| ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py |
| ; RUN: opt -S -force-vector-width=2 -force-vector-interleave=1 -passes=loop-vectorize,simplifycfg -verify-loop-info -simplifycfg-require-and-preserve-domtree=1 -force-widen-divrem-via-safe-divisor=0 < %s | FileCheck %s |
| ; RUN: opt -S -force-vector-width=1 -force-vector-interleave=2 -passes=loop-vectorize -verify-loop-info -force-widen-divrem-via-safe-divisor=0 < %s | FileCheck %s --check-prefix=UNROLL-NO-VF |
| |
| target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128" |
| |
| ; Test predication of non-void instructions, specifically (i) that these |
| ; instructions permit vectorization and (ii) the creation of an insertelement |
| ; and a Phi node. We check the full 2-element sequence for all predicate instructions. |
| define void @test(ptr nocapture %asd, ptr nocapture %aud, |
| ; CHECK-LABEL: @test( |
| ; CHECK-NEXT: entry: |
| ; CHECK-NEXT: [[SCEVGEP:%.*]] = getelementptr i8, ptr [[ASD:%.*]], i64 512 |
| ; CHECK-NEXT: [[SCEVGEP1:%.*]] = getelementptr i8, ptr [[AUD:%.*]], i64 512 |
| ; CHECK-NEXT: [[SCEVGEP2:%.*]] = getelementptr i8, ptr [[ASR:%.*]], i64 512 |
| ; CHECK-NEXT: [[SCEVGEP3:%.*]] = getelementptr i8, ptr [[AUR:%.*]], i64 512 |
| ; CHECK-NEXT: [[BOUND0:%.*]] = icmp ult ptr [[ASD]], [[SCEVGEP1]] |
| ; CHECK-NEXT: [[BOUND1:%.*]] = icmp ult ptr [[AUD]], [[SCEVGEP]] |
| ; CHECK-NEXT: [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]] |
| ; CHECK-NEXT: [[BOUND04:%.*]] = icmp ult ptr [[ASD]], [[SCEVGEP2]] |
| ; CHECK-NEXT: [[BOUND15:%.*]] = icmp ult ptr [[ASR]], [[SCEVGEP]] |
| ; CHECK-NEXT: [[FOUND_CONFLICT6:%.*]] = and i1 [[BOUND04]], [[BOUND15]] |
| ; CHECK-NEXT: [[CONFLICT_RDX:%.*]] = or i1 [[FOUND_CONFLICT]], [[FOUND_CONFLICT6]] |
| ; CHECK-NEXT: [[BOUND07:%.*]] = icmp ult ptr [[ASD]], [[SCEVGEP3]] |
| ; CHECK-NEXT: [[BOUND18:%.*]] = icmp ult ptr [[AUR]], [[SCEVGEP]] |
| ; CHECK-NEXT: [[FOUND_CONFLICT9:%.*]] = and i1 [[BOUND07]], [[BOUND18]] |
| ; CHECK-NEXT: [[CONFLICT_RDX10:%.*]] = or i1 [[CONFLICT_RDX]], [[FOUND_CONFLICT9]] |
| ; CHECK-NEXT: [[BOUND011:%.*]] = icmp ult ptr [[AUD]], [[SCEVGEP2]] |
| ; CHECK-NEXT: [[BOUND112:%.*]] = icmp ult ptr [[ASR]], [[SCEVGEP1]] |
| ; CHECK-NEXT: [[FOUND_CONFLICT13:%.*]] = and i1 [[BOUND011]], [[BOUND112]] |
| ; CHECK-NEXT: [[CONFLICT_RDX14:%.*]] = or i1 [[CONFLICT_RDX10]], [[FOUND_CONFLICT13]] |
| ; CHECK-NEXT: [[BOUND015:%.*]] = icmp ult ptr [[AUD]], [[SCEVGEP3]] |
| ; CHECK-NEXT: [[BOUND116:%.*]] = icmp ult ptr [[AUR]], [[SCEVGEP1]] |
| ; CHECK-NEXT: [[FOUND_CONFLICT17:%.*]] = and i1 [[BOUND015]], [[BOUND116]] |
| ; CHECK-NEXT: [[CONFLICT_RDX18:%.*]] = or i1 [[CONFLICT_RDX14]], [[FOUND_CONFLICT17]] |
| ; CHECK-NEXT: [[BOUND019:%.*]] = icmp ult ptr [[ASR]], [[SCEVGEP3]] |
| ; CHECK-NEXT: [[BOUND120:%.*]] = icmp ult ptr [[AUR]], [[SCEVGEP2]] |
| ; CHECK-NEXT: [[FOUND_CONFLICT21:%.*]] = and i1 [[BOUND019]], [[BOUND120]] |
| ; CHECK-NEXT: [[CONFLICT_RDX22:%.*]] = or i1 [[CONFLICT_RDX18]], [[FOUND_CONFLICT21]] |
| ; CHECK-NEXT: br i1 [[CONFLICT_RDX22]], label [[FOR_BODY:%.*]], label [[VECTOR_BODY:%.*]] |
| ; CHECK: vector.body: |
| ; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ [[INDEX_NEXT:%.*]], [[PRED_UREM_CONTINUE27:%.*]] ], [ 0, [[ENTRY:%.*]] ] |
| ; CHECK-NEXT: [[TMP0:%.*]] = add i64 [[INDEX]], 0 |
| ; CHECK-NEXT: [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[ASD]], i64 [[TMP0]] |
| ; CHECK-NEXT: [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[AUD]], i64 [[TMP0]] |
| ; CHECK-NEXT: [[TMP3:%.*]] = getelementptr inbounds i32, ptr [[ASR]], i64 [[TMP0]] |
| ; CHECK-NEXT: [[TMP4:%.*]] = getelementptr inbounds i32, ptr [[AUR]], i64 [[TMP0]] |
| ; CHECK-NEXT: [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0 |
| ; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP5]], align 4, !alias.scope [[META5:![0-9]+]], !noalias [[META8:![0-9]+]] |
| ; CHECK-NEXT: [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[TMP2]], i32 0 |
| ; CHECK-NEXT: [[WIDE_LOAD23:%.*]] = load <2 x i32>, ptr [[TMP6]], align 4, !alias.scope [[META12:![0-9]+]], !noalias [[META13:![0-9]+]] |
| ; CHECK-NEXT: [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[TMP3]], i32 0 |
| ; CHECK-NEXT: [[WIDE_LOAD24:%.*]] = load <2 x i32>, ptr [[TMP7]], align 4, !alias.scope [[META14:![0-9]+]], !noalias [[META15:![0-9]+]] |
| ; CHECK-NEXT: [[TMP8:%.*]] = getelementptr inbounds i32, ptr [[TMP4]], i32 0 |
| ; CHECK-NEXT: [[WIDE_LOAD25:%.*]] = load <2 x i32>, ptr [[TMP8]], align 4, !alias.scope [[META15]] |
| ; CHECK-NEXT: [[TMP9:%.*]] = add nsw <2 x i32> [[WIDE_LOAD]], splat (i32 23) |
| ; CHECK-NEXT: [[TMP10:%.*]] = add nsw <2 x i32> [[WIDE_LOAD23]], splat (i32 24) |
| ; CHECK-NEXT: [[TMP11:%.*]] = add nsw <2 x i32> [[WIDE_LOAD24]], splat (i32 25) |
| ; CHECK-NEXT: [[TMP12:%.*]] = add nsw <2 x i32> [[WIDE_LOAD25]], splat (i32 26) |
| ; CHECK-NEXT: [[TMP13:%.*]] = icmp slt <2 x i32> [[WIDE_LOAD]], splat (i32 100) |
| ; CHECK-NEXT: [[TMP14:%.*]] = extractelement <2 x i1> [[TMP13]], i32 0 |
| ; CHECK-NEXT: br i1 [[TMP14]], label [[PRED_UREM_IF:%.*]], label [[PRED_UREM_CONTINUE:%.*]] |
| ; CHECK: pred.urem.if: |
| ; CHECK-NEXT: [[TMP15:%.*]] = extractelement <2 x i32> [[TMP9]], i32 0 |
| ; CHECK-NEXT: [[TMP16:%.*]] = extractelement <2 x i32> [[WIDE_LOAD]], i32 0 |
| ; CHECK-NEXT: [[TMP17:%.*]] = sdiv i32 [[TMP15]], [[TMP16]] |
| ; CHECK-NEXT: [[TMP18:%.*]] = insertelement <2 x i32> poison, i32 [[TMP17]], i32 0 |
| ; CHECK-NEXT: [[TMP19:%.*]] = extractelement <2 x i32> [[TMP10]], i32 0 |
| ; CHECK-NEXT: [[TMP20:%.*]] = extractelement <2 x i32> [[WIDE_LOAD23]], i32 0 |
| ; CHECK-NEXT: [[TMP21:%.*]] = udiv i32 [[TMP19]], [[TMP20]] |
| ; CHECK-NEXT: [[TMP22:%.*]] = insertelement <2 x i32> poison, i32 [[TMP21]], i32 0 |
| ; CHECK-NEXT: [[TMP23:%.*]] = extractelement <2 x i32> [[TMP11]], i32 0 |
| ; CHECK-NEXT: [[TMP24:%.*]] = extractelement <2 x i32> [[WIDE_LOAD24]], i32 0 |
| ; CHECK-NEXT: [[TMP25:%.*]] = srem i32 [[TMP23]], [[TMP24]] |
| ; CHECK-NEXT: [[TMP26:%.*]] = insertelement <2 x i32> poison, i32 [[TMP25]], i32 0 |
| ; CHECK-NEXT: [[TMP27:%.*]] = extractelement <2 x i32> [[TMP12]], i32 0 |
| ; CHECK-NEXT: [[TMP28:%.*]] = extractelement <2 x i32> [[WIDE_LOAD25]], i32 0 |
| ; CHECK-NEXT: [[TMP29:%.*]] = urem i32 [[TMP27]], [[TMP28]] |
| ; CHECK-NEXT: [[TMP30:%.*]] = insertelement <2 x i32> poison, i32 [[TMP29]], i32 0 |
| ; CHECK-NEXT: br label [[PRED_UREM_CONTINUE]] |
| ; CHECK: pred.urem.continue: |
| ; CHECK-NEXT: [[TMP31:%.*]] = phi <2 x i32> [ poison, [[VECTOR_BODY]] ], [ [[TMP18]], [[PRED_UREM_IF]] ] |
| ; CHECK-NEXT: [[TMP32:%.*]] = phi <2 x i32> [ poison, [[VECTOR_BODY]] ], [ [[TMP22]], [[PRED_UREM_IF]] ] |
| ; CHECK-NEXT: [[TMP33:%.*]] = phi <2 x i32> [ poison, [[VECTOR_BODY]] ], [ [[TMP26]], [[PRED_UREM_IF]] ] |
| ; CHECK-NEXT: [[TMP34:%.*]] = phi <2 x i32> [ poison, [[VECTOR_BODY]] ], [ [[TMP30]], [[PRED_UREM_IF]] ] |
| ; CHECK-NEXT: [[TMP35:%.*]] = extractelement <2 x i1> [[TMP13]], i32 1 |
| ; CHECK-NEXT: br i1 [[TMP35]], label [[PRED_UREM_IF26:%.*]], label [[PRED_UREM_CONTINUE27]] |
| ; CHECK: pred.urem.if26: |
| ; CHECK-NEXT: [[TMP36:%.*]] = extractelement <2 x i32> [[TMP9]], i32 1 |
| ; CHECK-NEXT: [[TMP37:%.*]] = extractelement <2 x i32> [[WIDE_LOAD]], i32 1 |
| ; CHECK-NEXT: [[TMP38:%.*]] = sdiv i32 [[TMP36]], [[TMP37]] |
| ; CHECK-NEXT: [[TMP39:%.*]] = insertelement <2 x i32> [[TMP31]], i32 [[TMP38]], i32 1 |
| ; CHECK-NEXT: [[TMP40:%.*]] = extractelement <2 x i32> [[TMP10]], i32 1 |
| ; CHECK-NEXT: [[TMP41:%.*]] = extractelement <2 x i32> [[WIDE_LOAD23]], i32 1 |
| ; CHECK-NEXT: [[TMP42:%.*]] = udiv i32 [[TMP40]], [[TMP41]] |
| ; CHECK-NEXT: [[TMP43:%.*]] = insertelement <2 x i32> [[TMP32]], i32 [[TMP42]], i32 1 |
| ; CHECK-NEXT: [[TMP44:%.*]] = extractelement <2 x i32> [[TMP11]], i32 1 |
| ; CHECK-NEXT: [[TMP45:%.*]] = extractelement <2 x i32> [[WIDE_LOAD24]], i32 1 |
| ; CHECK-NEXT: [[TMP46:%.*]] = srem i32 [[TMP44]], [[TMP45]] |
| ; CHECK-NEXT: [[TMP47:%.*]] = insertelement <2 x i32> [[TMP33]], i32 [[TMP46]], i32 1 |
