8> poison, i8 1, i64 0), <vscale x 16 x i8> poison, <vscale x 16 x i32> zeroinitializer), <vscale x 16 x ptr> [[TMP8]], i32 1, <vscale x 16 x i1> [[ACTIVE_LANE_MASK]])
; PREDICATED_TAIL_FOLDING-NEXT: [[TMP9:%.*]] = icmp ugt <vscale x 16 x i32> [[VEC_IND]], [[BROADCAST_SPLAT]]
; PREDICATED_TAIL_FOLDING-NEXT: [[TMP13:%.*]] = select <vscale x 16 x i1> [[ACTIVE_LANE_MASK]], <vscale x 16 x i1> [[TMP9]], <vscale x 16 x i1> zeroinitializer
; PREDICATED_TAIL_FOLDING-NEXT: [[TMP10:%.*]] = or disjoint <vscale x 16 x i32> [[TMP6]], shufflevector (<vscale x 16 x i32> insertelement (<vscale x 16 x i32> poison, i32 1, i64 0), <vscale x 16 x i32> poison, <vscale x 16 x i32> zeroinitializer)
; PREDICATED_TAIL_FOLDING-NEXT: [[TMP11:%.*]] = zext nneg <vscale x 16 x i32> [[TMP10]] to <vscale x 16 x i64>
; PREDICATED_TAIL_FOLDING-NEXT: [[TMP12:%.*]] = getelementptr inbounds i8, ptr [[Q]], <vscale x 16 x i64> [[TMP11]]
; PREDICATED_TAIL_FOLDING-NEXT: call void @llvm.masked.scatter.nxv16i8.nxv16p0(<vscale x 16 x i8> shufflevector (<vscale x 16 x i8> insertelement (<vscale x 16 x i8> poison, i8 2, i64 0), <vscale x 16 x i8> poison, <vscale x 16 x i32> zeroinitializer), <vscale x 16 x ptr> [[TMP12]], i32 1, <vscale x 16 x i1> [[TMP13]])
; PREDICATED_TAIL_FOLDING-NEXT: [[INDEX_NEXT]] = add i32 [[INDEX]], [[TMP15]]
; PREDICATED_TAIL_FOLDING-NEXT: [[ACTIVE_LANE_MASK_NEXT]] = call <vscale x 16 x i1> @llvm.get.active.lane.mask.nxv16i1.i32(i32 [[INDEX]], i32 [[TMP2]])
; PREDICATED_TAIL_FOLDING-NEXT: [[VEC_IND_NEXT]] = add <vscale x 16 x i32> [[VEC_IND]], [[DOTSPLAT]]
; PREDICATED_TAIL_FOLDING-NEXT: [[TMP16:%.*]] = extractelement <vscale x 16 x i1> [[ACTIVE_LANE_MASK_NEXT]], i64 0
; PREDICATED_TAIL_FOLDING-NEXT: br i1 [[TMP16]], label [[VECTOR_BODY]], label [[MIDDLE_BLOCK:%.*]], !llvm.loop [[LOOP4:![0-9]+]]
; PREDICATED_TAIL_FOLDING: middle.block:
; PREDICATED_TAIL_FOLDING-NEXT: br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]]
; PREDICATED_TAIL_FOLDING: scalar.ph:
; PREDICATED_TAIL_FOLDING-NEXT: br label [[FOR_BODY:%.*]]
; PREDICATED_TAIL_FOLDING: for.body:
; PREDICATED_TAIL_FOLDING-NEXT: br i1 poison, label [[IF_THEN:%.*]], label [[FOR_INC:%.*]]
; PREDICATED_TAIL_FOLDING: if.then:
; PREDICATED_TAIL_FOLDING-NEXT: br label [[FOR_INC]]
; PREDICATED_TAIL_FOLDING: for.inc:
; PREDICATED_TAIL_FOLDING-NEXT: br i1 poison, label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]]
; PREDICATED_TAIL_FOLDING: for.end:
; PREDICATED_TAIL_FOLDING-NEXT: ret void
;
entry:
%conv = zext i8 %guard to i32
br label %for.body
for.body:
%ix.012 = phi i32 [ 0, %entry ], [ %inc, %for.inc ]
%mul = shl nuw nsw i32 %ix.012, 1
%arrayidx = getelementptr inbounds i8, ptr %q, i32 %mul
store i8 1, ptr %arrayidx, align 1
%cmp1 = icmp ugt i32 %ix.012, %conv
br i1 %cmp1, label %if.then, label %for.inc
if.then:
%add = or disjoint i32 %mul, 1
%arrayidx3 = getelementptr inbounds i8, ptr %q, i32 %add
store i8 2, ptr %arrayidx3, align 1
br label %for.inc
for.inc:
%inc = add nuw nsw i32 %ix.012, 1
%exitcond = icmp eq i32 %inc, 1024
br i1 %exitcond, label %for.end, label %for.body
for.end:
ret void
}
; Expected not to have interleave2/deinterleave2 instructions
;
; void masked_strided3(const unsigned char* restrict p,
; unsigned char* restrict q,
; unsigned char guard1,
; unsigned char guard2) {
; for(ix=0; ix < 1024; ++ix) {
; if (ix > guard1) {
; q[2*ix] = 1;
; }
; if (ix > guard2) {
; q[2*ix+1] = 2;
; }
; }
;}
define dso_local void @masked_strided3(ptr noalias nocapture readnone %p, ptr noalias nocapture %q, i8 zeroext %guard1, i8 zeroext %guard2) local_unnamed_addr #0 {
; SCALAR_TAIL_FOLDING-LABEL: define dso_local void @masked_strided3
; SCALAR_TAIL_FOLDING-SAME: (ptr noalias nocapture readnone [[P:%.*]], ptr noalias nocapture [[Q:%.*]], i8 zeroext [[GUARD1:%.*]], i8 zeroext [[GUARD2:%.*]]) local_unnamed_addr #[[ATTR0]] {
; SCALAR_TAIL_FOLDING-NEXT: entry:
; SCALAR_TAIL_FOLDING-NEXT: [[CONV:%.*]] = zext i8 [[GUARD1]] to i32
; SCALAR_TAIL_FOLDING-NEXT: [[CONV3:%.*]] = zext i8 [[GUARD2]] to i32
; SCALAR_TAIL_FOLDING-NEXT: [[TMP0:%.*]] = call i32 @llvm.vscale.i32()
; SCALAR_TAIL_FOLDING-NEXT: [[TMP1:%.*]] = shl i32 [[TMP0]], 4
; SCALAR_TAIL_FOLDING-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ugt i32 [[TMP1]], 1024
; SCALAR_TAIL_FOLDING-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; SCALAR_TAIL_FOLDING: vector.ph:
; SCALAR_TAIL_FOLDING-NEXT: [[TMP2:%.*]] = call i32 @llvm.vscale.i32()
; SCALAR_TAIL_FOLDING-NEXT: [[TMP3:%.*]] = shl i32 [[TMP2]], 4
; SCALAR_TAIL_FOLDING-NEXT: [[N_MOD_VF:%.*]] = urem i32 1024, [[TMP3]]
; SCALAR_TAIL_FOLDING-NEXT: [[N_VEC:%.*]] = sub nuw nsw i32 1024, [[N_MOD_VF]]
; SCALAR_TAIL_FOLDING-NEXT: [[TMP15:%.*]] = call i32 @llvm.vscale.i32()
; SCALAR_TAIL_FOLDING-NEXT: [[TMP16:%.*]] = shl i32 [[TMP15]], 4
; SCALAR_TAIL_FOLDING-NEXT: [[TMP4:%.*]] = call <vscale x 16 x i32> @llvm.stepvector.nxv16i32()
; SCALAR_TAIL_FOLDING-NEXT: [[DOTSPLATINSERT:%.*]] = insertelement <vscale x 16 x i32> poison, i32 [[TMP16]], i64 0
; SCALAR_TAIL_FOLDING-NEXT: [[DOTSPLAT:%.*]] = shufflevector <vscale x 16 x i32> [[DOTSPLATINSERT]], <vscale x 16 x i32> poison, <vscale x 16 x i32> zeroinitializer
; SCALAR_TAIL_FOLDING-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 16 x i32> poison, i32 [[CONV]], i64 0
; SCALAR_TAIL_FOLDING-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 16 x i32> [[BROADCAST_SPLATINSERT]], <vscale x 16 x i32> poison, <vscale x 16 x i32> zeroinitializer
; SCALAR_TAIL_FOLDING-NEXT: [[BROADCAST_SPLATINSERT1:%.*]] = insertelement <vscale x 16 x i32> poison, i32 [[CONV3]], i64 0
