240 lines · plain
1; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 62; RUN: opt < %s -O3 -S | FileCheck %s3; RUN: opt < %s -passes="default<O3>" -S | FileCheck %s4 5; Test that IR is optimal after vectorization/unrolling/CSE/canonicalization.6; In particular, there should be no fdivs inside loops because that is expensive.7 8; TODO: There is a CSE opportunity to reduce the hoisted fdivs after vectorization/unrolling.9; PR46115 - https://llvm.org/PR4611510 11target datalayout = "e-m:o-p270:32:32-p271:32:32-p272:64:64-i64:64-f80:128-n8:16:32:64-S128"12target triple = "x86_64-apple-macosx10.15.0"13 14define void @vdiv(ptr %x, ptr %y, double %a, i32 %N) #0 {15; CHECK-LABEL: define void @vdiv(16; CHECK-SAME: ptr writeonly captures(none) [[X:%.*]], ptr readonly captures(none) [[Y:%.*]], double [[A:%.*]], i32 [[N:%.*]]) local_unnamed_addr #[[ATTR0:[0-9]+]] {17; CHECK-NEXT: [[ENTRY:.*:]]18; CHECK-NEXT: [[CMP1:%.*]] = icmp sgt i32 [[N]], 019; CHECK-NEXT: br i1 [[CMP1]], label %[[FOR_BODY_PREHEADER:.*]], label %[[FOR_END:.*]]20; CHECK: [[FOR_BODY_PREHEADER]]:21; CHECK-NEXT: [[X4:%.*]] = ptrtoint ptr [[X]] to i6422; CHECK-NEXT: [[Y5:%.*]] = ptrtoint ptr [[Y]] to i6423; CHECK-NEXT: [[WIDE_TRIP_COUNT:%.*]] = zext nneg i32 [[N]] to i6424; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[N]], 425; CHECK-NEXT: [[TMP0:%.*]] = sub i64 [[X4]], [[Y5]]26; CHECK-NEXT: [[DIFF_CHECK:%.*]] = icmp ult i64 [[TMP0]], 12827; CHECK-NEXT: [[OR_COND:%.*]] = select i1 [[MIN_ITERS_CHECK]], i1 true, i1 [[DIFF_CHECK]]28; CHECK-NEXT: br i1 [[OR_COND]], label %[[FOR_BODY_PREHEADER9:.*]], label %[[VECTOR_PH:.*]]29; CHECK: [[VECTOR_PH]]:30; CHECK-NEXT: [[MIN_ITERS_CHECK6:%.*]] = icmp ult i32 [[N]], 1631; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK6]], label %[[VEC_EPILOG_PH:.*]], label %[[VECTOR_PH1:.*]]32; CHECK: [[VECTOR_PH1]]:33; CHECK-NEXT: [[N_VEC_REMAINING:%.*]] = and i64 [[WIDE_TRIP_COUNT]], 1234; CHECK-NEXT: [[N_VEC:%.*]] = and i64 [[WIDE_TRIP_COUNT]], 214748363235; CHECK-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x double> poison, double [[A]], i64 036; CHECK-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x double> [[BROADCAST_SPLATINSERT]], <4 x double> poison, <4 x i32> zeroinitializer37; CHECK-NEXT: [[TMP1:%.*]] = fdiv fast <4 x double> splat (double 1.000000e+00), [[BROADCAST_SPLAT]]38; CHECK-NEXT: [[TMP2:%.*]] = fdiv fast <4 x double> splat (double 1.000000e+00), [[BROADCAST_SPLAT]]39; CHECK-NEXT: [[TMP3:%.