| ; CHECK-NEXT: [[TMP48:%.*]] = extractelement <2 x i32> [[TMP12]], i32 1 |
| ; CHECK-NEXT: [[TMP49:%.*]] = extractelement <2 x i32> [[WIDE_LOAD25]], i32 1 |
| ; CHECK-NEXT: [[TMP50:%.*]] = urem i32 [[TMP48]], [[TMP49]] |
| ; CHECK-NEXT: [[TMP51:%.*]] = insertelement <2 x i32> [[TMP34]], i32 [[TMP50]], i32 1 |
| ; CHECK-NEXT: br label [[PRED_UREM_CONTINUE27]] |
| ; CHECK: pred.urem.continue27: |
| ; CHECK-NEXT: [[TMP52:%.*]] = phi <2 x i32> [ [[TMP31]], [[PRED_UREM_CONTINUE]] ], [ [[TMP39]], [[PRED_UREM_IF26]] ] |
| ; CHECK-NEXT: [[TMP53:%.*]] = phi <2 x i32> [ [[TMP32]], [[PRED_UREM_CONTINUE]] ], [ [[TMP43]], [[PRED_UREM_IF26]] ] |
| ; CHECK-NEXT: [[TMP54:%.*]] = phi <2 x i32> [ [[TMP33]], [[PRED_UREM_CONTINUE]] ], [ [[TMP47]], [[PRED_UREM_IF26]] ] |
| ; CHECK-NEXT: [[TMP55:%.*]] = phi <2 x i32> [ [[TMP34]], [[PRED_UREM_CONTINUE]] ], [ [[TMP51]], [[PRED_UREM_IF26]] ] |
| ; CHECK-NEXT: [[TMP56:%.*]] = xor <2 x i1> [[TMP13]], splat (i1 true) |
| ; CHECK-NEXT: [[PREDPHI:%.*]] = select <2 x i1> [[TMP56]], <2 x i32> [[TMP9]], <2 x i32> [[TMP52]] |
| ; CHECK-NEXT: [[PREDPHI28:%.*]] = select <2 x i1> [[TMP56]], <2 x i32> [[TMP10]], <2 x i32> [[TMP53]] |
| ; CHECK-NEXT: [[PREDPHI29:%.*]] = select <2 x i1> [[TMP56]], <2 x i32> [[TMP11]], <2 x i32> [[TMP54]] |
| ; CHECK-NEXT: [[PREDPHI30:%.*]] = select <2 x i1> [[TMP56]], <2 x i32> [[TMP12]], <2 x i32> [[TMP55]] |
| ; CHECK-NEXT: [[TMP57:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0 |
| ; CHECK-NEXT: store <2 x i32> [[PREDPHI]], ptr [[TMP57]], align 4, !alias.scope [[META5]], !noalias [[META8]] |
| ; CHECK-NEXT: [[TMP58:%.*]] = getelementptr inbounds i32, ptr [[TMP2]], i32 0 |
| ; CHECK-NEXT: store <2 x i32> [[PREDPHI28]], ptr [[TMP58]], align 4, !alias.scope [[META12]], !noalias [[META13]] |
| ; CHECK-NEXT: [[TMP59:%.*]] = getelementptr inbounds i32, ptr [[TMP3]], i32 0 |
| ; CHECK-NEXT: store <2 x i32> [[PREDPHI29]], ptr [[TMP59]], align 4, !alias.scope [[META14]], !noalias [[META15]] |
| ; CHECK-NEXT: [[TMP60:%.*]] = getelementptr inbounds i32, ptr [[TMP4]], i32 0 |
| ; CHECK-NEXT: store <2 x i32> [[PREDPHI30]], ptr [[TMP60]], align 4, !alias.scope [[META15]] |
| ; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2 |
| ; CHECK-NEXT: [[TMP61:%.*]] = icmp eq i64 [[INDEX_NEXT]], 128 |
| ; CHECK-NEXT: br i1 [[TMP61]], label [[FOR_COND_CLEANUP:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP16:![0-9]+]] |
| ; CHECK: for.cond.cleanup: |
| ; CHECK-NEXT: ret void |
| ; CHECK: for.body: |
| ; CHECK-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ [[INDVARS_IV_NEXT:%.*]], [[IF_END:%.*]] ], [ 0, [[ENTRY]] ] |
| ; CHECK-NEXT: [[ISD:%.*]] = getelementptr inbounds i32, ptr [[ASD]], i64 [[INDVARS_IV]] |
| ; CHECK-NEXT: [[IUD:%.*]] = getelementptr inbounds i32, ptr [[AUD]], i64 [[INDVARS_IV]] |
| ; CHECK-NEXT: [[ISR:%.*]] = getelementptr inbounds i32, ptr [[ASR]], i64 [[INDVARS_IV]] |
| ; CHECK-NEXT: [[IUR:%.*]] = getelementptr inbounds i32, ptr [[AUR]], i64 [[INDVARS_IV]] |
| ; CHECK-NEXT: [[LSD:%.*]] = load i32, ptr [[ISD]], align 4 |
| ; CHECK-NEXT: [[LUD:%.*]] = load i32, ptr [[IUD]], align 4 |
| ; CHECK-NEXT: [[LSR:%.*]] = load i32, ptr [[ISR]], align 4 |
| ; CHECK-NEXT: [[LUR:%.*]] = load i32, ptr [[IUR]], align 4 |
| ; CHECK-NEXT: [[PSD:%.*]] = add nsw i32 [[LSD]], 23 |
| ; CHECK-NEXT: [[PUD:%.*]] = add nsw i32 [[LUD]], 24 |
| ; CHECK-NEXT: [[PSR:%.*]] = add nsw i32 [[LSR]], 25 |
| ; CHECK-NEXT: [[PUR:%.*]] = add nsw i32 [[LUR]], 26 |
| ; CHECK-NEXT: [[CMP1:%.*]] = icmp slt i32 [[LSD]], 100 |
| ; CHECK-NEXT: br i1 [[CMP1]], label [[IF_THEN:%.*]], label [[IF_END]] |
| ; CHECK: if.then: |
| ; CHECK-NEXT: [[RSD:%.*]] = sdiv i32 [[PSD]], [[LSD]] |
| ; CHECK-NEXT: [[RUD:%.*]] = udiv i32 [[PUD]], [[LUD]] |
| ; CHECK-NEXT: [[RSR:%.*]] = srem i32 [[PSR]], [[LSR]] |
| ; CHECK-NEXT: [[RUR:%.*]] = urem i32 [[PUR]], [[LUR]] |
| ; CHECK-NEXT: br label [[IF_END]] |
| ; CHECK: if.end: |
| ; CHECK-NEXT: [[YSD_0:%.*]] = phi i32 [ [[RSD]], [[IF_THEN]] ], [ [[PSD]], [[FOR_BODY]] ] |
| ; CHECK-NEXT: [[YUD_0:%.*]] = phi i32 [ [[RUD]], [[IF_THEN]] ], [ [[PUD]], [[FOR_BODY]] ] |
| ; CHECK-NEXT: [[YSR_0:%.*]] = phi i32 [ [[RSR]], [[IF_THEN]] ], [ [[PSR]], [[FOR_BODY]] ] |
| ; CHECK-NEXT: [[YUR_0:%.*]] = phi i32 [ [[RUR]], [[IF_THEN]] ], [ [[PUR]], [[FOR_BODY]] ] |
| ; CHECK-NEXT: store i32 [[YSD_0]], ptr [[ISD]], align 4 |
| ; CHECK-NEXT: store i32 [[YUD_0]], ptr [[IUD]], align 4 |
| ; CHECK-NEXT: store i32 [[YSR_0]], ptr [[ISR]], align 4 |
| ; CHECK-NEXT: store i32 [[YUR_0]], ptr [[IUR]], align 4 |
| ; CHECK-NEXT: [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1 |
| ; CHECK-NEXT: [[EXITCOND:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], 128 |
| ; CHECK-NEXT: br i1 [[EXITCOND]], label [[FOR_COND_CLEANUP]], label [[FOR_BODY]], !llvm.loop [[LOOP19:![0-9]+]] |
| ; |
| ; UNROLL-NO-VF-LABEL: @test( |
| ; UNROLL-NO-VF-NEXT: entry: |
| ; UNROLL-NO-VF-NEXT: br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_MEMCHECK:%.*]] |
| ; UNROLL-NO-VF: vector.memcheck: |
| ; UNROLL-NO-VF-NEXT: [[SCEVGEP:%.*]] = getelementptr i8, ptr [[ASD:%.*]], i64 512 |
| ; UNROLL-NO-VF-NEXT: [[SCEVGEP1:%.*]] = getelementptr i8, ptr [[AUD:%.*]], i64 512 |
| ; UNROLL-NO-VF-NEXT: [[SCEVGEP2:%.*]] = getelementptr i8, ptr [[ASR:%.*]], i64 512 |
| ; UNROLL-NO-VF-NEXT: [[SCEVGEP3:%.*]] = getelementptr i8, ptr [[AUR:%.*]], i64 512 |
| ; UNROLL-NO-VF-NEXT: [[BOUND0:%.*]] = icmp ult ptr [[ASD]], [[SCEVGEP1]] |
| ; UNROLL-NO-VF-NEXT: [[BOUND1:%.*]] = icmp ult ptr [[AUD]], [[SCEVGEP]] |
| ; UNROLL-NO-VF-NEXT: [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]] |
| ; UNROLL-NO-VF-NEXT: [[BOUND04:%.*]] = icmp ult ptr [[ASD]], [[SCEVGEP2]] |
| ; UNROLL-NO-VF-NEXT: [[BOUND15:%.*]] = icmp ult ptr [[ASR]], [[SCEVGEP]] |
| ; UNROLL-NO-VF-NEXT: [[FOUND_CONFLICT6:%.*]] = and i1 [[BOUND04]], [[BOUND15]] |
| ; UNROLL-NO-VF-NEXT: [[CONFLICT_RDX:%.*]] = or i1 [[FOUND_CONFLICT]], [[FOUND_CONFLICT6]] |
| ; UNROLL-NO-VF-NEXT: [[BOUND07:%.*]] = icmp ult ptr [[ASD]], [[SCEVGEP3]] |
| ; UNROLL-NO-VF-NEXT: [[BOUND18:%.*]] = icmp ult ptr [[AUR]], [[SCEVGEP]] |
| ; UNROLL-NO-VF-NEXT: [[FOUND_CONFLICT9:%.*]] = and i1 [[BOUND07]], [[BOUND18]] |
| ; UNROLL-NO-VF-NEXT: [[CONFLICT_RDX10:%.*]] = or i1 [[CONFLICT_RDX]], [[FOUND_CONFLICT9]] |
| ; UNROLL-NO-VF-NEXT: [[BOUND011:%.*]] = icmp ult ptr [[AUD]], [[SCEVGEP2]] |
| ; UNROLL-NO-VF-NEXT: [[BOUND112:%.*]] = icmp ult ptr [[ASR]], [[SCEVGEP1]] |
| ; UNROLL-NO-VF-NEXT: [[FOUND_CONFLICT13:%.*]] = and i1 [[BOUND011]], [[BOUND112]] |
| ; UNROLL-NO-VF-NEXT: [[CONFLICT_RDX14:%.*]] = or i1 [[CONFLICT_RDX10]], [[FOUND_CONFLICT13]] |
| ; UNROLL-NO-VF-NEXT: [[BOUND015:%.*]] = icmp ult ptr [[AUD]], [[SCEVGEP3]] |
| ; UNROLL-NO-VF-NEXT: [[BOUND116:%.*]] = icmp ult ptr [[AUR]], [[SCEVGEP1]] |
| ; UNROLL-NO-VF-NEXT: [[FOUND_CONFLICT17:%.*]] = and i1 [[BOUND015]], [[BOUND116]] |
| ; UNROLL-NO-VF-NEXT: [[CONFLICT_RDX18:%.*]] = or i1 [[CONFLICT_RDX14]], [[FOUND_CONFLICT17]] |
| ; UNROLL-NO-VF-NEXT: [[BOUND019:%.*]] = icmp ult ptr [[ASR]], [[SCEVGEP3]] |
| ; UNROLL-NO-VF-NEXT: [[BOUND120:%.*]] = icmp ult ptr [[AUR]], [[SCEVGEP2]] |
| ; UNROLL-NO-VF-NEXT: [[FOUND_CONFLICT21:%.*]] = and i1 [[BOUND019]], [[BOUND120]] |
| ; UNROLL-NO-VF-NEXT: [[CONFLICT_RDX22:%.*]] = or i1 [[CONFLICT_RDX18]], [[FOUND_CONFLICT21]] |
| ; UNROLL-NO-VF-NEXT: br i1 [[CONFLICT_RDX22]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]] |
| ; UNROLL-NO-VF: vector.ph: |
| ; UNROLL-NO-VF-NEXT: br label [[VECTOR_BODY:%.*]] |
| ; UNROLL-NO-VF: vector.body: |
| ; UNROLL-NO-VF-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_UREM_CONTINUE24:%.*]] ] |
| ; UNROLL-NO-VF-NEXT: [[TMP0:%.*]] = add i64 [[INDEX]], 0 |
| ; UNROLL-NO-VF-NEXT: [[TMP1:%.*]] = add i64 [[INDEX]], 1 |
| ; UNROLL-NO-VF-NEXT: [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[ASD]], i64 [[TMP0]] |
| ; UNROLL-NO-VF-NEXT: [[TMP3:%.*]] = getelementptr inbounds i32, ptr [[ASD]], i64 [[TMP1]] |
| ; UNROLL-NO-VF-NEXT: [[TMP4:%.*]] = getelementptr inbounds i32, ptr [[AUD]], i64 [[TMP0]] |
