https://github.com/zahiraam created https://github.com/llvm/llvm-project/pull/184674
None >From 9ba8f5d495bada04be1e996f24212eb7cf5b22ea Mon Sep 17 00:00:00 2001 From: Ammarguellat <[email protected]> Date: Wed, 4 Mar 2026 12:06:37 -0800 Subject: [PATCH] [OpenMP] Enable collapse handling in triangular loops --- clang/lib/Sema/SemaOpenMP.cpp | 41 +- .../test/OpenMP/for_non_rectangular_codegen.c | 200 ++-- .../OpenMP/for_private_reduction_codegen.cpp | 80 +- clang/test/OpenMP/loop_collapse_codegen.cpp | 857 ++++++++++++++++++ 4 files changed, 1038 insertions(+), 140 deletions(-) create mode 100644 clang/test/OpenMP/loop_collapse_codegen.cpp diff --git a/clang/lib/Sema/SemaOpenMP.cpp b/clang/lib/Sema/SemaOpenMP.cpp index e90884a89bf6e..94c4993d151ec 100644 --- a/clang/lib/Sema/SemaOpenMP.cpp +++ b/clang/lib/Sema/SemaOpenMP.cpp @@ -8696,7 +8696,9 @@ static Expr * calculateNumIters(Sema &SemaRef, Scope *S, SourceLocation DefaultLoc, Expr *Lower, Expr *Upper, Expr *Step, QualType LCTy, bool TestIsStrictOp, bool RoundToStep, - llvm::MapVector<const Expr *, DeclRefExpr *> &Captures) { + llvm::MapVector<const Expr *, DeclRefExpr *> &Captures, + std::optional<unsigned> InitDependOnLC, + std::optional<unsigned> CondDependOnLC) { ExprResult NewStep = tryBuildCapture(SemaRef, Step, Captures, ".new_step"); if (!NewStep.isUsable()) return nullptr; @@ -8790,6 +8792,22 @@ calculateNumIters(Sema &SemaRef, Scope *S, SourceLocation DefaultLoc, return nullptr; ExprResult Diff; + + // For triangular loops, use already computed Upper and Lower bounds to + // calculate the number of iterations: Upper - Lower + 1. + if (TestIsStrictOp && InitDependOnLC.has_value() && + !CondDependOnLC.has_value()) { + Diff = SemaRef.BuildBinOp(S, DefaultLoc, BO_Sub, Upper, Lower); + if (!Diff.isUsable()) + return nullptr; + Diff = + SemaRef.BuildBinOp(S, DefaultLoc, BO_Add, Diff.get(), + SemaRef.ActOnIntegerConstant(DefaultLoc, 1).get()); + if (!Diff.isUsable()) + return nullptr; + return Diff.get(); + } + // If need to reorganize, then calculate the form as Upper - (Lower - Step [+ // 1]). if (NeedToReorganize) { @@ -9046,9 +9064,9 @@ Expr *OpenMPIterationSpaceChecker::buildNumIterations( if (!Upper || !Lower) return nullptr; - ExprResult Diff = calculateNumIters(SemaRef, S, DefaultLoc, Lower, Upper, - Step, VarType, TestIsStrictOp, - /*RoundToStep=*/true, Captures); + ExprResult Diff = calculateNumIters( + SemaRef, S, DefaultLoc, Lower, Upper, Step, VarType, TestIsStrictOp, + /*RoundToStep=*/true, Captures, InitDependOnLC, CondDependOnLC); if (!Diff.isUsable()) return nullptr; @@ -9117,7 +9135,6 @@ std::pair<Expr *, Expr *> OpenMPIterationSpaceChecker::buildMinMaxValues( UBNonRect ? UBExpr : tryBuildCapture(SemaRef, UBExpr, Captures).get(); if (!Upper || !Lower) return std::make_pair(nullptr, nullptr); - if (*TestIsLessOp) MinExpr = Lower; else @@ -9126,9 +9143,10 @@ std::pair<Expr *, Expr *> OpenMPIterationSpaceChecker::buildMinMaxValues( // Build minimum/maximum value based on number of iterations. QualType