*]] = fdiv fast <4 x double> splat (double 1.000000e+00), [[BROADCAST_SPLAT]]40; CHECK-NEXT: [[TMP4:%.*]] = fdiv fast <4 x double> splat (double 1.000000e+00), [[BROADCAST_SPLAT]]41; CHECK-NEXT: br label %[[VECTOR_BODY:.*]]42; CHECK: [[VECTOR_BODY]]:43; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH1]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]44; CHECK-NEXT: [[TMP5:%.*]] = getelementptr inbounds nuw double, ptr [[Y]], i64 [[INDEX]]45; CHECK-NEXT: [[TMP6:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP5]], i64 3246; CHECK-NEXT: [[TMP7:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP5]], i64 6447; CHECK-NEXT: [[TMP8:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP5]], i64 9648; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <4 x double>, ptr [[TMP5]], align 8, !tbaa [[DOUBLE_TBAA3:![0-9]+]]49; CHECK-NEXT: [[WIDE_LOAD6:%.*]] = load <4 x double>, ptr [[TMP6]], align 8, !tbaa [[DOUBLE_TBAA3]]50; CHECK-NEXT: [[WIDE_LOAD7:%.*]] = load <4 x double>, ptr [[TMP7]], align 8, !tbaa [[DOUBLE_TBAA3]]51; CHECK-NEXT: [[WIDE_LOAD8:%.*]] = load <4 x double>, ptr [[TMP8]], align 8, !tbaa [[DOUBLE_TBAA3]]52; CHECK-NEXT: [[TMP9:%.*]] = fmul fast <4 x double> [[WIDE_LOAD]], [[TMP1]]53; CHECK-NEXT: [[TMP10:%.*]] = fmul fast <4 x double> [[WIDE_LOAD6]], [[TMP2]]54; CHECK-NEXT: [[TMP11:%.*]] = fmul fast <4 x double> [[WIDE_LOAD7]], [[TMP3]]55; CHECK-NEXT: [[TMP12:%.*]] = fmul fast <4 x double> [[WIDE_LOAD8]], [[TMP4]]56; CHECK-NEXT: [[TMP13:%.*]] = getelementptr inbounds nuw double, ptr [[X]], i64 [[INDEX]]57; CHECK-NEXT: [[TMP14:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP13]], i64 3258; CHECK-NEXT: [[TMP15:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP13]], i64 6459; CHECK-NEXT: [[TMP16:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP13]], i64 9660; CHECK-NEXT: store <4 x double> [[TMP9]], ptr [[TMP13]], align 8, !tbaa [[DOUBLE_TBAA3]]61; CHECK-NEXT: store <4 x double> [[TMP10]], ptr [[TMP14]], align 8, !tbaa [[DOUBLE_TBAA3]]62; CHECK-NEXT: store <4 x double> [[TMP11]], ptr [[TMP15]], align 8, !tbaa [[DOUBLE_TBAA3]]63; CHECK-NEXT: store <4 x double> [[TMP12]], ptr [[TMP16]], align 8, !tbaa [[DOUBLE_TBAA3]]64; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 1665; CHECK-NEXT: [[TMP17:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]66; CHECK-NEXT: br i1 [[TMP17]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP7:![0-9]+]]67; CHECK: [[MIDDLE_BLOCK]]:68; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[N_VEC]], [[WIDE_TRIP_COUNT]]69; CHECK-NEXT: br i1 [[CMP_N]], label %[[FOR_END]], label %[[VEC_EPILOG_ITER_CHECK:.