| ; UNROLL-NO-VF-NEXT: [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[AUD]], i64 [[TMP1]] |
| ; UNROLL-NO-VF-NEXT: [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[ASR]], i64 [[TMP0]] |
| ; UNROLL-NO-VF-NEXT: [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[ASR]], i64 [[TMP1]] |
| ; UNROLL-NO-VF-NEXT: [[TMP8:%.*]] = getelementptr inbounds i32, ptr [[AUR]], i64 [[TMP0]] |
| ; UNROLL-NO-VF-NEXT: [[TMP9:%.*]] = getelementptr inbounds i32, ptr [[AUR]], i64 [[TMP1]] |
| ; UNROLL-NO-VF-NEXT: [[TMP10:%.*]] = load i32, ptr [[TMP2]], align 4, !alias.scope [[META5:![0-9]+]], !noalias [[META8:![0-9]+]] |
| ; UNROLL-NO-VF-NEXT: [[TMP11:%.*]] = load i32, ptr [[TMP3]], align 4, !alias.scope [[META5]], !noalias [[META8]] |
| ; UNROLL-NO-VF-NEXT: [[TMP12:%.*]] = load i32, ptr [[TMP4]], align 4, !alias.scope [[META12:![0-9]+]], !noalias [[META13:![0-9]+]] |
| ; UNROLL-NO-VF-NEXT: [[TMP13:%.*]] = load i32, ptr [[TMP5]], align 4, !alias.scope [[META12]], !noalias [[META13]] |
| ; UNROLL-NO-VF-NEXT: [[TMP14:%.*]] = load i32, ptr [[TMP6]], align 4, !alias.scope [[META14:![0-9]+]], !noalias [[META15:![0-9]+]] |
| ; UNROLL-NO-VF-NEXT: [[TMP15:%.*]] = load i32, ptr [[TMP7]], align 4, !alias.scope [[META14]], !noalias [[META15]] |
| ; UNROLL-NO-VF-NEXT: [[TMP16:%.*]] = load i32, ptr [[TMP8]], align 4, !alias.scope [[META15]] |
| ; UNROLL-NO-VF-NEXT: [[TMP17:%.*]] = load i32, ptr [[TMP9]], align 4, !alias.scope [[META15]] |
| ; UNROLL-NO-VF-NEXT: [[TMP18:%.*]] = add nsw i32 [[TMP10]], 23 |
| ; UNROLL-NO-VF-NEXT: [[TMP19:%.*]] = add nsw i32 [[TMP11]], 23 |
| ; UNROLL-NO-VF-NEXT: [[TMP20:%.*]] = add nsw i32 [[TMP12]], 24 |
| ; UNROLL-NO-VF-NEXT: [[TMP21:%.*]] = add nsw i32 [[TMP13]], 24 |
| ; UNROLL-NO-VF-NEXT: [[TMP22:%.*]] = add nsw i32 [[TMP14]], 25 |
| ; UNROLL-NO-VF-NEXT: [[TMP23:%.*]] = add nsw i32 [[TMP15]], 25 |
| ; UNROLL-NO-VF-NEXT: [[TMP24:%.*]] = add nsw i32 [[TMP16]], 26 |
| ; UNROLL-NO-VF-NEXT: [[TMP25:%.*]] = add nsw i32 [[TMP17]], 26 |
| ; UNROLL-NO-VF-NEXT: [[TMP26:%.*]] = icmp slt i32 [[TMP10]], 100 |
| ; UNROLL-NO-VF-NEXT: [[TMP27:%.*]] = icmp slt i32 [[TMP11]], 100 |
| ; UNROLL-NO-VF-NEXT: br i1 [[TMP26]], label [[PRED_UREM_IF:%.*]], label [[PRED_UREM_CONTINUE:%.*]] |
| ; UNROLL-NO-VF: pred.urem.if: |
| ; UNROLL-NO-VF-NEXT: [[TMP28:%.*]] = sdiv i32 [[TMP18]], [[TMP10]] |
| ; UNROLL-NO-VF-NEXT: [[TMP29:%.*]] = udiv i32 [[TMP20]], [[TMP12]] |
| ; UNROLL-NO-VF-NEXT: [[TMP30:%.*]] = srem i32 [[TMP22]], [[TMP14]] |
| ; UNROLL-NO-VF-NEXT: [[TMP31:%.*]] = urem i32 [[TMP24]], [[TMP16]] |
| ; UNROLL-NO-VF-NEXT: br label [[PRED_UREM_CONTINUE]] |
| ; UNROLL-NO-VF: pred.urem.continue: |
| ; UNROLL-NO-VF-NEXT: [[TMP32:%.*]] = phi i32 [ poison, [[VECTOR_BODY]] ], [ [[TMP28]], [[PRED_UREM_IF]] ] |
| ; UNROLL-NO-VF-NEXT: [[TMP33:%.*]] = phi i32 [ poison, [[VECTOR_BODY]] ], [ [[TMP29]], [[PRED_UREM_IF]] ] |
| ; UNROLL-NO-VF-NEXT: [[TMP34:%.*]] = phi i32 [ poison, [[VECTOR_BODY]] ], [ [[TMP30]], [[PRED_UREM_IF]] ] |
| ; UNROLL-NO-VF-NEXT: [[TMP35:%.*]] = phi i32 [ poison, [[VECTOR_BODY]] ], [ [[TMP31]], [[PRED_UREM_IF]] ] |
| ; UNROLL-NO-VF-NEXT: br i1 [[TMP27]], label [[PRED_UREM_IF23:%.*]], label [[PRED_UREM_CONTINUE24]] |
| ; UNROLL-NO-VF: pred.urem.if23: |
| ; UNROLL-NO-VF-NEXT: [[TMP36:%.*]] = sdiv i32 [[TMP19]], [[TMP11]] |
| ; UNROLL-NO-VF-NEXT: [[TMP37:%.*]] = udiv i32 [[TMP21]], [[TMP13]] |
| ; UNROLL-NO-VF-NEXT: [[TMP38:%.*]] = srem i32 [[TMP23]], [[TMP15]] |
| ; UNROLL-NO-VF-NEXT: [[TMP39:%.*]] = urem i32 [[TMP25]], [[TMP17]] |
| ; UNROLL-NO-VF-NEXT: br label [[PRED_UREM_CONTINUE24]] |
| ; UNROLL-NO-VF: pred.urem.continue24: |
| ; UNROLL-NO-VF-NEXT: [[TMP40:%.*]] = phi i32 [ poison, [[PRED_UREM_CONTINUE]] ], [ [[TMP36]], [[PRED_UREM_IF23]] ] |
| ; UNROLL-NO-VF-NEXT: [[TMP41:%.*]] = phi i32 [ poison, [[PRED_UREM_CONTINUE]] ], [ [[TMP37]], [[PRED_UREM_IF23]] ] |
| ; UNROLL-NO-VF-NEXT: [[TMP42:%.*]] = phi i32 [ poison, [[PRED_UREM_CONTINUE]] ], [ [[TMP38]], [[PRED_UREM_IF23]] ] |
| ; UNROLL-NO-VF-NEXT: [[TMP43:%.*]] = phi i32 [ poison, [[PRED_UREM_CONTINUE]] ], [ [[TMP39]], [[PRED_UREM_IF23]] ] |
| ; UNROLL-NO-VF-NEXT: [[TMP44:%.*]] = xor i1 [[TMP26]], true |
| ; UNROLL-NO-VF-NEXT: [[TMP45:%.*]] = xor i1 [[TMP27]], true |
| ; UNROLL-NO-VF-NEXT: [[PREDPHI:%.*]] = select i1 [[TMP44]], i32 [[TMP18]], i32 [[TMP32]] |
| ; UNROLL-NO-VF-NEXT: [[PREDPHI25:%.*]] = select i1 [[TMP45]], i32 [[TMP19]], i32 [[TMP40]] |
| ; UNROLL-NO-VF-NEXT: [[PREDPHI26:%.*]] = select i1 [[TMP44]], i32 [[TMP20]], i32 [[TMP33]] |
| ; UNROLL-NO-VF-NEXT: [[PREDPHI27:%.*]] = select i1 [[TMP45]], i32 [[TMP21]], i32 [[TMP41]] |
| ; UNROLL-NO-VF-NEXT: [[PREDPHI28:%.*]] = select i1 [[TMP44]], i32 [[TMP22]], i32 [[TMP34]] |
| ; UNROLL-NO-VF-NEXT: [[PREDPHI29:%.*]] = select i1 [[TMP45]], i32 [[TMP23]], i32 [[TMP42]] |
| ; UNROLL-NO-VF-NEXT: [[PREDPHI30:%.*]] = select i1 [[TMP44]], i32 [[TMP24]], i32 [[TMP35]] |
| ; UNROLL-NO-VF-NEXT: [[PREDPHI31:%.*]] = select i1 [[TMP45]], i32 [[TMP25]], i32 [[TMP43]] |
| ; UNROLL-NO-VF-NEXT: store i32 [[PREDPHI]], ptr [[TMP2]], align 4, !alias.scope [[META5]], !noalias [[META8]] |
| ; UNROLL-NO-VF-NEXT: store i32 [[PREDPHI25]], ptr [[TMP3]], align 4, !alias.scope [[META5]], !noalias [[META8]] |
| ; UNROLL-NO-VF-NEXT: store i32 [[PREDPHI26]], ptr [[TMP4]], align 4, !alias.scope [[META12]], !noalias [[META13]] |
| ; UNROLL-NO-VF-NEXT: store i32 [[PREDPHI27]], ptr [[TMP5]], align 4, !alias.scope [[META12]], !noalias [[META13]] |
| ; UNROLL-NO-VF-NEXT: store i32 [[PREDPHI28]], ptr [[TMP6]], align 4, !alias.scope [[META14]], !noalias [[META15]] |
| ; UNROLL-NO-VF-NEXT: store i32 [[PREDPHI29]], ptr [[TMP7]], align 4, !alias.scope [[META14]], !noalias [[META15]] |
| ; UNROLL-NO-VF-NEXT: store i32 [[PREDPHI30]], ptr [[TMP8]], align 4, !alias.scope [[META15]] |
| ; UNROLL-NO-VF-NEXT: store i32 [[PREDPHI31]], ptr [[TMP9]], align 4, !alias.scope [[META15]] |
| ; UNROLL-NO-VF-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2 |
| ; UNROLL-NO-VF-NEXT: [[TMP46:%.*]] = icmp eq i64 [[INDEX_NEXT]], 128 |
| ; UNROLL-NO-VF-NEXT: br i1 [[TMP46]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP16:![0-9]+]] |
| ; UNROLL-NO-VF: middle.block: |
| ; UNROLL-NO-VF-NEXT: br i1 true, label [[FOR_COND_CLEANUP:%.*]], label [[SCALAR_PH]] |
| ; UNROLL-NO-VF: scalar.ph: |
| ; UNROLL-NO-VF-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ 128, [[MIDDLE_BLOCK]] ], [ 0, [[VECTOR_MEMCHECK]] ], [ 0, [[ENTRY:%.*]] ] |
| ; UNROLL-NO-VF-NEXT: br label [[FOR_BODY:%.*]] |
| ; UNROLL-NO-VF: for.cond.cleanup: |
| ; UNROLL-NO-VF-NEXT: ret void |
| ; UNROLL-NO-VF: for.body: |
| ; UNROLL-NO-VF-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[IF_END:%.*]] ] |
| ; UNROLL-NO-VF-NEXT: [[ISD:%.*]] = getelementptr inbounds i32, ptr [[ASD]], i64 [[INDVARS_IV]] |
| ; UNROLL-NO-VF-NEXT: [[IUD:%.*]] = getelementptr inbounds i32, ptr [[AUD]], i64 [[INDVARS_IV]] |
| ; UNROLL-NO-VF-NEXT: [[ISR:%.*]] = getelementptr inbounds i32, ptr [[ASR]], i64 [[INDVARS_IV]] |
| ; UNROLL-NO-VF-NEXT: [[IUR:%.*]] = getelementptr inbounds i32, ptr [[AUR]], i64 [[INDVARS_IV]] |
| ; UNROLL-NO-VF-NEXT: [[LSD:%.*]] = load i32, ptr [[ISD]], align 4 |
| ; UNROLL-NO-VF-NEXT: [[LUD:%.*]] = load i32, ptr [[IUD]], align 4 |
| ; UNROLL-NO-VF-NEXT: [[LSR:%.*]] = load i32, ptr [[ISR]], align 4 |
| ; UNROLL-NO-VF-NEXT: [[LUR:%.*]] = load i32, ptr [[IUR]], align 4 |
| ; UNROLL-NO-VF-NEXT: [[PSD:%.*]] = add nsw i32 [[LSD]], 23 |
| ; UNROLL-NO-VF-NEXT: [[PUD:%.*]] = add nsw i32 [[LUD]], 24 |
| ; UNROLL-NO-VF-NEXT: [[PSR:%.*]] = add nsw i32 [[LSR]], 25 |
| ; UNROLL-NO-VF-NEXT: [[PUR:%.*]] = add nsw i32 [[LUR]], 26 |
| ; UNROLL-NO-VF-NEXT: [[CMP1:%.*]] = icmp slt i32 [[LSD]], 100 |
| ; UNROLL-NO-VF-NEXT: br i1 [[CMP1]], label [[IF_THEN:%.*]], label [[IF_END]] |
| ; UNROLL-NO-VF: if.then: |
| ; UNROLL-NO-VF-NEXT: [[RSD:%.*]] = sdiv i32 [[PSD]], [[LSD]] |
| ; UNROLL-NO-VF-NEXT: [[RUD:%.*]] = udiv i32 [[PUD]], [[LUD]] |
| ; UNROLL-NO-VF-NEXT: [[RSR:%.*]] = srem i32 [[PSR]], [[LSR]] |
| ; UNROLL-NO-VF-NEXT: [[RUR:%.*]] = urem i32 [[PUR]], [[LUR]] |