VarType = LCDecl->getType().getNonReferenceType(); - ExprResult Diff = calculateNumIters(SemaRef, S, DefaultLoc, Lower, Upper, - Step, VarType, TestIsStrictOp, - /*RoundToStep=*/false, Captures); + ExprResult Diff = calculateNumIters( + SemaRef, S, DefaultLoc, Lower, Upper, Step, VarType, TestIsStrictOp, + /*RoundToStep=*/false, Captures, InitDependOnLC, CondDependOnLC); + if (!Diff.isUsable()) return std::make_pair(nullptr, nullptr); @@ -9319,9 +9337,10 @@ Expr *OpenMPIterationSpaceChecker::buildOrderedLoopData( if (!Upper || !Lower) return nullptr; - ExprResult Diff = calculateNumIters( - SemaRef, S, DefaultLoc, Lower, Upper, Step, VarType, - /*TestIsStrictOp=*/false, /*RoundToStep=*/false, Captures); + ExprResult Diff = + calculateNumIters(SemaRef, S, DefaultLoc, Lower, Upper, Step, VarType, + /*TestIsStrictOp=*/false, /*RoundToStep=*/false, + Captures, InitDependOnLC, CondDependOnLC); if (!Diff.isUsable()) return nullptr; diff --git a/clang/test/OpenMP/for_non_rectangular_codegen.c b/clang/test/OpenMP/for_non_rectangular_codegen.c index 48dd37f6eddc0..b35a8f6e27b50 100644 --- a/clang/test/OpenMP/for_non_rectangular_codegen.c +++ b/clang/test/OpenMP/for_non_rectangular_codegen.c @@ -40,14 +40,14 @@ void collapsed(int mp) { // CHECK-NEXT: [[J:%.*]] = alloca i32, align 4 // CHECK-NEXT: [[I:%.*]] = alloca i32, align 4 // CHECK-NEXT: [[I0:%.*]] = alloca i32, align 4 -// CHECK-NEXT: [[_TMP15:%.*]] = alloca i32, align 4 +// CHECK-NEXT: [[_TMP13:%.*]] = alloca i32, align 4 // CHECK-NEXT: [[DOTOMP_LB:%.*]] = alloca i64, align 8 // CHECK-NEXT: [[DOTOMP_UB:%.*]] = alloca i64, align 8 // CHECK-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8 // CHECK-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4 -// CHECK-NEXT: [[J19:%.*]] = alloca i32, align 4 -// CHECK-NEXT: [[I20:%.*]] = alloca i32, align 4 -// CHECK-NEXT: [[I021:%.*]] = alloca i32, align 4 +// CHECK-NEXT: [[J17:%.*]] = alloca i32, align 4 +// CHECK-NEXT: [[I18:%.*]] = alloca i32, align 4 +// CHECK-NEXT: [[I019:%.*]] = alloca i32, align 4 // CHECK-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB2:[0-9]+]]) // CHECK-NEXT: store i32 [[MP]], ptr [[MP_ADDR]], align 4 // CHECK-NEXT: [[TMP1:%.*]] = load i32, ptr [[MP_ADDR]], align 4 @@ -90,22 +90,20 @@ void collapsed(int mp) { // CHECK-NEXT: [[TMP12:%.*]] = load i32, ptr [[DOTUPPER]], align 4 // CHECK-NEXT: [[TMP13:%.*]] = load i32, ptr [[DOTLOWER]], align 4 // CHECK-NEXT: [[SUB7:%.*]] = sub i32 [[TMP12]], [[TMP13]] -// CHECK-NEXT: [[SUB8:%.*]] = sub i32 [[SUB7]], 1 -// CHECK-NEXT: [[ADD9:%.*]] = add i32 [[SUB8]], 1 -// CHECK-NEXT: [[DIV10:%.*]] = udiv i32 [[ADD9]], 1 -// CHECK-NEXT: [[CONV11:%.*]] = zext i32 [[DIV10]] to i64 -// CHECK-NEXT: [[MUL12:%.*]] = mul nsw i64 [[CONV6]], [[CONV11]] -// CHECK-NEXT: [[MUL13:%.*]] = mul nsw i64 [[MUL12]], 10 -// CHECK-NEXT: [[SUB14:%.*]] = sub nsw i64 [[MUL13]], 1 -// CHECK-NEXT: store i64 [[SUB14]], ptr [[DOTCAPTURE_EXPR_3]], align 8 +// CHECK-NEXT: [[ADD8:%.*]] = add i32 [[SUB7]], 1 +// CHECK-NEXT: [[CONV9:%.