; SCALAR_TAIL_FOLDING-NEXT: [[BROADCAST_SPLAT2:%.*]] = shufflevector <vscale x 16 x i32> [[BROADCAST_SPLATINSERT1]], <vscale x 16 x i32> poison, <vscale x 16 x i32> zeroinitializer
; SCALAR_TAIL_FOLDING-NEXT: br label [[VECTOR_BODY:%.*]]
; SCALAR_TAIL_FOLDING: vector.body:
; SCALAR_TAIL_FOLDING-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; SCALAR_TAIL_FOLDING-NEXT: [[VEC_IND:%.*]] = phi <vscale x 16 x i32> [ [[TMP4]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
; SCALAR_TAIL_FOLDING-NEXT: [[TMP7:%.*]] = shl nuw nsw <vscale x 16 x i32> [[VEC_IND]], shufflevector (<vscale x 16 x i32> insertelement (<vscale x 16 x i32> poison, i32 1, i64 0), <vscale x 16 x i32> poison, <vscale x 16 x i32> zeroinitializer)
; SCALAR_TAIL_FOLDING-NEXT: [[TMP8:%.*]] = icmp ugt <vscale x 16 x i32> [[VEC_IND]], [[BROADCAST_SPLAT]]
; SCALAR_TAIL_FOLDING-NEXT: [[TMP9:%.*]] = zext nneg <vscale x 16 x i32> [[TMP7]] to <vscale x 16 x i64>
; SCALAR_TAIL_FOLDING-NEXT: [[TMP10:%.*]] = getelementptr inbounds i8, ptr [[Q]], <vscale x 16 x i64> [[TMP9]]
; SCALAR_TAIL_FOLDING-NEXT: call void @llvm.masked.scatter.nxv16i8.nxv16p0(<vscale x 16 x i8> shufflevector (<vscale x 16 x i8> insertelement (<vscale x 16 x i8> poison, i8 1, i64 0), <vscale x 16 x i8> poison, <vscale x 16 x i32> zeroinitializer), <vscale x 16 x ptr> [[TMP10]], i32 1, <vscale x 16 x i1> [[TMP8]])
; SCALAR_TAIL_FOLDING-NEXT: [[TMP11:%.*]] = icmp ugt <vscale x 16 x i32> [[VEC_IND]], [[BROADCAST_SPLAT2]]
; SCALAR_TAIL_FOLDING-NEXT: [[TMP12:%.*]] = or disjoint <vscale x 16 x i32> [[TMP7]], shufflevector (<vscale x 16 x i32> insertelement (<vscale x 16 x i32> poison, i32 1, i64 0), <vscale x 16 x i32> poison, <vscale x 16 x i32> zeroinitializer)
; SCALAR_TAIL_FOLDING-NEXT: [[TMP13:%.*]] = zext nneg <vscale x 16 x i32> [[TMP12]] to <vscale x 16 x i64>
; SCALAR_TAIL_FOLDING-NEXT: [[TMP14:%.*]] = getelementptr inbounds i8, ptr [[Q]], <vscale x 16 x i64> [[TMP13]]
; SCALAR_TAIL_FOLDING-NEXT: call void @llvm.masked.scatter.nxv16i8.nxv16p0(<vscale x 16 x i8> shufflevector (<vscale x 16 x i8> insertelement (<vscale x 16 x i8> poison, i8 2, i64 0), <vscale x 16 x i8> poison, <vscale x 16 x i32> zeroinitializer), <vscale x 16 x ptr> [[TMP14]], i32 1, <vscale x 16 x i1> [[TMP11]])
; SCALAR_TAIL_FOLDING-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], [[TMP16]]
; SCALAR_TAIL_FOLDING-NEXT: [[VEC_IND_NEXT]] = add <vscale x 16 x i32> [[VEC_IND]], [[DOTSPLAT]]
; SCALAR_TAIL_FOLDING-NEXT: [[TMP17:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
; SCALAR_TAIL_FOLDING-NEXT: br i1 [[TMP17]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
; SCALAR_TAIL_FOLDING: middle.block:
; SCALAR_TAIL_FOLDING-NEXT: [[CMP_N:%.*]] = icmp eq i32 [[N_MOD_VF]], 0
; SCALAR_TAIL_FOLDING-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
; SCALAR_TAIL_FOLDING: scalar.ph:
; SCALAR_TAIL_FOLDING-NEXT: [[BC_RESUME_VAL:%.*]] = phi i32 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
; SCALAR_TAIL_FOLDING-NEXT: br label [[FOR_BODY:%.*]]
; SCALAR_TAIL_FOLDING: for.body:
; SCALAR_TAIL_FOLDING-NEXT: [[IX_018:%.*]] = phi i32 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INC:%.*]], [[FOR_INC:%.*]] ]
; SCALAR_TAIL_FOLDING-NEXT: [[MUL:%.*]] = shl nuw nsw i32 [[IX_018]], 1
; SCALAR_TAIL_FOLDING-NEXT: [[CMP1:%.*]] = icmp ugt i32 [[IX_018]], [[CONV]]
; SCALAR_TAIL_FOLDING-NEXT: br i1 [[CMP1]], label [[IF_THEN:%.*]], label [[IF_END:%.*]]
; SCALAR_TAIL_FOLDING: if.then:
; SCALAR_TAIL_FOLDING-NEXT: [[TMP18:%.*]] = zext nneg i32 [[MUL]] to i64
; SCALAR_TAIL_FOLDING-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i8, ptr [[Q]], i64 [[TMP18]]
; SCALAR_TAIL_FOLDING-NEXT: store i8 1, ptr [[ARRAYIDX]], align 1
; SCALAR_TAIL_FOLDING-NEXT: br label [[IF_END]]
; SCALAR_TAIL_FOLDING: if.end:
; SCALAR_TAIL_FOLDING-NEXT: [[CMP4:%.*]] = icmp ugt i32 [[IX_018]], [[CONV3]]
; SCALAR_TAIL_FOLDING-NEXT: br i1 [[CMP4]], label [[IF_THEN6:%.*]], label [[FOR_INC]]
; SCALAR_TAIL_FOLDING: if.then6:
; SCALAR_TAIL_FOLDING-NEXT: [[ADD:%.*]] = or disjoint i32 [[MUL]], 1
; SCALAR_TAIL_FOLDING-NEXT: [[TMP19:%.*]] = zext nneg i32 [[ADD]] to i64
; SCALAR_TAIL_FOLDING-NEXT: [[ARRAYIDX7:%.*]] = getelementptr inbounds i8, ptr [[Q]], i64 [[TMP19]]
; SCALAR_TAIL_FOLDING-NEXT: store i8 2, ptr [[ARRAYIDX7]], align 1
; SCALAR_TAIL_FOLDING-NEXT: br label [[FOR_INC]]
; SCALAR_TAIL_FOLDING: for.inc:
; SCALAR_TAIL_FOLDING-NEXT: [[INC]] = add nuw nsw i32 [[IX_018]], 1
; SCALAR_TAIL_FOLDING-NEXT: [[EXITCOND:%.*]] = icmp eq i32 [[INC]], 1024
; SCALAR_TAIL_FOLDING-NEXT: br i1 [[EXITCOND]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP7:![0-9]+]]
; SCALAR_TAIL_FOLDING: for.end:
; SCALAR_TAIL_FOLDING-NEXT: ret void
;
; PREDICATED_TAIL_FOLDING-LABEL: define dso_local void @masked_strided3
; PREDICATED_TAIL_FOLDING-SAME: (ptr noalias nocapture readnone [[P:%.*]], ptr noalias nocapture [[Q:%.*]], i8 zeroext [[GUARD1:%.*]], i8 zeroext [[GUARD2:%.*]]) local_unnamed_addr #[[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: [[CONV3:%.*]] = zext i8 [[GUARD2]] to i32
; PREDICATED_TAIL_FOLDING-NEXT: [[CONV:%.*]] = zext i8 [[GUARD1]] to i32
; PREDICATED_TAIL_FOLDING-NEXT: [[TMP16:%.*]] = call i32 @llvm.vscale.i32()
; PREDICATED_TAIL_FOLDING-NEXT: [[TMP17:%.*]] = shl i32 [[TMP16]], 4
; PREDICATED_TAIL_FOLDING-NEXT: [[TMP0:%.*]] = call i32 @llvm.vscale.i32()
; PREDICATED_TAIL_FOLDING-NEXT: [[TMP1:%.*]] = shl i32 [[TMP0]], 4
; PREDICATED_TAIL_FOLDING-NEXT: [[TMP2:%.*]] = call i32 @llvm.usub.sat.i32(i32 1024, i32 [[TMP1]])
; PREDICATED_TAIL_FOLDING-NEXT: [[ACTIVE_LANE_MASK_ENTRY:%.*]] = call <vscale x 16 x i1> @llvm.get.active.lane.mask.nxv16i1.i32(i32 0, i32 1024)
; PREDICATED_TAIL_FOLDING-NEXT: [[TMP3:%.*]] = call <vscale x 16 x i32> @llvm.stepvector.nxv16i32()