| ; UNROLL-NO-VF-NEXT: br label [[IF_END]] |
| ; UNROLL-NO-VF: if.end: |
| ; UNROLL-NO-VF-NEXT: [[YSD_0:%.*]] = phi i32 [ [[RSD]], [[IF_THEN]] ], [ [[PSD]], [[FOR_BODY]] ] |
| ; UNROLL-NO-VF-NEXT: [[YUD_0:%.*]] = phi i32 [ [[RUD]], [[IF_THEN]] ], [ [[PUD]], [[FOR_BODY]] ] |
| ; UNROLL-NO-VF-NEXT: [[YSR_0:%.*]] = phi i32 [ [[RSR]], [[IF_THEN]] ], [ [[PSR]], [[FOR_BODY]] ] |
| ; UNROLL-NO-VF-NEXT: [[YUR_0:%.*]] = phi i32 [ [[RUR]], [[IF_THEN]] ], [ [[PUR]], [[FOR_BODY]] ] |
| ; UNROLL-NO-VF-NEXT: store i32 [[YSD_0]], ptr [[ISD]], align 4 |
| ; UNROLL-NO-VF-NEXT: store i32 [[YUD_0]], ptr [[IUD]], align 4 |
| ; UNROLL-NO-VF-NEXT: store i32 [[YSR_0]], ptr [[ISR]], align 4 |
| ; UNROLL-NO-VF-NEXT: store i32 [[YUR_0]], ptr [[IUR]], align 4 |
| ; UNROLL-NO-VF-NEXT: [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1 |
| ; UNROLL-NO-VF-NEXT: [[EXITCOND:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], 128 |
| ; UNROLL-NO-VF-NEXT: br i1 [[EXITCOND]], label [[FOR_COND_CLEANUP]], label [[FOR_BODY]], !llvm.loop [[LOOP19:![0-9]+]] |
| ; |
| ptr nocapture %asr, ptr nocapture %aur) { |
| entry: |
| br label %for.body |
| |
| for.cond.cleanup: ; preds = %if.end |
| ret void |
| |
| for.body: ; preds = %if.end, %entry |
| %indvars.iv = phi i64 [ 0, %entry ], [ %indvars.iv.next, %if.end ] |
| %isd = getelementptr inbounds i32, ptr %asd, i64 %indvars.iv |
| %iud = getelementptr inbounds i32, ptr %aud, i64 %indvars.iv |
| %isr = getelementptr inbounds i32, ptr %asr, i64 %indvars.iv |
| %iur = getelementptr inbounds i32, ptr %aur, i64 %indvars.iv |
| %lsd = load i32, ptr %isd, align 4 |
| %lud = load i32, ptr %iud, align 4 |
| %lsr = load i32, ptr %isr, align 4 |
| %lur = load i32, ptr %iur, align 4 |
| %psd = add nsw i32 %lsd, 23 |
| %pud = add nsw i32 %lud, 24 |
| %psr = add nsw i32 %lsr, 25 |
| %pur = add nsw i32 %lur, 26 |
| %cmp1 = icmp slt i32 %lsd, 100 |
| br i1 %cmp1, label %if.then, label %if.end |
| |
| if.then: ; preds = %for.body |
| %rsd = sdiv i32 %psd, %lsd |
| %rud = udiv i32 %pud, %lud |
| %rsr = srem i32 %psr, %lsr |
| %rur = urem i32 %pur, %lur |
| br label %if.end |
| |
| if.end: ; preds = %if.then, %for.body |
| %ysd.0 = phi i32 [ %rsd, %if.then ], [ %psd, %for.body ] |
| %yud.0 = phi i32 [ %rud, %if.then ], [ %pud, %for.body ] |
| %ysr.0 = phi i32 [ %rsr, %if.then ], [ %psr, %for.body ] |
| %yur.0 = phi i32 [ %rur, %if.then ], [ %pur, %for.body ] |
| store i32 %ysd.0, ptr %isd, align 4 |
| store i32 %yud.0, ptr %iud, align 4 |
| store i32 %ysr.0, ptr %isr, align 4 |
| store i32 %yur.0, ptr %iur, align 4 |
| %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1 |
| %exitcond = icmp eq i64 %indvars.iv.next, 128 |
| br i1 %exitcond, label %for.cond.cleanup, label %for.body |
| } |
| |
| define void @test_scalar2scalar(ptr nocapture %asd, ptr nocapture %bsd) { |
| ; CHECK-LABEL: @test_scalar2scalar( |
| ; CHECK-NEXT: entry: |
| ; CHECK-NEXT: [[SCEVGEP:%.*]] = getelementptr i8, ptr [[ASD:%.*]], i64 512 |
| ; CHECK-NEXT: [[SCEVGEP1:%.*]] = getelementptr i8, ptr [[BSD:%.*]], i64 512 |
| ; CHECK-NEXT: [[BOUND0:%.*]] = icmp ult ptr [[ASD]], [[SCEVGEP1]] |
| ; CHECK-NEXT: [[BOUND1:%.*]] = icmp ult ptr [[BSD]], [[SCEVGEP]] |
| ; CHECK-NEXT: [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]] |
| ; CHECK-NEXT: br i1 [[FOUND_CONFLICT]], label [[FOR_BODY:%.*]], label [[VECTOR_BODY:%.*]] |
| ; CHECK: vector.body: |
| ; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ [[INDEX_NEXT:%.*]], [[PRED_SDIV_CONTINUE4:%.*]] ], [ 0, [[ENTRY:%.*]] ] |
| ; CHECK-NEXT: [[TMP0:%.*]] = add i64 [[INDEX]], 0 |
| ; CHECK-NEXT: [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[ASD]], i64 [[TMP0]] |
| ; CHECK-NEXT: [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0 |
| ; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP2]], align 4, !alias.scope [[META20:![0-9]+]], !noalias [[META23:![0-9]+]] |
| ; CHECK-NEXT: [[TMP3:%.*]] = getelementptr inbounds i32, ptr [[BSD]], i64 [[TMP0]] |
| ; CHECK-NEXT: [[TMP4:%.*]] = getelementptr inbounds i32, ptr [[TMP3]], i32 0 |
| ; CHECK-NEXT: [[WIDE_LOAD2:%.*]] = load <2 x i32>, ptr [[TMP4]], align 4, !alias.scope [[META23]] |
| ; CHECK-NEXT: [[TMP5:%.*]] = add nsw <2 x i32> [[WIDE_LOAD]], splat (i32 23) |
| ; CHECK-NEXT: [[TMP6:%.*]] = icmp slt <2 x i32> [[WIDE_LOAD]], splat (i32 100) |
| ; CHECK-NEXT: [[TMP7:%.*]] = extractelement <2 x i1> [[TMP6]], i32 0 |
| ; CHECK-NEXT: br i1 [[TMP7]], label [[PRED_SDIV_IF:%.*]], label [[PRED_SDIV_CONTINUE:%.*]] |
| ; CHECK: pred.sdiv.if: |
| ; CHECK-NEXT: [[TMP8:%.*]] = extractelement <2 x i32> [[TMP5]], i32 0 |
| ; CHECK-NEXT: [[TMP9:%.*]] = extractelement <2 x i32> [[WIDE_LOAD]], i32 0 |
| ; CHECK-NEXT: [[TMP10:%.*]] = sdiv i32 [[TMP8]], [[TMP9]] |
| ; CHECK-NEXT: [[TMP11:%.*]] = extractelement <2 x i32> [[WIDE_LOAD2]], i32 0 |
| ; CHECK-NEXT: [[TMP12:%.*]] = sdiv i32 [[TMP11]], [[TMP10]] |
| ; CHECK-NEXT: [[TMP13:%.*]] = insertelement <2 x i32> poison, i32 [[TMP12]], i32 0 |
| ; CHECK-NEXT: br label [[PRED_SDIV_CONTINUE]] |
| ; CHECK: pred.sdiv.continue: |
| ; CHECK-NEXT: [[TMP15:%.*]] = phi <2 x i32> [ poison, [[VECTOR_BODY]] ], [ [[TMP13]], [[PRED_SDIV_IF]] ] |
| ; CHECK-NEXT: [[TMP16:%.*]] = extractelement <2 x i1> [[TMP6]], i32 1 |
| ; CHECK-NEXT: br i1 [[TMP16]], label [[PRED_SDIV_IF3:%.*]], label [[PRED_SDIV_CONTINUE4]] |
| ; CHECK: pred.sdiv.if3: |
| ; CHECK-NEXT: [[TMP17:%.*]] = extractelement <2 x i32> [[TMP5]], i32 1 |
| ; CHECK-NEXT: [[TMP18:%.*]] = extractelement <2 x i32> [[WIDE_LOAD]], i32 1 |
| ; CHECK-NEXT: [[TMP19:%.*]] = sdiv i32 [[TMP17]], [[TMP18]] |
| ; CHECK-NEXT: [[TMP20:%.*]] = extractelement <2 x i32> [[WIDE_LOAD2]], i32 1 |
| ; CHECK-NEXT: [[TMP21:%.*]] = sdiv i32 [[TMP20]], [[TMP19]] |
| ; CHECK-NEXT: [[TMP22:%.*]] = insertelement <2 x i32> [[TMP15]], i32 [[TMP21]], i32 1 |
| ; CHECK-NEXT: br label [[PRED_SDIV_CONTINUE4]] |
| ; CHECK: pred.sdiv.continue4: |
| ; CHECK-NEXT: [[TMP24:%.*]] = phi <2 x i32> [ [[TMP15]], [[PRED_SDIV_CONTINUE]] ], [ [[TMP22]], [[PRED_SDIV_IF3]] ] |
| ; CHECK-NEXT: [[TMP25:%.*]] = xor <2 x i1> [[TMP6]], splat (i1 true) |
| ; CHECK-NEXT: [[PREDPHI:%.*]] = select <2 x i1> [[TMP25]], <2 x i32> [[TMP5]], <2 x i32> [[TMP24]] |
| ; CHECK-NEXT: [[TMP26:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0 |
| ; CHECK-NEXT: store <2 x i32> [[PREDPHI]], ptr [[TMP26]], align 4, !alias.scope [[META20]], !noalias [[META23]] |
| ; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2 |
| ; CHECK-NEXT: [[TMP27:%.*]] = icmp eq i64 [[INDEX_NEXT]], 128 |
| ; CHECK-NEXT: br i1 [[TMP27]], label [[FOR_COND_CLEANUP:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP25:![0-9]+]] |
| ; CHECK: for.cond.cleanup: |
| ; CHECK-NEXT: ret void |
| ; CHECK: for.body: |
| ; CHECK-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ [[INDVARS_IV_NEXT:%.*]], [[IF_END:%.*]] ], [ 0, [[ENTRY]] ] |
| ; CHECK-NEXT: [[ISD:%.*]] = getelementptr inbounds i32, ptr [[ASD]], i64 [[INDVARS_IV]] |
| ; CHECK-NEXT: [[LSD:%.*]] = load i32, ptr [[ISD]], align 4 |
| ; CHECK-NEXT: [[ISD_B:%.*]] = getelementptr inbounds i32, ptr [[BSD]], i64 [[INDVARS_IV]] |
| ; CHECK-NEXT: [[LSD_B:%.*]] = load i32, ptr [[ISD_B]], align 4 |
| ; CHECK-NEXT: [[PSD:%.*]] = add nsw i32 [[LSD]], 23 |
| ; CHECK-NEXT: [[CMP1:%.*]] = icmp slt i32 [[LSD]], 100 |
| ; CHECK-NEXT: br i1 [[CMP1]], label [[IF_THEN:%.*]], label [[IF_END]] |
| ; CHECK: if.then: |
| ; CHECK-NEXT: [[SD1:%.*]] = sdiv i32 [[PSD]], [[LSD]] |
| ; CHECK-NEXT: [[RSD:%.*]] = sdiv i32 [[LSD_B]], [[SD1]] |
| ; CHECK-NEXT: br label [[IF_END]] |
| ; CHECK: if.end: |
| ; CHECK-NEXT: [[YSD_0:%.*]] = phi i32 [ [[RSD]], [[IF_THEN]] ], [ [[PSD]], [[FOR_BODY]] ] |
| ; CHECK-NEXT: store i32 [[YSD_0]], ptr [[ISD]], align 4 |
| ; CHECK-NEXT: [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1 |
| ; CHECK-NEXT: [[EXITCOND:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], 128 |
| ; CHECK-NEXT: br i1 [[EXITCOND]], label [[FOR_COND_CLEANUP]], label [[FOR_BODY]], !llvm.loop [[LOOP26:![0-9]+]] |
| ; |