*]] = zext i32 [[ADD8]] to i64 +// CHECK-NEXT: [[MUL10:%.*]] = mul nsw i64 [[CONV6]], [[CONV9]] +// CHECK-NEXT: [[MUL11:%.*]] = mul nsw i64 [[MUL10]], 10 +// CHECK-NEXT: [[SUB12:%.*]] = sub nsw i64 [[MUL11]], 1 +// CHECK-NEXT: store i64 [[SUB12]], ptr [[DOTCAPTURE_EXPR_3]], align 8 // CHECK-NEXT: store i32 0, ptr [[J]], align 4 // CHECK-NEXT: [[TMP14:%.*]] = load i32, ptr [[J]], align 4 // CHECK-NEXT: store i32 [[TMP14]], ptr [[I]], align 4 // CHECK-NEXT: store i32 0, ptr [[I0]], align 4 -// CHECK-NEXT: store i32 0, ptr [[_TMP15]], align 4 +// CHECK-NEXT: store i32 0, ptr [[_TMP13]], align 4 // CHECK-NEXT: [[TMP15:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4 -// CHECK-NEXT: [[CMP16:%.*]] = icmp slt i32 0, [[TMP15]] -// CHECK-NEXT: br i1 [[CMP16]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]] +// CHECK-NEXT: [[CMP14:%.*]] = icmp slt i32 0, [[TMP15]] +// CHECK-NEXT: br i1 [[CMP14]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]] // CHECK: omp.precond.then: // CHECK-NEXT: store i64 0, ptr [[DOTOMP_LB]], align 8 // CHECK-NEXT: [[TMP16:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_3]], align 8 @@ -115,136 +113,122 @@ void collapsed(int mp) { // CHECK-NEXT: call void @__kmpc_for_static_init_8(ptr @[[GLOB1:[0-9]+]], i32 [[TMP0]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i64 1, i64 1) // CHECK-NEXT: [[TMP17:%.*]] = load i64, ptr [[DOTOMP_UB]], align 8 // CHECK-NEXT: [[TMP18:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_3]], align 8 -// CHECK-NEXT: [[CMP22:%.*]] = icmp sgt i64 [[TMP17]], [[TMP18]] -// CHECK-NEXT: br i1 [[CMP22]], label [[COND_TRUE24:%.*]], label [[COND_FALSE25:%.*]] -// CHECK: cond.true24: +// CHECK-NEXT: [[CMP20:%.*]] = icmp sgt i64 [[TMP17]], [[TMP18]] +// CHECK-NEXT: br i1 [[CMP20]], label [[COND_TRUE22:%.*]], label [[COND_FALSE23:%.*]] +// CHECK: cond.true22: // CHECK-NEXT: [[TMP19:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_3]], align 8 -// CHECK-NEXT: br label [[COND_END26:%.*]] -// CHECK: cond.false25: +// CHECK-NEXT: br label [[COND_END24:%.*]] +// CHECK: cond.false23: // CHECK-NEXT: [[TMP20:%.*]] = load i64, ptr [[DOTOMP_UB]], align 8 -// CHECK-NEXT: br label [[COND_END26]] -// CHECK: cond.end26: -// CHECK-NEXT: [[COND27:%.*]] = phi i64 [ [[TMP19]], [[COND_TRUE24]] ], [ [[TMP20]], [[COND_FALSE25]] ] -// CHECK-NEXT: store i64 [[COND27]], ptr [[DOTOMP_UB]], align 8 +// CHECK-NEXT: br label [[COND_END24]] +// CHECK: cond.end24: +// CHECK-NEXT: [[COND25:%.*]] = phi i64 [ [[TMP19]], [[COND_TRUE22]] ], [ [[TMP20]], [[COND_FALSE23]] ] +// CHECK-NEXT: store i64 [[COND25]], ptr [[DOTOMP_UB]], align 8 // CHECK-NEXT: [[TMP21:%.*]] = load i64, ptr [[DOTOMP_LB]], align 8 // CHECK-NEXT: store i64 [[TMP21]], ptr [[DOTOMP_IV]], align 8 // CHECK-NEXT: br label [[OMP_INNER_FOR_COND:%.*]] // CHECK: omp.inner.for.cond: // CHECK-NEXT: [[TMP22:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8 // CHECK-NEXT: [[TMP23:%.*]] = load i64, ptr [[DOTOMP_UB]], align 8 -// CHECK-NEXT: [[CMP28:%.