*]]70; CHECK: [[VEC_EPILOG_ITER_CHECK]]:71; CHECK-NEXT: [[MIN_EPILOG_ITERS_CHECK:%.*]] = icmp eq i64 [[N_VEC_REMAINING]], 072; CHECK-NEXT: br i1 [[MIN_EPILOG_ITERS_CHECK]], label %[[FOR_BODY_PREHEADER9]], label %[[VEC_EPILOG_PH]]73; CHECK: [[VEC_EPILOG_PH]]:74; CHECK-NEXT: [[VEC_EPILOG_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[VEC_EPILOG_ITER_CHECK]] ], [ 0, %[[VECTOR_PH]] ]75; CHECK-NEXT: [[N_VEC11:%.*]] = and i64 [[WIDE_TRIP_COUNT]], 214748364476; CHECK-NEXT: [[BROADCAST_SPLATINSERT14:%.*]] = insertelement <4 x double> poison, double [[A]], i64 077; CHECK-NEXT: [[BROADCAST_SPLAT15:%.*]] = shufflevector <4 x double> [[BROADCAST_SPLATINSERT14]], <4 x double> poison, <4 x i32> zeroinitializer78; CHECK-NEXT: [[TMP38:%.*]] = fdiv fast <4 x double> splat (double 1.000000e+00), [[BROADCAST_SPLAT15]]79; CHECK-NEXT: br label %[[VEC_EPILOG_VECTOR_BODY:.*]]80; CHECK: [[VEC_EPILOG_VECTOR_BODY]]:81; CHECK-NEXT: [[INDEX12:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], %[[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT16:%.*]], %[[VEC_EPILOG_VECTOR_BODY]] ]82; CHECK-NEXT: [[TMP39:%.*]] = getelementptr inbounds nuw double, ptr [[Y]], i64 [[INDEX12]]83; CHECK-NEXT: [[WIDE_LOAD13:%.*]] = load <4 x double>, ptr [[TMP39]], align 8, !tbaa [[DOUBLE_TBAA3]]84; CHECK-NEXT: [[TMP40:%.*]] = fmul fast <4 x double> [[WIDE_LOAD13]], [[TMP38]]85; CHECK-NEXT: [[TMP41:%.*]] = getelementptr inbounds nuw double, ptr [[X]], i64 [[INDEX12]]86; CHECK-NEXT: store <4 x double> [[TMP40]], ptr [[TMP41]], align 8, !tbaa [[DOUBLE_TBAA3]]87; CHECK-NEXT: [[INDEX_NEXT16]] = add nuw i64 [[INDEX12]], 488; CHECK-NEXT: [[TMP42:%.*]] = icmp eq i64 [[INDEX_NEXT16]], [[N_VEC11]]89; CHECK-NEXT: br i1 [[TMP42]], label %[[VEC_EPILOG_MIDDLE_BLOCK:.*]], label %[[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]90; CHECK: [[VEC_EPILOG_MIDDLE_BLOCK]]:91; CHECK-NEXT: [[CMP_N17:%.*]] = icmp eq i64 [[N_VEC11]], [[WIDE_TRIP_COUNT]]92; CHECK-NEXT: br i1 [[CMP_N17]], label %[[FOR_END]], label %[[FOR_BODY_PREHEADER9]]93; CHECK: [[FOR_BODY_PREHEADER9]]:94; CHECK-NEXT: [[INDVARS_IV_PH:%.*]] = phi i64 [ 0, %[[FOR_BODY_PREHEADER]] ], [ [[N_VEC]], %[[VEC_EPILOG_ITER_CHECK]] ], [ [[N_VEC11]], %[[VEC_EPILOG_MIDDLE_BLOCK]] ]95; CHECK-NEXT: [[TMP43:%.*]] = sub nsw i64 [[WIDE_TRIP_COUNT]], [[INDVARS_IV_PH]]96; CHECK-NEXT: [[XTRAITER:%.*]] = and i64 [[TMP43]], 797; CHECK-NEXT: [[LCMP_MOD_NOT:%.