| ; UNROLL-NO-VF-LABEL: @test_scalar2scalar( |
| ; UNROLL-NO-VF-NEXT: entry: |
| ; UNROLL-NO-VF-NEXT: br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_MEMCHECK:%.*]] |
| ; UNROLL-NO-VF: vector.memcheck: |
| ; UNROLL-NO-VF-NEXT: [[SCEVGEP:%.*]] = getelementptr i8, ptr [[ASD:%.*]], i64 512 |
| ; UNROLL-NO-VF-NEXT: [[SCEVGEP1:%.*]] = getelementptr i8, ptr [[BSD:%.*]], i64 512 |
| ; UNROLL-NO-VF-NEXT: [[BOUND0:%.*]] = icmp ult ptr [[ASD]], [[SCEVGEP1]] |
| ; UNROLL-NO-VF-NEXT: [[BOUND1:%.*]] = icmp ult ptr [[BSD]], [[SCEVGEP]] |
| ; UNROLL-NO-VF-NEXT: [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]] |
| ; UNROLL-NO-VF-NEXT: br i1 [[FOUND_CONFLICT]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]] |
| ; UNROLL-NO-VF: vector.ph: |
| ; UNROLL-NO-VF-NEXT: br label [[VECTOR_BODY:%.*]] |
| ; UNROLL-NO-VF: vector.body: |
| ; UNROLL-NO-VF-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_SDIV_CONTINUE3:%.*]] ] |
| ; UNROLL-NO-VF-NEXT: [[TMP0:%.*]] = add i64 [[INDEX]], 0 |
| ; UNROLL-NO-VF-NEXT: [[TMP1:%.*]] = add i64 [[INDEX]], 1 |
| ; UNROLL-NO-VF-NEXT: [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[ASD]], i64 [[TMP0]] |
| ; UNROLL-NO-VF-NEXT: [[TMP3:%.*]] = getelementptr inbounds i32, ptr [[ASD]], i64 [[TMP1]] |
| ; UNROLL-NO-VF-NEXT: [[TMP4:%.*]] = load i32, ptr [[TMP2]], align 4, !alias.scope [[META20:![0-9]+]], !noalias [[META23:![0-9]+]] |
| ; UNROLL-NO-VF-NEXT: [[TMP5:%.*]] = load i32, ptr [[TMP3]], align 4, !alias.scope [[META20]], !noalias [[META23]] |
| ; UNROLL-NO-VF-NEXT: [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[BSD]], i64 [[TMP0]] |
| ; UNROLL-NO-VF-NEXT: [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[BSD]], i64 [[TMP1]] |
| ; UNROLL-NO-VF-NEXT: [[TMP8:%.*]] = load i32, ptr [[TMP6]], align 4, !alias.scope [[META23]] |
| ; UNROLL-NO-VF-NEXT: [[TMP9:%.*]] = load i32, ptr [[TMP7]], align 4, !alias.scope [[META23]] |
| ; UNROLL-NO-VF-NEXT: [[TMP10:%.*]] = add nsw i32 [[TMP4]], 23 |
| ; UNROLL-NO-VF-NEXT: [[TMP11:%.*]] = add nsw i32 [[TMP5]], 23 |
| ; UNROLL-NO-VF-NEXT: [[TMP12:%.*]] = icmp slt i32 [[TMP4]], 100 |
| ; UNROLL-NO-VF-NEXT: [[TMP13:%.*]] = icmp slt i32 [[TMP5]], 100 |
| ; UNROLL-NO-VF-NEXT: br i1 [[TMP12]], label [[PRED_SDIV_IF:%.*]], label [[PRED_SDIV_CONTINUE:%.*]] |
| ; UNROLL-NO-VF: pred.sdiv.if: |
| ; UNROLL-NO-VF-NEXT: [[TMP14:%.*]] = sdiv i32 [[TMP10]], [[TMP4]] |
| ; UNROLL-NO-VF-NEXT: [[TMP15:%.*]] = sdiv i32 [[TMP8]], [[TMP14]] |
| ; UNROLL-NO-VF-NEXT: br label [[PRED_SDIV_CONTINUE]] |
| ; UNROLL-NO-VF: pred.sdiv.continue: |
| ; UNROLL-NO-VF-NEXT: [[TMP17:%.*]] = phi i32 [ poison, [[VECTOR_BODY]] ], [ [[TMP15]], [[PRED_SDIV_IF]] ] |
| ; UNROLL-NO-VF-NEXT: br i1 [[TMP13]], label [[PRED_SDIV_IF2:%.*]], label [[PRED_SDIV_CONTINUE3]] |
| ; UNROLL-NO-VF: pred.sdiv.if2: |
| ; UNROLL-NO-VF-NEXT: [[TMP18:%.*]] = sdiv i32 [[TMP11]], [[TMP5]] |
| ; UNROLL-NO-VF-NEXT: [[TMP19:%.*]] = sdiv i32 [[TMP9]], [[TMP18]] |
| ; UNROLL-NO-VF-NEXT: br label [[PRED_SDIV_CONTINUE3]] |
| ; UNROLL-NO-VF: pred.sdiv.continue3: |
| ; UNROLL-NO-VF-NEXT: [[TMP21:%.*]] = phi i32 [ poison, [[PRED_SDIV_CONTINUE]] ], [ [[TMP19]], [[PRED_SDIV_IF2]] ] |
| ; UNROLL-NO-VF-NEXT: [[TMP22:%.*]] = xor i1 [[TMP12]], true |
| ; UNROLL-NO-VF-NEXT: [[TMP23:%.*]] = xor i1 [[TMP13]], true |
| ; UNROLL-NO-VF-NEXT: [[PREDPHI:%.*]] = select i1 [[TMP22]], i32 [[TMP10]], i32 [[TMP17]] |
| ; UNROLL-NO-VF-NEXT: [[PREDPHI4:%.*]] = select i1 [[TMP23]], i32 [[TMP11]], i32 [[TMP21]] |
| ; UNROLL-NO-VF-NEXT: store i32 [[PREDPHI]], ptr [[TMP2]], align 4, !alias.scope [[META20]], !noalias [[META23]] |
| ; UNROLL-NO-VF-NEXT: store i32 [[PREDPHI4]], ptr [[TMP3]], align 4, !alias.scope [[META20]], !noalias [[META23]] |
| ; UNROLL-NO-VF-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2 |
| ; UNROLL-NO-VF-NEXT: [[TMP24:%.*]] = icmp eq i64 [[INDEX_NEXT]], 128 |
| ; UNROLL-NO-VF-NEXT: br i1 [[TMP24]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP25:![0-9]+]] |
| ; UNROLL-NO-VF: middle.block: |
| ; UNROLL-NO-VF-NEXT: br i1 true, label [[FOR_COND_CLEANUP:%.*]], label [[SCALAR_PH]] |
| ; UNROLL-NO-VF: scalar.ph: |
| ; UNROLL-NO-VF-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ 128, [[MIDDLE_BLOCK]] ], [ 0, [[VECTOR_MEMCHECK]] ], [ 0, [[ENTRY:%.*]] ] |
| ; UNROLL-NO-VF-NEXT: br label [[FOR_BODY:%.*]] |
| ; UNROLL-NO-VF: for.cond.cleanup: |
| ; UNROLL-NO-VF-NEXT: ret void |
| ; UNROLL-NO-VF: for.body: |
| ; UNROLL-NO-VF-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[IF_END:%.*]] ] |
| ; UNROLL-NO-VF-NEXT: [[ISD:%.*]] = getelementptr inbounds i32, ptr [[ASD]], i64 [[INDVARS_IV]] |
| ; UNROLL-NO-VF-NEXT: [[LSD:%.*]] = load i32, ptr [[ISD]], align 4 |
| ; UNROLL-NO-VF-NEXT: [[ISD_B:%.*]] = getelementptr inbounds i32, ptr [[BSD]], i64 [[INDVARS_IV]] |
| ; UNROLL-NO-VF-NEXT: [[LSD_B:%.*]] = load i32, ptr [[ISD_B]], align 4 |
| ; UNROLL-NO-VF-NEXT: [[PSD:%.*]] = add nsw i32 [[LSD]], 23 |
| ; UNROLL-NO-VF-NEXT: [[CMP1:%.*]] = icmp slt i32 [[LSD]], 100 |
| ; UNROLL-NO-VF-NEXT: br i1 [[CMP1]], label [[IF_THEN:%.*]], label [[IF_END]] |
| ; UNROLL-NO-VF: if.then: |
| ; UNROLL-NO-VF-NEXT: [[SD1:%.*]] = sdiv i32 [[PSD]], [[LSD]] |
| ; UNROLL-NO-VF-NEXT: [[RSD:%.*]] = sdiv i32 [[LSD_B]], [[SD1]] |
| ; UNROLL-NO-VF-NEXT: br label [[IF_END]] |
| ; UNROLL-NO-VF: if.end: |
| ; UNROLL-NO-VF-NEXT: [[YSD_0:%.*]] = phi i32 [ [[RSD]], [[IF_THEN]] ], [ [[PSD]], [[FOR_BODY]] ] |
| ; UNROLL-NO-VF-NEXT: store i32 [[YSD_0]], ptr [[ISD]], align 4 |
| ; UNROLL-NO-VF-NEXT: [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1 |
| ; UNROLL-NO-VF-NEXT: [[EXITCOND:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], 128 |
| ; UNROLL-NO-VF-NEXT: br i1 [[EXITCOND]], label [[FOR_COND_CLEANUP]], label [[FOR_BODY]], !llvm.loop [[LOOP26:![0-9]+]] |
| ; |
| entry: |
| br label %for.body |
| |
| for.cond.cleanup: ; preds = %if.end |
| ret void |
| |
| |
| for.body: ; preds = %if.end, %entry |
| %indvars.iv = phi i64 [ 0, %entry ], [ %indvars.iv.next, %if.end ] |
| %isd = getelementptr inbounds i32, ptr %asd, i64 %indvars.iv |
| %lsd = load i32, ptr %isd, align 4 |
| %isd.b = getelementptr inbounds i32, ptr %bsd, i64 %indvars.iv |
| %lsd.b = load i32, ptr %isd.b, align 4 |
| %psd = add nsw i32 %lsd, 23 |
| %cmp1 = icmp slt i32 %lsd, 100 |
| br i1 %cmp1, label %if.then, label %if.end |
| |
| if.then: ; preds = %for.body |
| %sd1 = sdiv i32 %psd, %lsd |
| %rsd = sdiv i32 %lsd.b, %sd1 |
| br label %if.end |
| |
| if.end: ; preds = %if.then, %for.body |
| %ysd.0 = phi i32 [ %rsd, %if.then ], [ %psd, %for.body ] |
| store i32 %ysd.0, ptr %isd, align 4 |
| %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1 |
| %exitcond = icmp eq i64 %indvars.iv.next, 128 |
| br i1 %exitcond, label %for.cond.cleanup, label %for.body |
| } |
| |
| define void @pr30172(ptr nocapture %asd, ptr nocapture %bsd) !dbg !5 {; |
| ; CHECK-LABEL: @pr30172( |
| ; CHECK-NEXT: entry: |
| ; CHECK-NEXT: [[SCEVGEP:%.*]] = getelementptr i8, ptr [[ASD:%.*]], i64 512 |
| ; CHECK-NEXT: [[SCEVGEP1:%.*]] = getelementptr i8, ptr [[BSD:%.*]], i64 512 |
| ; CHECK-NEXT: [[BOUND0:%.*]] = icmp ult ptr [[ASD]], [[SCEVGEP1]] |
| ; CHECK-NEXT: [[BOUND1:%.*]] = icmp ult ptr [[BSD]], [[SCEVGEP]] |
| ; CHECK-NEXT: [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]] |
| ; CHECK-NEXT: br i1 [[FOUND_CONFLICT]], label [[FOR_BODY:%.*]], label [[VECTOR_BODY:%.*]] |
| ; CHECK: vector.body: |
| ; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ [[INDEX_NEXT:%.*]], [[PRED_SDIV_CONTINUE4:%.*]] ], [ 0, [[ENTRY:%.*]] ] |
| ; CHECK-NEXT: [[TMP0:%.*]] = add i64 [[INDEX]], 0 |
| ; CHECK-NEXT: [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[ASD]], i64 [[TMP0]] |
| ; CHECK-NEXT: [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0 |
| ; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP2]], align 4, !alias.scope [[META29:![0-9]+]], !noalias [[META32:![0-9]+]] |
| ; CHECK-NEXT: [[TMP3:%.*]] = getelementptr inbounds i32, ptr [[BSD]], i64 [[TMP0]] |
| ; CHECK-NEXT: [[TMP4:%.*]] = getelementptr inbounds i32, ptr [[TMP3]], i32 0 |