*]] = icmp sle i64 [[TMP22]], [[TMP23]] -// CHECK-NEXT: br i1 [[CMP28]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] +// CHECK-NEXT: [[CMP26:%.*]] = icmp sle i64 [[TMP22]], [[TMP23]] +// CHECK-NEXT: br i1 [[CMP26]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] // CHECK: omp.inner.for.body: // CHECK-NEXT: [[TMP24:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8 // CHECK-NEXT: [[TMP25:%.*]] = load i32, ptr [[DOTUPPER]], align 4 // CHECK-NEXT: [[TMP26:%.*]] = load i32, ptr [[DOTLOWER]], align 4 -// CHECK-NEXT: [[SUB30:%.*]] = sub i32 [[TMP25]], [[TMP26]] -// CHECK-NEXT: [[SUB31:%.*]] = sub i32 [[SUB30]], 1 -// CHECK-NEXT: [[ADD32:%.*]] = add i32 [[SUB31]], 1 -// CHECK-NEXT: [[DIV33:%.*]] = udiv i32 [[ADD32]], 1 -// CHECK-NEXT: [[MUL34:%.*]] = mul i32 1, [[DIV33]] -// CHECK-NEXT: [[MUL35:%.*]] = mul i32 [[MUL34]], 10 -// CHECK-NEXT: [[CONV36:%.*]] = zext i32 [[MUL35]] to i64 -// CHECK-NEXT: [[DIV37:%.*]] = sdiv i64 [[TMP24]], [[CONV36]] -// CHECK-NEXT: [[MUL38:%.*]] = mul nsw i64 [[DIV37]], 1 -// CHECK-NEXT: [[ADD39:%.*]] = add nsw i64 0, [[MUL38]] -// CHECK-NEXT: [[CONV40:%.*]] = trunc i64 [[ADD39]] to i32 -// CHECK-NEXT: store i32 [[CONV40]], ptr [[J19]], align 4 -// CHECK-NEXT: [[TMP27:%.*]] = load i32, ptr [[J19]], align 4 -// CHECK-NEXT: [[CONV41:%.*]] = sext i32 [[TMP27]] to i64 +// CHECK-NEXT: [[SUB28:%.*]] = sub i32 [[TMP25]], [[TMP26]] +// CHECK-NEXT: [[ADD29:%.*]] = add i32 [[SUB28]], 1 +// CHECK-NEXT: [[MUL30:%.*]] = mul i32 1, [[ADD29]] +// CHECK-NEXT: [[MUL31:%.*]] = mul i32 [[MUL30]], 10 +// CHECK-NEXT: [[CONV32:%.*]] = zext i32 [[MUL31]] to i64 +// CHECK-NEXT: [[DIV33:%.*]] = sdiv i64 [[TMP24]], [[CONV32]] +// CHECK-NEXT: [[MUL34:%.*]] = mul nsw i64 [[DIV33]], 1 +// CHECK-NEXT: [[ADD35:%.*]] = add nsw i64 0, [[MUL34]] +// CHECK-NEXT: [[CONV36:%.*]] = trunc i64 [[ADD35]] to i32 +// CHECK-NEXT: store i32 [[CONV36]], ptr [[J17]], align 4 +// CHECK-NEXT: [[TMP27:%.*]] = load i32, ptr [[J17]], align 4 +// CHECK-NEXT: [[CONV37:%.*]] = sext i32 [[TMP27]] to i64 // CHECK-NEXT: [[TMP28:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8 // CHECK-NEXT: [[TMP29:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8 // CHECK-NEXT: [[TMP30:%.*]] = load i32, ptr [[DOTUPPER]], align 4 // CHECK-NEXT: [[TMP31:%.*]] = load i32, ptr [[DOTLOWER]], align 4 -// CHECK-NEXT: [[SUB42:%.*]] = sub i32 [[TMP30]], [[TMP31]] -// CHECK-NEXT: [[SUB43:%.*]] = sub i32 [[SUB42]], 1 -// CHECK-NEXT: [[ADD44:%.*]] = add i32 [[SUB43]], 1 -// CHECK-NEXT: [[DIV45:%.*]] = udiv i32 [[ADD44]], 1 -// CHECK-NEXT: [[MUL46:%.*]] = mul i32 1, [[DIV45]] -// CHECK-NEXT: [[MUL47:%.*]] = mul i32 [[MUL46]], 10 -// CHECK-NEXT: [[CONV48:%.*]] = zext i32 [[MUL47]] to i64 -// CHECK-NEXT: [[DIV49:%.*]] = sdiv i64 [[TMP29]], [[CONV48]] +// CHECK-NEXT: [[SUB38:%.*]] = sub i32 [[TMP30]], [[TMP31]] +// CHECK-NEXT: [[ADD39:%.*]] = add i32 [[SUB38]], 1 +// CHECK-NEXT: [[MUL40:%.*]] = mul i32 1, [[ADD39]] +// CHECK-NEXT: [[MUL41:%.*]] = mul i32 [[MUL40]], 10 +// CHECK-NEXT: [[CONV42:%.*]] = zext i32 [[MUL41]] to i64 +// CHECK-NEXT: [[DIV43:%.