*]] = icmp eq i64 [[XTRAITER]], 098; CHECK-NEXT: br i1 [[LCMP_MOD_NOT]], label %[[FOR_BODY_PROL_LOOPEXIT:.*]], label %[[FOR_BODY_PROL_PREHEADER:.*]]99; CHECK: [[FOR_BODY_PROL_PREHEADER]]:100; CHECK-NEXT: [[TMP18:%.*]] = fdiv fast double 1.000000e+00, [[A]]101; CHECK-NEXT: br label %[[FOR_BODY_PROL:.*]]102; CHECK: [[FOR_BODY_PROL]]:103; CHECK-NEXT: [[INDVARS_IV_PROL:%.*]] = phi i64 [ [[INDVARS_IV_NEXT_PROL:%.*]], %[[FOR_BODY_PROL]] ], [ [[INDVARS_IV_PH]], %[[FOR_BODY_PROL_PREHEADER]] ]104; CHECK-NEXT: [[PROL_ITER:%.*]] = phi i64 [ [[PROL_ITER_NEXT:%.*]], %[[FOR_BODY_PROL]] ], [ 0, %[[FOR_BODY_PROL_PREHEADER]] ]105; CHECK-NEXT: [[ARRAYIDX_PROL:%.*]] = getelementptr inbounds nuw double, ptr [[Y]], i64 [[INDVARS_IV_PROL]]106; CHECK-NEXT: [[T0_PROL:%.*]] = load double, ptr [[ARRAYIDX_PROL]], align 8, !tbaa [[DOUBLE_TBAA3]]107; CHECK-NEXT: [[TMP19:%.*]] = fmul fast double [[T0_PROL]], [[TMP18]]108; CHECK-NEXT: [[ARRAYIDX2_PROL:%.*]] = getelementptr inbounds nuw double, ptr [[X]], i64 [[INDVARS_IV_PROL]]109; CHECK-NEXT: store double [[TMP19]], ptr [[ARRAYIDX2_PROL]], align 8, !tbaa [[DOUBLE_TBAA3]]110; CHECK-NEXT: [[INDVARS_IV_NEXT_PROL]] = add nuw nsw i64 [[INDVARS_IV_PROL]], 1111; CHECK-NEXT: [[PROL_ITER_NEXT]] = add i64 [[PROL_ITER]], 1112; CHECK-NEXT: [[PROL_ITER_CMP_NOT:%.*]] = icmp eq i64 [[PROL_ITER_NEXT]], [[XTRAITER]]113; CHECK-NEXT: br i1 [[PROL_ITER_CMP_NOT]], label %[[FOR_BODY_PROL_LOOPEXIT]], label %[[FOR_BODY_PROL]], !llvm.loop [[LOOP11:![0-9]+]]114; CHECK: [[FOR_BODY_PROL_LOOPEXIT]]:115; CHECK-NEXT: [[INDVARS_IV_UNR:%.*]] = phi i64 [ [[INDVARS_IV_PH]], %[[FOR_BODY_PREHEADER9]] ], [ [[INDVARS_IV_NEXT_PROL]], %[[FOR_BODY_PROL]] ]116; CHECK-NEXT: [[TMP20:%.*]] = sub nsw i64 [[INDVARS_IV_PH]], [[WIDE_TRIP_COUNT]]117; CHECK-NEXT: [[TMP21:%.*]] = icmp ugt i64 [[TMP20]], -8118; CHECK-NEXT: br i1 [[TMP21]], label %[[FOR_END]], label %[[FOR_BODY_PREHEADER9_NEW:.*]]119; CHECK: [[FOR_BODY_PREHEADER9_NEW]]:120; CHECK-NEXT: [[TMP22:%.*]] = fdiv fast double 1.000000e+00, [[A]]121; CHECK-NEXT: [[TMP23:%.*]] = fdiv fast double 1.000000e+00, [[A]]122; CHECK-NEXT: [[TMP24:%.*]] = fdiv fast double 1.000000e+00, [[A]]123; CHECK-NEXT: [[TMP25:%.*]] = fdiv fast double 1.000000e+00, [[A]]124; CHECK-NEXT: [[TMP26:%.*]] = fdiv fast double 1.000000e+00, [[A]]125; CHECK-NEXT: [[TMP27:%.*]] = fdiv fast double 1.000000e+00, [[A]]126; CHECK-NEXT: [[TMP28:%.*]] = fdiv fast double 1.000000e+00, [[A]]127; CHECK-NEXT: [[TMP29:%.