| ; CHECK-NEXT: [[WIDE_LOAD2:%.*]] = load <2 x i32>, ptr [[TMP4]], align 4, !alias.scope [[META32]] |
| ; CHECK-NEXT: [[TMP5:%.*]] = add nsw <2 x i32> [[WIDE_LOAD]], splat (i32 23) |
| ; CHECK-NEXT: [[TMP6:%.*]] = icmp slt <2 x i32> [[WIDE_LOAD]], splat (i32 100) |
| ; CHECK-NEXT: [[TMP7:%.*]] = xor <2 x i1> [[TMP6]], splat (i1 true), !dbg [[DBG34:![0-9]+]] |
| ; CHECK-NEXT: [[TMP8:%.*]] = icmp sge <2 x i32> [[WIDE_LOAD]], splat (i32 200) |
| ; CHECK-NEXT: [[TMP9:%.*]] = select <2 x i1> [[TMP7]], <2 x i1> [[TMP8]], <2 x i1> zeroinitializer, !dbg [[DBG35:![0-9]+]] |
| ; CHECK-NEXT: [[TMP10:%.*]] = or <2 x i1> [[TMP9]], [[TMP6]] |
| ; CHECK-NEXT: [[TMP11:%.*]] = extractelement <2 x i1> [[TMP10]], i32 0 |
| ; CHECK-NEXT: br i1 [[TMP11]], label [[PRED_SDIV_IF:%.*]], label [[PRED_SDIV_CONTINUE:%.*]] |
| ; CHECK: pred.sdiv.if: |
| ; CHECK-NEXT: [[TMP12:%.*]] = extractelement <2 x i32> [[TMP5]], i32 0 |
| ; CHECK-NEXT: [[TMP13:%.*]] = extractelement <2 x i32> [[WIDE_LOAD]], i32 0 |
| ; CHECK-NEXT: [[TMP14:%.*]] = sdiv i32 [[TMP12]], [[TMP13]] |
| ; CHECK-NEXT: [[TMP15:%.*]] = extractelement <2 x i32> [[WIDE_LOAD2]], i32 0 |
| ; CHECK-NEXT: [[TMP16:%.*]] = sdiv i32 [[TMP15]], [[TMP14]] |
| ; CHECK-NEXT: [[TMP17:%.*]] = insertelement <2 x i32> poison, i32 [[TMP16]], i32 0 |
| ; CHECK-NEXT: br label [[PRED_SDIV_CONTINUE]] |
| ; CHECK: pred.sdiv.continue: |
| ; CHECK-NEXT: [[TMP19:%.*]] = phi <2 x i32> [ poison, [[VECTOR_BODY]] ], [ [[TMP17]], [[PRED_SDIV_IF]] ] |
| ; CHECK-NEXT: [[TMP20:%.*]] = extractelement <2 x i1> [[TMP10]], i32 1 |
| ; CHECK-NEXT: br i1 [[TMP20]], label [[PRED_SDIV_IF3:%.*]], label [[PRED_SDIV_CONTINUE4]] |
| ; CHECK: pred.sdiv.if3: |
| ; CHECK-NEXT: [[TMP21:%.*]] = extractelement <2 x i32> [[TMP5]], i32 1 |
| ; CHECK-NEXT: [[TMP22:%.*]] = extractelement <2 x i32> [[WIDE_LOAD]], i32 1 |
| ; CHECK-NEXT: [[TMP23:%.*]] = sdiv i32 [[TMP21]], [[TMP22]] |
| ; CHECK-NEXT: [[TMP24:%.*]] = extractelement <2 x i32> [[WIDE_LOAD2]], i32 1 |
| ; CHECK-NEXT: [[TMP25:%.*]] = sdiv i32 [[TMP24]], [[TMP23]] |
| ; CHECK-NEXT: [[TMP26:%.*]] = insertelement <2 x i32> [[TMP19]], i32 [[TMP25]], i32 1 |
| ; CHECK-NEXT: br label [[PRED_SDIV_CONTINUE4]] |
| ; CHECK: pred.sdiv.continue4: |
| ; CHECK-NEXT: [[TMP28:%.*]] = phi <2 x i32> [ [[TMP19]], [[PRED_SDIV_CONTINUE]] ], [ [[TMP26]], [[PRED_SDIV_IF3]] ] |
| ; CHECK-NEXT: [[TMP27:%.*]] = xor <2 x i1> [[TMP8]], splat (i1 true), !dbg [[DBG35]] |
| ; CHECK-NEXT: [[TMP30:%.*]] = select <2 x i1> [[TMP7]], <2 x i1> [[TMP27]], <2 x i1> zeroinitializer, !dbg [[DBG35]] |
| ; CHECK-NEXT: [[PREDPHI:%.*]] = select <2 x i1> [[TMP30]], <2 x i32> [[TMP5]], <2 x i32> [[TMP28]] |
| ; CHECK-NEXT: [[TMP31:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0 |
| ; CHECK-NEXT: store <2 x i32> [[PREDPHI]], ptr [[TMP31]], align 4, !alias.scope [[META29]], !noalias [[META32]] |
| ; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2 |
| ; CHECK-NEXT: [[TMP32:%.*]] = icmp eq i64 [[INDEX_NEXT]], 128 |
| ; CHECK-NEXT: br i1 [[TMP32]], label [[FOR_COND_CLEANUP:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP36:![0-9]+]] |
| ; CHECK: for.cond.cleanup: |
| ; CHECK-NEXT: ret void |
| ; CHECK: for.body: |
| ; CHECK-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ [[INDVARS_IV_NEXT:%.*]], [[IF_END:%.*]] ], [ 0, [[ENTRY]] ] |
| ; CHECK-NEXT: [[ISD:%.*]] = getelementptr inbounds i32, ptr [[ASD]], i64 [[INDVARS_IV]] |
| ; CHECK-NEXT: [[LSD:%.*]] = load i32, ptr [[ISD]], align 4 |
| ; CHECK-NEXT: [[ISD_B:%.*]] = getelementptr inbounds i32, ptr [[BSD]], i64 [[INDVARS_IV]] |
| ; CHECK-NEXT: [[LSD_B:%.*]] = load i32, ptr [[ISD_B]], align 4 |
| ; CHECK-NEXT: [[PSD:%.*]] = add nsw i32 [[LSD]], 23 |
| ; CHECK-NEXT: [[CMP1:%.*]] = icmp slt i32 [[LSD]], 100 |
| ; CHECK-NEXT: [[CMP2:%.*]] = icmp sge i32 [[LSD]], 200 |
| ; CHECK-NEXT: [[OR_COND:%.*]] = or i1 [[CMP1]], [[CMP2]], !dbg [[DBG34]] |
| ; CHECK-NEXT: br i1 [[OR_COND]], label [[IF_THEN:%.*]], label [[IF_END]], !dbg [[DBG34]] |
| ; CHECK: if.then: |
| ; CHECK-NEXT: [[SD1:%.*]] = sdiv i32 [[PSD]], [[LSD]] |
| ; CHECK-NEXT: [[RSD:%.*]] = sdiv i32 [[LSD_B]], [[SD1]] |
| ; CHECK-NEXT: br label [[IF_END]] |
| ; CHECK: if.end: |
| ; CHECK-NEXT: [[YSD_0:%.*]] = phi i32 [ [[RSD]], [[IF_THEN]] ], [ [[PSD]], [[FOR_BODY]] ] |
| ; CHECK-NEXT: store i32 [[YSD_0]], ptr [[ISD]], align 4 |
| ; CHECK-NEXT: [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1 |
| ; CHECK-NEXT: [[EXITCOND:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], 128 |
| ; CHECK-NEXT: br i1 [[EXITCOND]], label [[FOR_COND_CLEANUP]], label [[FOR_BODY]], !llvm.loop [[LOOP37:![0-9]+]] |
| ; |
| ; UNROLL-NO-VF-LABEL: @pr30172( |
| ; UNROLL-NO-VF-NEXT: entry: |
| ; UNROLL-NO-VF-NEXT: br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_MEMCHECK:%.*]] |
| ; UNROLL-NO-VF: vector.memcheck: |
| ; UNROLL-NO-VF-NEXT: [[SCEVGEP:%.*]] = getelementptr i8, ptr [[ASD:%.*]], i64 512 |
| ; UNROLL-NO-VF-NEXT: [[SCEVGEP1:%.*]] = getelementptr i8, ptr [[BSD:%.*]], i64 512 |
| ; UNROLL-NO-VF-NEXT: [[BOUND0:%.*]] = icmp ult ptr [[ASD]], [[SCEVGEP1]] |
| ; UNROLL-NO-VF-NEXT: [[BOUND1:%.*]] = icmp ult ptr [[BSD]], [[SCEVGEP]] |
| ; UNROLL-NO-VF-NEXT: [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]] |
| ; UNROLL-NO-VF-NEXT: br i1 [[FOUND_CONFLICT]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]] |
| ; UNROLL-NO-VF: vector.ph: |
| ; UNROLL-NO-VF-NEXT: br label [[VECTOR_BODY:%.*]] |
| ; UNROLL-NO-VF: vector.body: |
| ; UNROLL-NO-VF-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_SDIV_CONTINUE3:%.*]] ] |
| ; UNROLL-NO-VF-NEXT: [[TMP0:%.*]] = add i64 [[INDEX]], 0 |
| ; UNROLL-NO-VF-NEXT: [[TMP1:%.*]] = add i64 [[INDEX]], 1 |
| ; UNROLL-NO-VF-NEXT: [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[ASD]], i64 [[TMP0]] |
| ; UNROLL-NO-VF-NEXT: [[TMP3:%.*]] = getelementptr inbounds i32, ptr [[ASD]], i64 [[TMP1]] |
| ; UNROLL-NO-VF-NEXT: [[TMP4:%.*]] = load i32, ptr [[TMP2]], align 4, !alias.scope [[META29:![0-9]+]], !noalias [[META32:![0-9]+]] |
| ; UNROLL-NO-VF-NEXT: [[TMP5:%.*]] = load i32, ptr [[TMP3]], align 4, !alias.scope [[META29]], !noalias [[META32]] |
| ; UNROLL-NO-VF-NEXT: [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[BSD]], i64 [[TMP0]] |
| ; UNROLL-NO-VF-NEXT: [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[BSD]], i64 [[TMP1]] |
| ; UNROLL-NO-VF-NEXT: [[TMP8:%.*]] = load i32, ptr [[TMP6]], align 4, !alias.scope [[META32]] |
| ; UNROLL-NO-VF-NEXT: [[TMP9:%.*]] = load i32, ptr [[TMP7]], align 4, !alias.scope [[META32]] |
| ; UNROLL-NO-VF-NEXT: [[TMP10:%.*]] = add nsw i32 [[TMP4]], 23 |
| ; UNROLL-NO-VF-NEXT: [[TMP11:%.*]] = add nsw i32 [[TMP5]], 23 |
| ; UNROLL-NO-VF-NEXT: [[TMP12:%.*]] = icmp slt i32 [[TMP4]], 100 |
| ; UNROLL-NO-VF-NEXT: [[TMP13:%.*]] = icmp slt i32 [[TMP5]], 100 |
| ; UNROLL-NO-VF-NEXT: [[TMP14:%.*]] = xor i1 [[TMP12]], true, !dbg [[DBG34:![0-9]+]] |
| ; UNROLL-NO-VF-NEXT: [[TMP15:%.*]] = xor i1 [[TMP13]], true, !dbg [[DBG34]] |
| ; UNROLL-NO-VF-NEXT: [[TMP16:%.*]] = icmp sge i32 [[TMP4]], 200, !dbg [[DBG34]] |
| ; UNROLL-NO-VF-NEXT: [[TMP17:%.*]] = icmp sge i32 [[TMP5]], 200, !dbg [[DBG34]] |
| ; UNROLL-NO-VF-NEXT: [[TMP18:%.*]] = select i1 [[TMP14]], i1 [[TMP16]], i1 false, !dbg [[DBG35:![0-9]+]] |
| ; UNROLL-NO-VF-NEXT: [[TMP19:%.*]] = select i1 [[TMP15]], i1 [[TMP17]], i1 false, !dbg [[DBG35]] |
| ; UNROLL-NO-VF-NEXT: [[TMP20:%.*]] = or i1 [[TMP18]], [[TMP12]] |
| ; UNROLL-NO-VF-NEXT: [[TMP21:%.*]] = or i1 [[TMP19]], [[TMP13]] |
| ; UNROLL-NO-VF-NEXT: br i1 [[TMP20]], label [[PRED_SDIV_IF:%.*]], label [[PRED_SDIV_CONTINUE:%.*]] |
| ; UNROLL-NO-VF: pred.sdiv.if: |
| ; UNROLL-NO-VF-NEXT: [[TMP22:%.*]] = sdiv i32 [[TMP10]], [[TMP4]] |
| ; UNROLL-NO-VF-NEXT: [[TMP23:%.*]] = sdiv i32 [[TMP8]], [[TMP22]] |
| ; UNROLL-NO-VF-NEXT: br label [[PRED_SDIV_CONTINUE]] |
| ; UNROLL-NO-VF: pred.sdiv.continue: |
| ; UNROLL-NO-VF-NEXT: [[TMP25:%.*]] = phi i32 [ poison, [[VECTOR_BODY]] ], [ [[TMP23]], [[PRED_SDIV_IF]] ] |
| ; UNROLL-NO-VF-NEXT: br i1 [[TMP21]], label [[PRED_SDIV_IF2:%.*]], label [[PRED_SDIV_CONTINUE3]] |