*]] = sdiv i64 [[TMP29]], [[CONV42]] // CHECK-NEXT: [[TMP32:%.*]] = load i32, ptr [[DOTUPPER]], align 4 // CHECK-NEXT: [[TMP33:%.*]] = load i32, ptr [[DOTLOWER]], align 4 -// CHECK-NEXT: [[SUB50:%.*]] = sub i32 [[TMP32]], [[TMP33]] -// CHECK-NEXT: [[SUB51:%.*]] = sub i32 [[SUB50]], 1 -// CHECK-NEXT: [[ADD52:%.*]] = add i32 [[SUB51]], 1 -// CHECK-NEXT: [[DIV53:%.*]] = udiv i32 [[ADD52]], 1 -// CHECK-NEXT: [[MUL54:%.*]] = mul i32 1, [[DIV53]] -// CHECK-NEXT: [[MUL55:%.*]] = mul i32 [[MUL54]], 10 -// CHECK-NEXT: [[CONV56:%.*]] = zext i32 [[MUL55]] to i64 -// CHECK-NEXT: [[MUL57:%.*]] = mul nsw i64 [[DIV49]], [[CONV56]] -// CHECK-NEXT: [[SUB58:%.*]] = sub nsw i64 [[TMP28]], [[MUL57]] -// CHECK-NEXT: [[DIV59:%.*]] = sdiv i64 [[SUB58]], 10 -// CHECK-NEXT: [[MUL60:%.*]] = mul nsw i64 [[DIV59]], 1 -// CHECK-NEXT: [[ADD61:%.*]] = add nsw i64 [[CONV41]], [[MUL60]] -// CHECK-NEXT: [[CONV62:%.*]] = trunc i64 [[ADD61]] to i32 -// CHECK-NEXT: store i32 [[CONV62]], ptr [[I20]], align 4 +// CHECK-NEXT: [[SUB44:%.*]] = sub i32 [[TMP32]], [[TMP33]] +// CHECK-NEXT: [[ADD45:%.*]] = add i32 [[SUB44]], 1 +// CHECK-NEXT: [[MUL46:%.*]] = mul i32 1, [[ADD45]] +// CHECK-NEXT: [[MUL47:%.*]] = mul i32 [[MUL46]], 10 +// CHECK-NEXT: [[CONV48:%.*]] = zext i32 [[MUL47]] to i64 +// CHECK-NEXT: [[MUL49:%.*]] = mul nsw i64 [[DIV43]], [[CONV48]] +// CHECK-NEXT: [[SUB50:%.*]] = sub nsw i64 [[TMP28]], [[MUL49]] +// CHECK-NEXT: [[DIV51:%.*]] = sdiv i64 [[SUB50]], 10 +// CHECK-NEXT: [[MUL52:%.*]] = mul nsw i64 [[DIV51]], 1 +// CHECK-NEXT: [[ADD53:%.*]] = add nsw i64 [[CONV37]], [[MUL52]] +// CHECK-NEXT: [[CONV54:%.*]] = trunc i64 [[ADD53]] to i32 +// CHECK-NEXT: store i32 [[CONV54]], ptr [[I18]], align 4 // CHECK-NEXT: [[TMP34:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8 // CHECK-NEXT: [[TMP35:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8 // CHECK-NEXT: [[TMP36:%.*]] = load i32, ptr [[DOTUPPER]], align 4 // CHECK-NEXT: [[TMP37:%.*]] = load i32, ptr [[DOTLOWER]], align 4 -// CHECK-NEXT: [[SUB63:%.*]] = sub i32 [[TMP36]], [[TMP37]] -// CHECK-NEXT: [[SUB64:%.*]] = sub i32 [[SUB63]], 1 -// CHECK-NEXT: [[ADD65:%.*]] = add i32 [[SUB64]], 1 -// CHECK-NEXT: [[DIV66:%.*]] = udiv i32 [[ADD65]], 1 -// CHECK-NEXT: [[MUL67:%.*]] = mul i32 1, [[DIV66]] -// CHECK-NEXT: [[MUL68:%.*]] = mul i32 [[MUL67]], 10 -// CHECK-NEXT: [[CONV69:%.*]] = zext i32 [[MUL68]] to i64 -// CHECK-NEXT: [[DIV70:%.*]] = sdiv i64 [[TMP35]], [[CONV69]] +// CHECK-NEXT: [[SUB55:%.*]] = sub i32 [[TMP36]], [[TMP37]] +// CHECK-NEXT: [[ADD56:%.*]] = add i32 [[SUB55]], 1 +// CHECK-NEXT: [[MUL57:%.*]] = mul i32 1, [[ADD56]] +// CHECK-NEXT: [[MUL58:%.*]] = mul i32 [[MUL57]], 10 +// CHECK-NEXT: [[CONV59:%.*]] = zext i32 [[MUL58]] to i64 +// CHECK-NEXT: [[DIV60:%.*]] = sdiv i64 [[TMP35]], [[CONV59]] // CHECK-NEXT: [[TMP38:%.*]] = load i32, ptr [[DOTUPPER]], align 4 // CHECK-NEXT: [[TMP39:%.*]] = load i32, ptr [[DOTLOWER]], align 4 -// CHECK-NEXT: [[SUB71:%.*]] = sub i32 [[TMP38]], [[TMP39]] -// CHECK-NEXT: [[SUB72:%.*]] = sub i32 [[SUB71]], 1 -// CHECK-NEXT: [[ADD73:%.