*]] = fdiv fast double 1.000000e+00, [[A]]128; CHECK-NEXT: br label %[[FOR_BODY:.*]]129; CHECK: [[FOR_BODY]]:130; CHECK-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ [[INDVARS_IV_UNR]], %[[FOR_BODY_PREHEADER9_NEW]] ], [ [[INDVARS_IV_NEXT_7:%.*]], %[[FOR_BODY]] ]131; CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds nuw double, ptr [[Y]], i64 [[INDVARS_IV]]132; CHECK-NEXT: [[T0:%.*]] = load double, ptr [[ARRAYIDX]], align 8, !tbaa [[DOUBLE_TBAA3]]133; CHECK-NEXT: [[TMP30:%.*]] = fmul fast double [[T0]], [[TMP22]]134; CHECK-NEXT: [[ARRAYIDX2:%.*]] = getelementptr inbounds nuw double, ptr [[X]], i64 [[INDVARS_IV]]135; CHECK-NEXT: store double [[TMP30]], ptr [[ARRAYIDX2]], align 8, !tbaa [[DOUBLE_TBAA3]]136; CHECK-NEXT: [[INDVARS_IV_NEXT:%.*]] = add nuw nsw i64 [[INDVARS_IV]], 1137; CHECK-NEXT: [[ARRAYIDX_1:%.*]] = getelementptr inbounds nuw double, ptr [[Y]], i64 [[INDVARS_IV_NEXT]]138; CHECK-NEXT: [[T0_1:%.*]] = load double, ptr [[ARRAYIDX_1]], align 8, !tbaa [[DOUBLE_TBAA3]]139; CHECK-NEXT: [[TMP31:%.*]] = fmul fast double [[T0_1]], [[TMP23]]140; CHECK-NEXT: [[ARRAYIDX2_1:%.*]] = getelementptr inbounds nuw double, ptr [[X]], i64 [[INDVARS_IV_NEXT]]141; CHECK-NEXT: store double [[TMP31]], ptr [[ARRAYIDX2_1]], align 8, !tbaa [[DOUBLE_TBAA3]]142; CHECK-NEXT: [[INDVARS_IV_NEXT_1:%.*]] = add nuw nsw i64 [[INDVARS_IV]], 2143; CHECK-NEXT: [[ARRAYIDX_2:%.*]] = getelementptr inbounds nuw double, ptr [[Y]], i64 [[INDVARS_IV_NEXT_1]]144; CHECK-NEXT: [[T0_2:%.*]] = load double, ptr [[ARRAYIDX_2]], align 8, !tbaa [[DOUBLE_TBAA3]]145; CHECK-NEXT: [[TMP32:%.*]] = fmul fast double [[T0_2]], [[TMP24]]146; CHECK-NEXT: [[ARRAYIDX2_2:%.*]] = getelementptr inbounds nuw double, ptr [[X]], i64 [[INDVARS_IV_NEXT_1]]147; CHECK-NEXT: store double [[TMP32]], ptr [[ARRAYIDX2_2]], align 8, !tbaa [[DOUBLE_TBAA3]]148; CHECK-NEXT: [[INDVARS_IV_NEXT_2:%.*]] = add nuw nsw i64 [[INDVARS_IV]], 3149; CHECK-NEXT: [[ARRAYIDX_3:%.*]] = getelementptr inbounds nuw double, ptr [[Y]], i64 [[INDVARS_IV_NEXT_2]]150; CHECK-NEXT: [[T0_3:%.*]] = load double, ptr [[ARRAYIDX_3]], align 8, !tbaa [[DOUBLE_TBAA3]]151; CHECK-NEXT: [[TMP33:%.*]] = fmul fast double [[T0_3]], [[TMP25]]152; CHECK-NEXT: [[ARRAYIDX2_3:%.*]] = getelementptr inbounds nuw double, ptr [[X]], i64 [[INDVARS_IV_NEXT_2]]153; CHECK-NEXT: store double [[TMP33]], ptr [[ARRAYIDX2_3]], align 8, !tbaa [[DOUBLE_TBAA3]]154; CHECK-NEXT: [[INDVARS_IV_NEXT_3:%.*]] = add nuw nsw i64 [[INDVARS_IV]], 4155; CHECK-NEXT: [[ARRAYIDX_4:%.