; PREDICATED_TAIL_FOLDING-NEXT: [[DOTSPLATINSERT:%.*]] = insertelement <vscale x 16 x i32> poison, i32 [[TMP17]], i64 0
; PREDICATED_TAIL_FOLDING-NEXT: [[DOTSPLAT:%.*]] = shufflevector <vscale x 16 x i32> [[DOTSPLATINSERT]], <vscale x 16 x i32> poison, <vscale x 16 x i32> zeroinitializer
; PREDICATED_TAIL_FOLDING-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 16 x i32> poison, i32 [[CONV]], i64 0
; PREDICATED_TAIL_FOLDING-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 16 x i32> [[BROADCAST_SPLATINSERT]], <vscale x 16 x i32> poison, <vscale x 16 x i32> zeroinitializer
; PREDICATED_TAIL_FOLDING-NEXT: [[BROADCAST_SPLATINSERT1:%.*]] = insertelement <vscale x 16 x i32> poison, i32 [[CONV3]], i64 0
; PREDICATED_TAIL_FOLDING-NEXT: [[BROADCAST_SPLAT2:%.*]] = shufflevector <vscale x 16 x i32> [[BROADCAST_SPLATINSERT1]], <vscale x 16 x i32> poison, <vscale x 16 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: [[ACTIVE_LANE_MASK:%.*]] = phi <vscale x 16 x i1> [ [[ACTIVE_LANE_MASK_ENTRY]], [[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT:%.*]], [[VECTOR_BODY]] ]
; PREDICATED_TAIL_FOLDING-NEXT: [[VEC_IND:%.*]] = phi <vscale x 16 x i32> [ [[TMP3]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
; PREDICATED_TAIL_FOLDING-NEXT: [[TMP6:%.*]] = shl nuw nsw <vscale x 16 x i32> [[VEC_IND]], shufflevector (<vscale x 16 x i32> insertelement (<vscale x 16 x i32> poison, i32 1, i64 0), <vscale x 16 x i32> poison, <vscale x 16 x i32> zeroinitializer)
; PREDICATED_TAIL_FOLDING-NEXT: [[TMP7:%.*]] = icmp ugt <vscale x 16 x i32> [[VEC_IND]], [[BROADCAST_SPLAT]]
; PREDICATED_TAIL_FOLDING-NEXT: [[TMP10:%.*]] = select <vscale x 16 x i1> [[ACTIVE_LANE_MASK]], <vscale x 16 x i1> [[TMP7]], <vscale x 16 x i1> zeroinitializer
; PREDICATED_TAIL_FOLDING-NEXT: [[TMP8:%.*]] = zext nneg <vscale x 16 x i32> [[TMP6]] to <vscale x 16 x i64>
; PREDICATED_TAIL_FOLDING-NEXT: [[TMP9:%.*]] = getelementptr inbounds i8, ptr [[Q]], <vscale x 16 x i64> [[TMP8]]
; PREDICATED_TAIL_FOLDING-NEXT: call void @llvm.masked.scatter.nxv16i8.nxv16p0(<vscale x 16 x i8> shufflevector (<vscale x 16 x i8> insertelement (<vscale x 16 x i8> poison, i8 1, i64 0), <vscale x 16 x i8> poison, <vscale x 16 x i32> zeroinitializer), <vscale x 16 x ptr> [[TMP9]], i32 1, <vscale x 16 x i1> [[TMP10]])
; PREDICATED_TAIL_FOLDING-NEXT: [[TMP11:%.*]] = icmp ugt <vscale x 16 x i32> [[VEC_IND]], [[BROADCAST_SPLAT2]]
; PREDICATED_TAIL_FOLDING-NEXT: [[TMP15:%.*]] = select <vscale x 16 x i1> [[ACTIVE_LANE_MASK]], <vscale x 16 x i1> [[TMP11]], <vscale x 16 x i1> zeroinitializer
; PREDICATED_TAIL_FOLDING-NEXT: [[TMP12:%.*]] = or disjoint <vscale x 16 x i32> [[TMP6]], shufflevector (<vscale x 16 x i32> insertelement (<vscale x 16 x i32> poison, i32 1, i64 0), <vscale x 16 x i32> poison, <vscale x 16 x i32> zeroinitializer)