*]] = add i32 [[SUB72]], 1 -// CHECK-NEXT: [[DIV74:%.*]] = udiv i32 [[ADD73]], 1 -// CHECK-NEXT: [[MUL75:%.*]] = mul i32 1, [[DIV74]] -// CHECK-NEXT: [[MUL76:%.*]] = mul i32 [[MUL75]], 10 -// CHECK-NEXT: [[CONV77:%.*]] = zext i32 [[MUL76]] to i64 -// CHECK-NEXT: [[MUL78:%.*]] = mul nsw i64 [[DIV70]], [[CONV77]] -// CHECK-NEXT: [[SUB79:%.*]] = sub nsw i64 [[TMP34]], [[MUL78]] +// CHECK-NEXT: [[SUB61:%.*]] = sub i32 [[TMP38]], [[TMP39]] +// CHECK-NEXT: [[ADD62:%.*]] = add i32 [[SUB61]], 1 +// CHECK-NEXT: [[MUL63:%.*]] = mul i32 1, [[ADD62]] +// CHECK-NEXT: [[MUL64:%.*]] = mul i32 [[MUL63]], 10 +// CHECK-NEXT: [[CONV65:%.*]] = zext i32 [[MUL64]] to i64 +// CHECK-NEXT: [[MUL66:%.*]] = mul nsw i64 [[DIV60]], [[CONV65]] +// CHECK-NEXT: [[SUB67:%.*]] = sub nsw i64 [[TMP34]], [[MUL66]] // CHECK-NEXT: [[TMP40:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8 // CHECK-NEXT: [[TMP41:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8 // CHECK-NEXT: [[TMP42:%.*]] = load i32, ptr [[DOTUPPER]], align 4 // CHECK-NEXT: [[TMP43:%.*]] = load i32, ptr [[DOTLOWER]], align 4 -// CHECK-NEXT: [[SUB80:%.*]] = sub i32 [[TMP42]], [[TMP43]] -// CHECK-NEXT: [[SUB81:%.*]] = sub i32 [[SUB80]], 1 -// CHECK-NEXT: [[ADD82:%.*]] = add i32 [[SUB81]], 1 -// CHECK-NEXT: [[DIV83:%.*]] = udiv i32 [[ADD82]], 1 -// CHECK-NEXT: [[MUL84:%.*]] = mul i32 1, [[DIV83]] -// CHECK-NEXT: [[MUL85:%.*]] = mul i32 [[MUL84]], 10 -// CHECK-NEXT: [[CONV86:%.*]] = zext i32 [[MUL85]] to i64 -// CHECK-NEXT: [[DIV87:%.*]] = sdiv i64 [[TMP41]], [[CONV86]] +// CHECK-NEXT: [[SUB68:%.*]] = sub i32 [[TMP42]], [[TMP43]] +// CHECK-NEXT: [[ADD69:%.*]] = add i32 [[SUB68]], 1 +// CHECK-NEXT: [[MUL70:%.*]] = mul i32 1, [[ADD69]] +// CHECK-NEXT: [[MUL71:%.*]] = mul i32 [[MUL70]], 10 +// CHECK-NEXT: [[CONV72:%.*]] = zext i32 [[MUL71]] to i64 +// CHECK-NEXT: [[DIV73:%.*]] = sdiv i64 [[TMP41]], [[CONV72]] // CHECK-NEXT: [[TMP44:%.*]] = load i32, ptr [[DOTUPPER]], align 4 // CHECK-NEXT: [[TMP45:%.*]] = load i32, ptr [[DOTLOWER]], align 4 -// CHECK-NEXT: [[SUB88:%.*]] = sub i32 [[TMP44]], [[TMP45]] -// CHECK-NEXT: [[SUB89:%.*]] = sub i32 [[SUB88]], 1 -// CHECK-NEXT: [[ADD90:%.*]] = add i32 [[SUB89]], 1 -// CHECK-NEXT: [[DIV91:%.*]] = udiv i32 [[ADD90]], 1 -// CHECK-NEXT: [[MUL92:%.*]] = mul i32 1, [[DIV91]] -// CHECK-NEXT: [[MUL93:%.*]] = mul i32 [[MUL92]], 10 -// CHECK-NEXT: [[CONV94:%.*]] = zext i32 [[MUL93]] to i64 -// CHECK-NEXT: [[MUL95:%.*]] = mul nsw i64 [[DIV87]], [[CONV94]] -// CHECK-NEXT: [[SUB96:%.*]] = sub nsw i64 [[TMP40]], [[MUL95]] -// CHECK-NEXT: [[DIV97:%.*]] = sdiv i64 [[SUB96]], 10 -// CHECK-NEXT: [[MUL98:%.*]] = mul nsw i64 [[DIV97]], 10 -// CHECK-NEXT: [[SUB99:%.*]] = sub nsw i64 [[SUB79]], [[MUL98]] -// CHECK-NEXT: [[MUL100:%.*]] = mul nsw i64 [[SUB99]], 1 -// CHECK-NEXT: [[ADD101:%.*]] = add nsw i64 0, [[MUL100]] -// CHECK-NEXT: [[CONV102:%.*]] = trunc i64 [[ADD101]] to i32 -// CHECK-NEXT: store i32 [[CONV102]], ptr [[I021]], align 4 -// CHECK-NEXT: [[TMP46:%.*]] = load i32, ptr [[I20]], align 4 +// CHECK-NEXT: [[SUB74:%.