*]] = getelementptr inbounds nuw double, ptr [[Y]], i64 [[INDVARS_IV_NEXT_3]]156; CHECK-NEXT: [[T0_4:%.*]] = load double, ptr [[ARRAYIDX_4]], align 8, !tbaa [[DOUBLE_TBAA3]]157; CHECK-NEXT: [[TMP34:%.*]] = fmul fast double [[T0_4]], [[TMP26]]158; CHECK-NEXT: [[ARRAYIDX2_4:%.*]] = getelementptr inbounds nuw double, ptr [[X]], i64 [[INDVARS_IV_NEXT_3]]159; CHECK-NEXT: store double [[TMP34]], ptr [[ARRAYIDX2_4]], align 8, !tbaa [[DOUBLE_TBAA3]]160; CHECK-NEXT: [[INDVARS_IV_NEXT_4:%.*]] = add nuw nsw i64 [[INDVARS_IV]], 5161; CHECK-NEXT: [[ARRAYIDX_5:%.*]] = getelementptr inbounds nuw double, ptr [[Y]], i64 [[INDVARS_IV_NEXT_4]]162; CHECK-NEXT: [[T0_5:%.*]] = load double, ptr [[ARRAYIDX_5]], align 8, !tbaa [[DOUBLE_TBAA3]]163; CHECK-NEXT: [[TMP35:%.*]] = fmul fast double [[T0_5]], [[TMP27]]164; CHECK-NEXT: [[ARRAYIDX2_5:%.*]] = getelementptr inbounds nuw double, ptr [[X]], i64 [[INDVARS_IV_NEXT_4]]165; CHECK-NEXT: store double [[TMP35]], ptr [[ARRAYIDX2_5]], align 8, !tbaa [[DOUBLE_TBAA3]]166; CHECK-NEXT: [[INDVARS_IV_NEXT_5:%.*]] = add nuw nsw i64 [[INDVARS_IV]], 6167; CHECK-NEXT: [[ARRAYIDX_6:%.*]] = getelementptr inbounds nuw double, ptr [[Y]], i64 [[INDVARS_IV_NEXT_5]]168; CHECK-NEXT: [[T0_6:%.*]] = load double, ptr [[ARRAYIDX_6]], align 8, !tbaa [[DOUBLE_TBAA3]]169; CHECK-NEXT: [[TMP36:%.*]] = fmul fast double [[T0_6]], [[TMP28]]170; CHECK-NEXT: [[ARRAYIDX2_6:%.*]] = getelementptr inbounds nuw double, ptr [[X]], i64 [[INDVARS_IV_NEXT_5]]171; CHECK-NEXT: store double [[TMP36]], ptr [[ARRAYIDX2_6]], align 8, !tbaa [[DOUBLE_TBAA3]]172; CHECK-NEXT: [[INDVARS_IV_NEXT_6:%.*]] = add nuw nsw i64 [[INDVARS_IV]], 7173; CHECK-NEXT: [[ARRAYIDX_7:%.*]] = getelementptr inbounds nuw double, ptr [[Y]], i64 [[INDVARS_IV_NEXT_6]]174; CHECK-NEXT: [[T0_7:%.*]] = load double, ptr [[ARRAYIDX_7]], align 8, !tbaa [[DOUBLE_TBAA3]]175; CHECK-NEXT: [[TMP37:%.*]] = fmul fast double [[T0_7]], [[TMP29]]176; CHECK-NEXT: [[ARRAYIDX2_7:%.*]] = getelementptr inbounds nuw double, ptr [[X]], i64 [[INDVARS_IV_NEXT_6]]177; CHECK-NEXT: store double [[TMP37]], ptr [[ARRAYIDX2_7]], align 8, !tbaa [[DOUBLE_TBAA3]]178; CHECK-NEXT: [[INDVARS_IV_NEXT_7]] = add nuw nsw i64 [[INDVARS_IV]], 8179; CHECK-NEXT: [[EXITCOND_NOT_7:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT_7]], [[WIDE_TRIP_COUNT]]180; CHECK-NEXT: br i1 [[EXITCOND_NOT_7]], label %[[FOR_END]], label %[[FOR_BODY]], !llvm.loop [[LOOP13:![0-9]+]]181; CHECK: [[FOR_END]]:182; CHECK-NEXT: ret void183;184entry:185 %div = fdiv fast