; PREDICATED_TAIL_FOLDING-NEXT: [[TMP13:%.*]] = zext nneg <vscale x 16 x i32> [[TMP12]] to <vscale x 16 x i64>
; PREDICATED_TAIL_FOLDING-NEXT: [[TMP14:%.*]] = getelementptr inbounds i8, ptr [[Q]], <vscale x 16 x i64> [[TMP13]]
; PREDICATED_TAIL_FOLDING-NEXT: call void @llvm.masked.scatter.nxv16i8.nxv16p0(<vscale x 16 x i8> shufflevector (<vscale x 16 x i8> insertelement (<vscale x 16 x i8> poison, i8 2, i64 0), <vscale x 16 x i8> poison, <vscale x 16 x i32> zeroinitializer), <vscale x 16 x ptr> [[TMP14]], i32 1, <vscale x 16 x i1> [[TMP15]])
; PREDICATED_TAIL_FOLDING-NEXT: [[INDEX_NEXT]] = add i32 [[INDEX]], [[TMP17]]
; PREDICATED_TAIL_FOLDING-NEXT: [[ACTIVE_LANE_MASK_NEXT]] = call <vscale x 16 x i1> @llvm.get.active.lane.mask.nxv16i1.i32(i32 [[INDEX]], i32 [[TMP2]])
; PREDICATED_TAIL_FOLDING-NEXT: [[VEC_IND_NEXT]] = add <vscale x 16 x i32> [[VEC_IND]], [[DOTSPLAT]]
; PREDICATED_TAIL_FOLDING-NEXT: [[TMP18:%.*]] = extractelement <vscale x 16 x i1> [[ACTIVE_LANE_MASK_NEXT]], i64 0
; PREDICATED_TAIL_FOLDING-NEXT: br i1 [[TMP18]], label [[VECTOR_BODY]], label [[MIDDLE_BLOCK:%.*]], !llvm.loop [[LOOP6:![0-9]+]]
; PREDICATED_TAIL_FOLDING: middle.block:
; PREDICATED_TAIL_FOLDING-NEXT: br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]]
; PREDICATED_TAIL_FOLDING: scalar.ph:
; PREDICATED_TAIL_FOLDING-NEXT: br label [[FOR_BODY:%.*]]
; PREDICATED_TAIL_FOLDING: for.body:
; PREDICATED_TAIL_FOLDING-NEXT: br i1 poison, label [[IF_THEN:%.*]], label [[IF_END:%.*]]
; PREDICATED_TAIL_FOLDING: if.then:
; PREDICATED_TAIL_FOLDING-NEXT: br label [[IF_END]]
; PREDICATED_TAIL_FOLDING: if.end:
; PREDICATED_TAIL_FOLDING-NEXT: br i1 poison, label [[IF_THEN6:%.*]], label [[FOR_INC:%.*]]
; PREDICATED_TAIL_FOLDING: if.then6:
; PREDICATED_TAIL_FOLDING-NEXT: br label [[FOR_INC]]
; PREDICATED_TAIL_FOLDING: for.inc:
; PREDICATED_TAIL_FOLDING-NEXT: br i1 poison, label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP7:![0-9]+]]
; PREDICATED_TAIL_FOLDING: for.end:
; PREDICATED_TAIL_FOLDING-NEXT: ret void
;
entry:
%conv = zext i8 %guard1 to i32
%conv3 = zext i8 %guard2 to i32
br label %for.body
for.body:
%ix.018 = phi i32 [ 0, %entry ], [ %inc, %for.inc ]
%mul = shl nuw nsw i32 %ix.018, 1
%cmp1 = icmp ugt i32 %ix.018, %conv
br i1 %cmp1, label %if.then, label %if.end
if.then:
%arrayidx = getelementptr inbounds i8, ptr %q, i32 %mul
store i8 1, ptr %arrayidx, align 1
br label %if.end
if.end:
%cmp4 = icmp ugt i32 %ix.018, %conv3
br i1 %cmp4, label %if.then6, label %for.inc
if.then6:
%add = or disjoint i32 %mul, 1
%arrayidx7 = getelementptr inbounds i8, ptr %q, i32 %add
store i8 2, ptr %arrayidx7, align 1
br label %for.inc
for.inc:
%inc = add nuw nsw i32 %ix.018, 1
%exitcond = icmp eq i32 %inc, 1024
br i1 %exitcond, label %for.end, label %for.body
for.end:
ret void
}