*]] = sub i32 [[TMP44]], [[TMP45]] +// CHECK-NEXT: [[ADD75:%.*]] = add i32 [[SUB74]], 1 +// CHECK-NEXT: [[MUL76:%.*]] = mul i32 1, [[ADD75]] +// CHECK-NEXT: [[MUL77:%.*]] = mul i32 [[MUL76]], 10 +// CHECK-NEXT: [[CONV78:%.*]] = zext i32 [[MUL77]] to i64 +// CHECK-NEXT: [[MUL79:%.*]] = mul nsw i64 [[DIV73]], [[CONV78]] +// CHECK-NEXT: [[SUB80:%.*]] = sub nsw i64 [[TMP40]], [[MUL79]] +// CHECK-NEXT: [[DIV81:%.*]] = sdiv i64 [[SUB80]], 10 +// CHECK-NEXT: [[MUL82:%.*]] = mul nsw i64 [[DIV81]], 10 +// CHECK-NEXT: [[SUB83:%.*]] = sub nsw i64 [[SUB67]], [[MUL82]] +// CHECK-NEXT: [[MUL84:%.*]] = mul nsw i64 [[SUB83]], 1 +// CHECK-NEXT: [[ADD85:%.*]] = add nsw i64 0, [[MUL84]] +// CHECK-NEXT: [[CONV86:%.*]] = trunc i64 [[ADD85]] to i32 +// CHECK-NEXT: store i32 [[CONV86]], ptr [[I019]], align 4 +// CHECK-NEXT: [[TMP46:%.*]] = load i32, ptr [[I18]], align 4 // CHECK-NEXT: [[TMP47:%.*]] = load i32, ptr [[MP_ADDR]], align 4 -// CHECK-NEXT: [[CMP103:%.*]] = icmp slt i32 [[TMP46]], [[TMP47]] -// CHECK-NEXT: br i1 [[CMP103]], label [[OMP_BODY_NEXT:%.*]], label [[OMP_BODY_CONTINUE:%.*]] +// CHECK-NEXT: [[CMP87:%.*]] = icmp slt i32 [[TMP46]], [[TMP47]] +// CHECK-NEXT: br i1 [[CMP87]], label [[OMP_BODY_NEXT:%.*]], label [[OMP_BODY_CONTINUE:%.*]] // CHECK: omp.body.next: // CHECK-NEXT: br label [[OMP_BODY_CONTINUE]] // CHECK: omp.body.continue: // CHECK-NEXT: br label [[OMP_INNER_FOR_INC:%.*]] // CHECK: omp.inner.for.inc: // CHECK-NEXT: [[TMP48:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8 -// CHECK-NEXT: [[ADD105:%.*]] = add nsw i64 [[TMP48]], 1 -// CHECK-NEXT: store i64 [[ADD105]], ptr [[DOTOMP_IV]], align 8 +// CHECK-NEXT: [[ADD89:%.*]] = add nsw i64 [[TMP48]], 1 +// CHECK-NEXT: store i64 [[ADD89]], ptr [[DOTOMP_IV]], align 8 // CHECK-NEXT: br label [[OMP_INNER_FOR_COND]] // CHECK: omp.inner.for.end: // CHECK-NEXT: br label [[OMP_LOOP_EXIT:%.*]] diff --git a/clang/test/OpenMP/for_private_reduction_codegen.cpp b/clang/test/OpenMP/for_private_reduction_codegen.cpp index c8a6863299fb3..85ff39e3a44f6 100644 --- a/clang/test/OpenMP/for_private_reduction_codegen.cpp +++ b/clang/test/OpenMP/for_private_reduction_codegen.cpp @@ -69,6 +69,7 @@ int main(void) { return 0; } + //. // CHECK: @.omp.reduction..internal_pivate_.result.result_996 = common global %class.Sum zeroinitializer, align 4 // CHECK: @.omp.reduction..internal_pivate_.sum_v.sum_v_1188 = common global i32 0, align 4 @@ -79,12 +80,12 @@ int main(void) { // CHECK-SAME: () #[[ATTR0:[0-9]+]] { // CHECK-NEXT: entry: // CHECK-NEXT: [[RESULT:%.*]] = alloca [[CLASS_SUM:%.*]], align 4 -// CHECK-NEXT: [[ARRAY:%.*]] = alloca [10 x %class.Sum], align 16 +// CHECK-NEXT: [[ARRAY:%.*]] = alloca [10 x [[CLASS_SUM]]], align 16 // CHECK-NEXT: [[I:%.*]] = alloca i32, align 4 // CHECK-NEXT: [[REF_TMP:%.