| ; UNROLL-NO-VF: pred.sdiv.if2: |
| ; UNROLL-NO-VF-NEXT: [[TMP26:%.*]] = sdiv i32 [[TMP11]], [[TMP5]] |
| ; UNROLL-NO-VF-NEXT: [[TMP27:%.*]] = sdiv i32 [[TMP9]], [[TMP26]] |
| ; UNROLL-NO-VF-NEXT: br label [[PRED_SDIV_CONTINUE3]] |
| ; UNROLL-NO-VF: pred.sdiv.continue3: |
| ; UNROLL-NO-VF-NEXT: [[TMP29:%.*]] = phi i32 [ poison, [[PRED_SDIV_CONTINUE]] ], [ [[TMP27]], [[PRED_SDIV_IF2]] ] |
| ; UNROLL-NO-VF-NEXT: [[TMP28:%.*]] = xor i1 [[TMP16]], true, !dbg [[DBG35]] |
| ; UNROLL-NO-VF-NEXT: [[TMP30:%.*]] = xor i1 [[TMP17]], true, !dbg [[DBG35]] |
| ; UNROLL-NO-VF-NEXT: [[TMP32:%.*]] = select i1 [[TMP14]], i1 [[TMP28]], i1 false, !dbg [[DBG35]] |
| ; UNROLL-NO-VF-NEXT: [[TMP33:%.*]] = select i1 [[TMP15]], i1 [[TMP30]], i1 false, !dbg [[DBG35]] |
| ; UNROLL-NO-VF-NEXT: [[PREDPHI:%.*]] = select i1 [[TMP32]], i32 [[TMP10]], i32 [[TMP25]] |
| ; UNROLL-NO-VF-NEXT: [[PREDPHI4:%.*]] = select i1 [[TMP33]], i32 [[TMP11]], i32 [[TMP29]] |
| ; UNROLL-NO-VF-NEXT: store i32 [[PREDPHI]], ptr [[TMP2]], align 4, !alias.scope [[META29]], !noalias [[META32]] |
| ; UNROLL-NO-VF-NEXT: store i32 [[PREDPHI4]], ptr [[TMP3]], align 4, !alias.scope [[META29]], !noalias [[META32]] |
| ; UNROLL-NO-VF-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2 |
| ; UNROLL-NO-VF-NEXT: [[TMP34:%.*]] = icmp eq i64 [[INDEX_NEXT]], 128 |
| ; UNROLL-NO-VF-NEXT: br i1 [[TMP34]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP36:![0-9]+]] |
| ; UNROLL-NO-VF: middle.block: |
| ; UNROLL-NO-VF-NEXT: br i1 true, label [[FOR_COND_CLEANUP:%.*]], label [[SCALAR_PH]] |
| ; UNROLL-NO-VF: scalar.ph: |
| ; UNROLL-NO-VF-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ 128, [[MIDDLE_BLOCK]] ], [ 0, [[VECTOR_MEMCHECK]] ], [ 0, [[ENTRY:%.*]] ] |
| ; UNROLL-NO-VF-NEXT: br label [[FOR_BODY:%.*]] |
| ; UNROLL-NO-VF: for.cond.cleanup: |
| ; UNROLL-NO-VF-NEXT: ret void |
| ; UNROLL-NO-VF: for.body: |
| ; UNROLL-NO-VF-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[IF_END:%.*]] ] |
| ; UNROLL-NO-VF-NEXT: [[ISD:%.*]] = getelementptr inbounds i32, ptr [[ASD]], i64 [[INDVARS_IV]] |
| ; UNROLL-NO-VF-NEXT: [[LSD:%.*]] = load i32, ptr [[ISD]], align 4 |
| ; UNROLL-NO-VF-NEXT: [[ISD_B:%.*]] = getelementptr inbounds i32, ptr [[BSD]], i64 [[INDVARS_IV]] |
| ; UNROLL-NO-VF-NEXT: [[LSD_B:%.*]] = load i32, ptr [[ISD_B]], align 4 |
| ; UNROLL-NO-VF-NEXT: [[PSD:%.*]] = add nsw i32 [[LSD]], 23 |
| ; UNROLL-NO-VF-NEXT: [[CMP1:%.*]] = icmp slt i32 [[LSD]], 100 |
| ; UNROLL-NO-VF-NEXT: br i1 [[CMP1]], label [[IF_THEN:%.*]], label [[CHECKBB:%.*]], !dbg [[DBG34]] |
| ; UNROLL-NO-VF: checkbb: |
| ; UNROLL-NO-VF-NEXT: [[CMP2:%.*]] = icmp sge i32 [[LSD]], 200 |
| ; UNROLL-NO-VF-NEXT: br i1 [[CMP2]], label [[IF_THEN]], label [[IF_END]], !dbg [[DBG35]] |
| ; UNROLL-NO-VF: if.then: |
| ; UNROLL-NO-VF-NEXT: [[SD1:%.*]] = sdiv i32 [[PSD]], [[LSD]] |
| ; UNROLL-NO-VF-NEXT: [[RSD:%.*]] = sdiv i32 [[LSD_B]], [[SD1]] |
| ; UNROLL-NO-VF-NEXT: br label [[IF_END]] |
| ; UNROLL-NO-VF: if.end: |
| ; UNROLL-NO-VF-NEXT: [[YSD_0:%.*]] = phi i32 [ [[RSD]], [[IF_THEN]] ], [ [[PSD]], [[CHECKBB]] ] |
| ; UNROLL-NO-VF-NEXT: store i32 [[YSD_0]], ptr [[ISD]], align 4 |
| ; UNROLL-NO-VF-NEXT: [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1 |
| ; UNROLL-NO-VF-NEXT: [[EXITCOND:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], 128 |
| ; UNROLL-NO-VF-NEXT: br i1 [[EXITCOND]], label [[FOR_COND_CLEANUP]], label [[FOR_BODY]], !llvm.loop [[LOOP37:![0-9]+]] |
| ; |
| entry: |
| br label %for.body |
| |
| for.cond.cleanup: ; preds = %if.end |
| ret void |
| |
| for.body: ; preds = %if.end, %entry |
| %indvars.iv = phi i64 [ 0, %entry ], [ %indvars.iv.next, %if.end ] |
| %isd = getelementptr inbounds i32, ptr %asd, i64 %indvars.iv |
| %lsd = load i32, ptr %isd, align 4 |
| %isd.b = getelementptr inbounds i32, ptr %bsd, i64 %indvars.iv |
| %lsd.b = load i32, ptr %isd.b, align 4 |
| %psd = add nsw i32 %lsd, 23 |
| %cmp1 = icmp slt i32 %lsd, 100 |
| br i1 %cmp1, label %if.then, label %checkbb, !dbg !7 |
| |
| checkbb: ; preds = %for.body |
| %cmp2 = icmp sge i32 %lsd, 200 |
| br i1 %cmp2, label %if.then, label %if.end, !dbg !8 |
| |
| if.then: ; preds = %checkbb, %for.body |
| %sd1 = sdiv i32 %psd, %lsd |
| %rsd = sdiv i32 %lsd.b, %sd1 |
| br label %if.end |
| |
| if.end: ; preds = %if.then, %checkbb |
| %ysd.0 = phi i32 [ %rsd, %if.then ], [ %psd, %checkbb ] |
| store i32 %ysd.0, ptr %isd, align 4 |
| %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1 |
| %exitcond = icmp eq i64 %indvars.iv.next, 128 |
| br i1 %exitcond, label %for.cond.cleanup, label %for.body |
| } |
| |
| define i32 @predicated_udiv_scalarized_operand(ptr %a, i1 %c, i32 %x, i64 %n) { |
| ; CHECK-LABEL: @predicated_udiv_scalarized_operand( |
| ; CHECK-NEXT: entry: |
| ; CHECK-NEXT: [[SMAX:%.*]] = call i64 @llvm.smax.i64(i64 [[N:%.*]], i64 1) |
| ; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[SMAX]], 2 |
| ; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]] |
| ; CHECK: vector.ph: |
| ; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[SMAX]], 2 |
| ; CHECK-NEXT: [[N_VEC:%.*]] = sub i64 [[SMAX]], [[N_MOD_VF]] |
| ; CHECK-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x i1> poison, i1 [[C:%.*]], i64 0 |
| ; CHECK-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <2 x i1> [[BROADCAST_SPLATINSERT]], <2 x i1> poison, <2 x i32> zeroinitializer |
| ; CHECK-NEXT: br label [[VECTOR_BODY:%.*]] |
| ; CHECK: vector.body: |
| ; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_UDIV_CONTINUE2:%.*]] ] |
| ; CHECK-NEXT: [[VEC_PHI:%.*]] = phi <2 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP18:%.*]], [[PRED_UDIV_CONTINUE2]] ] |
| ; CHECK-NEXT: [[TMP0:%.*]] = add i64 [[INDEX]], 0 |
| ; CHECK-NEXT: [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[TMP0]] |
| ; CHECK-NEXT: [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0 |
| ; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP2]], align 4 |
| ; CHECK-NEXT: [[TMP3:%.*]] = extractelement <2 x i1> [[BROADCAST_SPLAT]], i32 0 |
| ; CHECK-NEXT: br i1 [[TMP3]], label [[PRED_UDIV_IF:%.*]], label [[PRED_UDIV_CONTINUE:%.*]] |
| ; CHECK: pred.udiv.if: |
| ; CHECK-NEXT: [[TMP4:%.*]] = extractelement <2 x i32> [[WIDE_LOAD]], i32 0 |
| ; CHECK-NEXT: [[TMP5:%.*]] = add nsw i32 [[TMP4]], [[X:%.*]] |
| ; CHECK-NEXT: [[TMP6:%.*]] = extractelement <2 x i32> [[WIDE_LOAD]], i32 0 |
| ; CHECK-NEXT: [[TMP7:%.*]] = udiv i32 [[TMP6]], [[TMP5]] |
| ; CHECK-NEXT: [[TMP8:%.*]] = insertelement <2 x i32> poison, i32 [[TMP7]], i32 0 |
| ; CHECK-NEXT: br label [[PRED_UDIV_CONTINUE]] |
| ; CHECK: pred.udiv.continue: |
| ; CHECK-NEXT: [[TMP9:%.*]] = phi <2 x i32> [ poison, [[VECTOR_BODY]] ], [ [[TMP8]], [[PRED_UDIV_IF]] ] |
| ; CHECK-NEXT: [[TMP10:%.*]] = extractelement <2 x i1> [[BROADCAST_SPLAT]], i32 1 |
| ; CHECK-NEXT: br i1 [[TMP10]], label [[PRED_UDIV_IF1:%.*]], label [[PRED_UDIV_CONTINUE2]] |
| ; CHECK: pred.udiv.if1: |
| ; CHECK-NEXT: [[TMP11:%.*]] = extractelement <2 x i32> [[WIDE_LOAD]], i32 1 |
| ; CHECK-NEXT: [[TMP12:%.*]] = add nsw i32 [[TMP11]], [[X]] |
| ; CHECK-NEXT: [[TMP13:%.*]] = extractelement <2 x i32> [[WIDE_LOAD]], i32 1 |
| ; CHECK-NEXT: [[TMP14:%.*]] = udiv i32 [[TMP13]], [[TMP12]] |
| ; CHECK-NEXT: [[TMP15:%.*]] = insertelement <2 x i32> [[TMP9]], i32 [[TMP14]], i32 1 |
| ; CHECK-NEXT: br label [[PRED_UDIV_CONTINUE2]] |
| ; CHECK: pred.udiv.continue2: |
| ; CHECK-NEXT: [[TMP16:%.*]] = phi <2 x i32> [ [[TMP9]], [[PRED_UDIV_CONTINUE]] ], [ [[TMP15]], [[PRED_UDIV_IF1]] ] |
| ; CHECK-NEXT: [[PREDPHI:%.*]] = select <2 x i1> [[BROADCAST_SPLAT]], <2 x i32> [[TMP16]], <2 x i32> [[WIDE_LOAD]] |
| ; CHECK-NEXT: [[TMP18]] = add <2 x i32> [[VEC_PHI]], [[PREDPHI]] |
| ; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2 |
| ; CHECK-NEXT: [[TMP19:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] |
| ; CHECK-NEXT: br i1 [[TMP19]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP38:![0-9]+]] |
| ; CHECK: middle.block: |
| ; CHECK-NEXT: [[TMP20:%.*]] = call i32 @llvm.vector.reduce.add.v2i32(<2 x i32> [[TMP18]]) |
| ; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[SMAX]], [[N_VEC]] |
| ; CHECK-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]] |
| ; CHECK: scalar.ph: |
| ; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ] |
| ; CHECK-NEXT: [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP20]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ] |
| ; CHECK-NEXT: br label [[FOR_BODY:%.*]] |
| ; CHECK: for.body: |
| ; CHECK-NEXT: [[I:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[I_NEXT:%.*]], [[FOR_INC:%.*]] ] |
| ; CHECK-NEXT: [[R:%.*]] = phi i32 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[T6:%.*]], [[FOR_INC]] ] |
| ; CHECK-NEXT: [[T0:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[I]] |
| ; CHECK-NEXT: [[T2:%.*]] = load i32, ptr [[T0]], align 4 |
| ; CHECK-NEXT: br i1 [[C]], label [[IF_THEN:%.*]], label [[FOR_INC]] |
| ; CHECK: if.then: |
| ; CHECK-NEXT: [[T3:%.*]] = add nsw i32 [[T2]], [[X]] |
| ; CHECK-NEXT: [[T4:%.*]] = udiv i32 [[T2]], [[T3]] |
| ; CHECK-NEXT: br label [[FOR_INC]] |
| ; CHECK: for.inc: |
| ; CHECK-NEXT: [[T5:%.*]] = phi i32 [ [[T2]], [[FOR_BODY]] ], [ [[T4]], [[IF_THEN]] ] |
| ; CHECK-NEXT: [[T6]] = add i32 [[R]], [[T5]] |
| ; CHECK-NEXT: [[I_NEXT]] = add nuw nsw i64 [[I]], 1 |
| ; CHECK-NEXT: [[COND:%.*]] = icmp slt i64 [[I_NEXT]], [[N]] |
| ; CHECK-NEXT: br i1 [[COND]], label [[FOR_BODY]], label [[FOR_END]], !llvm.loop [[LOOP39:![0-9]+]] |
| ; CHECK: for.end: |
| ; CHECK-NEXT: [[T7:%.*]] = phi i32 [ [[T6]], [[FOR_INC]] ], [ [[TMP20]], [[MIDDLE_BLOCK]] ] |
| ; CHECK-NEXT: ret i32 [[T7]] |
| ; |
| ; UNROLL-NO-VF-LABEL: @predicated_udiv_scalarized_operand( |
| ; UNROLL-NO-VF-NEXT: entry: |
| ; UNROLL-NO-VF-NEXT: [[SMAX:%.*]] = call i64 @llvm.smax.i64(i64 [[N:%.*]], i64 1) |
| ; UNROLL-NO-VF-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[SMAX]], 2 |
| ; UNROLL-NO-VF-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]] |
| ; UNROLL-NO-VF: vector.ph: |
| ; UNROLL-NO-VF-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[SMAX]], 2 |
| ; UNROLL-NO-VF-NEXT: [[N_VEC:%.*]] = sub i64 [[SMAX]], [[N_MOD_VF]] |
| ; UNROLL-NO-VF-NEXT: br label [[VECTOR_BODY:%.*]] |
| ; UNROLL-NO-VF: vector.body: |
| ; UNROLL-NO-VF-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_UDIV_CONTINUE3:%.*]] ] |
| ; UNROLL-NO-VF-NEXT: [[VEC_PHI:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[TMP14:%.*]], [[PRED_UDIV_CONTINUE3]] ] |
| ; UNROLL-NO-VF-NEXT: [[VEC_PHI1:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[TMP15:%.*]], [[PRED_UDIV_CONTINUE3]] ] |
| ; UNROLL-NO-VF-NEXT: [[TMP0:%.*]] = add i64 [[INDEX]], 0 |
| ; UNROLL-NO-VF-NEXT: [[TMP1:%.*]] = add i64 [[INDEX]], 1 |
| ; UNROLL-NO-VF-NEXT: [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[TMP0]] |
| ; UNROLL-NO-VF-NEXT: [[TMP3:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP1]] |
| ; UNROLL-NO-VF-NEXT: [[TMP4:%.*]] = load i32, ptr [[TMP2]], align 4 |
| ; UNROLL-NO-VF-NEXT: [[TMP5:%.*]] = load i32, ptr [[TMP3]], align 4 |
| ; UNROLL-NO-VF-NEXT: br i1 [[C:%.*]], label [[PRED_UDIV_IF:%.*]], label [[PRED_UDIV_CONTINUE:%.*]] |
| ; UNROLL-NO-VF: pred.udiv.if: |
| ; UNROLL-NO-VF-NEXT: [[TMP6:%.*]] = add nsw i32 [[TMP4]], [[X:%.*]] |
| ; UNROLL-NO-VF-NEXT: [[TMP7:%.*]] = udiv i32 [[TMP4]], [[TMP6]] |
| ; UNROLL-NO-VF-NEXT: br label [[PRED_UDIV_CONTINUE]] |
| ; UNROLL-NO-VF: pred.udiv.continue: |
| ; UNROLL-NO-VF-NEXT: [[TMP8:%.*]] = phi i32 [ poison, [[VECTOR_BODY]] ], [ [[TMP7]], [[PRED_UDIV_IF]] ] |
| ; UNROLL-NO-VF-NEXT: br i1 [[C]], label [[PRED_UDIV_IF2:%.*]], label [[PRED_UDIV_CONTINUE3]] |
| ; UNROLL-NO-VF: pred.udiv.if2: |
| ; UNROLL-NO-VF-NEXT: [[TMP9:%.*]] = add nsw i32 [[TMP5]], [[X]] |
| ; UNROLL-NO-VF-NEXT: [[TMP10:%.*]] = udiv i32 [[TMP5]], [[TMP9]] |
| ; UNROLL-NO-VF-NEXT: br label [[PRED_UDIV_CONTINUE3]] |
| ; UNROLL-NO-VF: pred.udiv.continue3: |
| ; UNROLL-NO-VF-NEXT: [[TMP11:%.*]] = phi i32 [ poison, [[PRED_UDIV_CONTINUE]] ], [ [[TMP10]], [[PRED_UDIV_IF2]] ] |
| ; UNROLL-NO-VF-NEXT: [[PREDPHI:%.*]] = select i1 [[C]], i32 [[TMP8]], i32 [[TMP4]] |
| ; UNROLL-NO-VF-NEXT: [[PREDPHI4:%.*]] = select i1 [[C]], i32 [[TMP11]], i32 [[TMP5]] |
| ; UNROLL-NO-VF-NEXT: [[TMP14]] = add i32 [[VEC_PHI]], [[PREDPHI]] |
| ; UNROLL-NO-VF-NEXT: [[TMP15]] = add i32 [[VEC_PHI1]], [[PREDPHI4]] |
| ; UNROLL-NO-VF-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2 |
| ; UNROLL-NO-VF-NEXT: [[TMP16:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] |
| ; UNROLL-NO-VF-NEXT: br i1 [[TMP16]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP38:![0-9]+]] |
| ; UNROLL-NO-VF: middle.block: |
| ; UNROLL-NO-VF-NEXT: [[BIN_RDX:%.*]] = add i32 [[TMP15]], [[TMP14]] |
| ; UNROLL-NO-VF-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[SMAX]], [[N_VEC]] |
| ; UNROLL-NO-VF-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]] |
| ; UNROLL-NO-VF: scalar.ph: |
| ; UNROLL-NO-VF-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ] |
| ; UNROLL-NO-VF-NEXT: [[BC_MERGE_RDX:%.*]] = phi i32 [ [[BIN_RDX]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ] |
| ; UNROLL-NO-VF-NEXT: br label [[FOR_BODY:%.*]] |
| ; UNROLL-NO-VF: for.body: |
| ; UNROLL-NO-VF-NEXT: [[I:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[I_NEXT:%.*]], [[FOR_INC:%.*]] ] |
| ; UNROLL-NO-VF-NEXT: [[R:%.*]] = phi i32 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[T6:%.*]], [[FOR_INC]] ] |
| ; UNROLL-NO-VF-NEXT: [[T0:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[I]] |
| ; UNROLL-NO-VF-NEXT: [[T2:%.*]] = load i32, ptr [[T0]], align 4 |
| ; UNROLL-NO-VF-NEXT: br i1 [[C]], label [[IF_THEN:%.*]], label [[FOR_INC]] |
| ; UNROLL-NO-VF: if.then: |
| ; UNROLL-NO-VF-NEXT: [[T3:%.*]] = add nsw i32 [[T2]], [[X]] |
| ; UNROLL-NO-VF-NEXT: [[T4:%.*]] = udiv i32 [[T2]], [[T3]] |
| ; UNROLL-NO-VF-NEXT: br label [[FOR_INC]] |
| ; UNROLL-NO-VF: for.inc: |
| ; UNROLL-NO-VF-NEXT: [[T5:%.*]] = phi i32 [ [[T2]], [[FOR_BODY]] ], [ [[T4]], [[IF_THEN]] ] |
| ; UNROLL-NO-VF-NEXT: [[T6]] = add i32 [[R]], [[T5]] |
| ; UNROLL-NO-VF-NEXT: [[I_NEXT]] = add nuw nsw i64 [[I]], 1 |
| ; UNROLL-NO-VF-NEXT: [[COND:%.*]] = icmp slt i64 [[I_NEXT]], [[N]] |
| ; UNROLL-NO-VF-NEXT: br i1 [[COND]], label [[FOR_BODY]], label [[FOR_END]], !llvm.loop [[LOOP39:![0-9]+]] |
| ; UNROLL-NO-VF: for.end: |
| ; UNROLL-NO-VF-NEXT: [[T7:%.*]] = phi i32 [ [[T6]], [[FOR_INC]] ], [ [[BIN_RDX]], [[MIDDLE_BLOCK]] ] |
| ; UNROLL-NO-VF-NEXT: ret i32 [[T7]] |
| ; |
| entry: |
| br label %for.body |
| |
| |
| ; Test predicating an instruction that feeds a vectorizable use, when unrolled |
| ; but not vectorized. Derived from pr34248 reproducer. |
| for.body: |
| %i = phi i64 [ 0, %entry ], [ %i.next, %for.inc ] |
| %r = phi i32 [ 0, %entry ], [ %t6, %for.inc ] |
| %t0 = getelementptr inbounds i32, ptr %a, i64 %i |
| %t2 = load i32, ptr %t0, align 4 |
| br i1 %c, label %if.then, label %for.inc |
| |
| if.then: |
| %t3 = add nsw i32 %t2, %x |
| %t4 = udiv i32 %t2, %t3 |
| br label %for.inc |
| |
| for.inc: |
| %t5 = phi i32 [ %t2, %for.body ], [ %t4, %if.then] |
| %t6 = add i32 %r, %t5 |
| %i.next = add nuw nsw i64 %i, 1 |
| %cond = icmp slt i64 %i.next, %n |
| br i1 %cond, label %for.body, label %for.end |
| |
| for.end: |
| %t7 = phi i32 [ %t6, %for.inc ] |
| ret i32 %t7 |
| } |
| |
| !llvm.dbg.cu = !{!0} |
| !llvm.module.flags = !{!3, !4} |
| |
| !0 = distinct !DICompileUnit(language: DW_LANG_C99, file: !1, producer: "clang", isOptimized: true, runtimeVersion: 0, emissionKind: NoDebug, enums: !2) |
| !1 = !DIFile(filename: "/tmp/s.c", directory: "/tmp") |
| !2 = !{} |
| !3 = !{i32 2, !"Debug Info Version", i32 3} |
| !4 = !{i32 7, !"PIC Level", i32 2} |
| !5 = distinct !DISubprogram(name: "f", scope: !1, file: !1, line: 4, type: !6, scopeLine: 4, flags: DIFlagPrototyped, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !0, retainedNodes: !2) |
| !6 = !DISubroutineType(types: !2) |
| !7 = !DILocation(line: 5, column: 21, scope: !5) |
| !8 = !DILocation(line: 5, column: 3, scope: !5) |
| |
| |