double 1.0, %a186 br label %for.cond187 188for.cond:189 %n.0 = phi i32 [ 0, %entry ], [ %inc, %for.inc ]190 %cmp = icmp slt i32 %n.0, %N191 br i1 %cmp, label %for.body, label %for.cond.cleanup192 193for.cond.cleanup:194 br label %for.end195 196for.body:197 %idxprom = sext i32 %n.0 to i64198 %arrayidx = getelementptr inbounds double, ptr %y, i64 %idxprom199 %t0 = load double, ptr %arrayidx, align 8, !tbaa !3200 %mul = fmul fast double %t0, %div201 %idxprom1 = sext i32 %n.0 to i64202 %arrayidx2 = getelementptr inbounds double, ptr %x, i64 %idxprom1203 store double %mul, ptr %arrayidx2, align 8, !tbaa !3204 br label %for.inc205 206for.inc:207 %inc = add nsw i32 %n.0, 1208 br label %for.cond209 210for.end:211 ret void212}213 214attributes #0 = { nounwind ssp uwtable "correctly-rounded-divide-sqrt-fp-math"="false" "disable-tail-calls"="false" "frame-pointer"="all" "less-precise-fpmad"="false" "min-legal-vector-width"="0" "no-jump-tables"="false" "no-trapping-math"="false" "stack-protector-buffer-size"="8" "target-cpu"="skylake-avx512" "target-features"="+adx,+aes,+avx,+avx2,+avx512bw,+avx512cd,+avx512dq,+avx512f,+avx512vl,+bmi,+bmi2,+clflushopt,+clwb,+cx16,+cx8,+f16c,+fma,+fsgsbase,+fxsr,+invpcid,+lzcnt,+mmx,+movbe,+pclmul,+pku,+popcnt,+prfchw,+rdrnd,+rdseed,+sahf,+sse,+sse2,+sse3,+sse4.1,+sse4.2,+ssse3,+x87,+xsave,+xsavec,+xsaveopt,+xsaves" "use-soft-float"="false" }215 216!llvm.module.flags = !{!0, !1}217!llvm.ident = !{!2}218 219!0 = !{i32 1, !"wchar_size", i32 4}220!1 = !{i32 7, !"PIC Level", i32 2}221!2 = !{!"clang version 11.0.0 (https://github.com/llvm/llvm-project.git 45ebe38ffc40bb7221fc587bfb4481cf7f53ebbc)"}222!3 = !{!4, !4, i64 0}223!4 = !{!"double", !5, i64 0}224!5 = !{!"omnipotent char", !6, i64 0}225!6 = !{!"Simple C/C++ TBAA"}226 227;.228; CHECK: [[DOUBLE_TBAA3]] = !{[[META4:![0-9]+]], [[META4]], i64 0}229; CHECK: [[META4]] = !{!"double", [[META5:![0-9]+]], i64 0}230; CHECK: [[META5]] = !{!"omnipotent char", [[META6:![0-9]+]], i64 0}231; CHECK: [[META6]] = !{!"Simple C/C++ TBAA"}232; CHECK: [[LOOP7]] = distinct !{[[LOOP7]], [[META8:![0-9]+]], [[META9:![0-9]+]]}233; CHECK: [[META8]] = !{!"llvm.loop.isvectorized", i32 1}234; CHECK: [[META9]] = !{!"llvm.loop.unroll.runtime.disable"}235; CHECK: [[LOOP10]] = distinct !{[[LOOP10]], [[META8]], [[META9]]}236; CHECK: [[LOOP11]] = distinct !{[[LOOP11]], [[META12:![0-9]+]]}237; CHECK: [[META12]] = !{!"llvm.loop.unroll.disable"}238; CHECK: [[LOOP13]] = distinct !{[[LOOP13]], [[META8]]}239;.240