*]] = alloca [[CLASS_SUM]], align 4 // CHECK-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB3:[0-9]+]]) // CHECK-NEXT: call void @_ZN3SumC1Ei(ptr noundef nonnull align 4 dereferenceable(4) [[RESULT]], i32 noundef 0) -// CHECK-NEXT: [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [10 x %class.Sum], ptr [[ARRAY]], i32 0, i32 0 +// CHECK-NEXT: [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [10 x [[CLASS_SUM]]], ptr [[ARRAY]], i32 0, i32 0 // CHECK-NEXT: [[ARRAYCTOR_END:%.*]] = getelementptr inbounds [[CLASS_SUM]], ptr [[ARRAY_BEGIN]], i64 10 // CHECK-NEXT: br label [[ARRAYCTOR_LOOP:%.*]] // CHECK: arrayctor.loop: @@ -105,7 +106,7 @@ int main(void) { // CHECK-NEXT: call void @_ZN3SumC1Ei(ptr noundef nonnull align 4 dereferenceable(4) [[REF_TMP]], i32 noundef [[TMP2]]) // CHECK-NEXT: [[TMP3:%.*]] = load i32, ptr [[I]], align 4 // CHECK-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP3]] to i64 -// CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x %class.Sum], ptr [[ARRAY]], i64 0, i64 [[IDXPROM]] +// CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x [[CLASS_SUM]]], ptr [[ARRAY]], i64 0, i64 [[IDXPROM]] // CHECK-NEXT: call void @llvm.memcpy.p0.p0.i64(ptr align 4 [[ARRAYIDX]], ptr align 4 [[REF_TMP]], i64 4, i1 false) // CHECK-NEXT: br label [[FOR_INC:%.*]] // CHECK: for.inc: @@ -152,7 +153,7 @@ int main(void) { // CHECK-NEXT: store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8 // CHECK-NEXT: store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8 // CHECK-NEXT: store ptr [[ARRAY]], ptr [[ARRAY_ADDR]], align 8 -// CHECK-NEXT: [[TMP0:%.*]] = load ptr, ptr [[ARRAY_ADDR]], align 8 +// CHECK-NEXT: [[TMP0:%.*]] = load ptr, ptr [[ARRAY_ADDR]], align 8, !nonnull [[META5:![0-9]+]], !align [[META6:![0-9]+]] // CHECK-NEXT: call void @_ZN3SumC1Ei(ptr noundef nonnull align 4 dereferenceable(4) [[RESULT]], i32 noundef 0) // CHECK-NEXT: store i32 0, ptr [[DOTOMP_LB]], align 4 // CHECK-NEXT: store i32 9, ptr [[DOTOMP_UB]], align 4 @@ -188,7 +189,7 @@ int main(void) { // CHECK-NEXT: store i32 [[ADD]], ptr [[I]], align 4 // CHECK-NEXT: [[TMP9:%.*]] = load i32, ptr [[I]], align 4 // CHECK-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP9]] to i64 -// CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x %class.Sum], ptr [[TMP0]], i64 0, i64 [[IDXPROM]] +// CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x [[CLASS_SUM]]], ptr [[TMP0]], i64 0, i64 [[IDXPROM]] // CHECK-NEXT: [[CALL:%.*]] = call i32 @_ZNK3SumplERKS_(ptr noundef nonnull align 4 dereferenceable(4) [[RESULT1]], ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYIDX]]) // CHECK-NEXT: [[COERCE_DIVE:%.*]] = getelementptr inbounds nuw [[CLASS_SUM]], ptr [[REF_TMP]], i32 0, i32 0 // CHECK-NEXT: store i32 [[CALL]], ptr [[COERCE_DIVE]], align 4 @@ -259,7 +260,7 @@ int main(void) { // CHECK-NEXT: store ptr [[THIS]], ptr [[THIS_ADDR]], align 8 // CHECK-NEXT: store ptr [[RHS]], ptr [[RHS_ADDR]], align 8 // CHECK-NEXT: [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8 -// CHECK-NEXT: [[TMP0:%.*]] = load ptr, ptr [[RHS_ADDR]], align 8 +// CHECK-NEXT: [[TMP0:%.*]] = load ptr, ptr [[RHS_ADDR]], align 8, !nonnull [[META5]], !align [[META6]] // CHECK-NEXT: [[VAL:%.*]] = getelementptr inbounds nuw [[CLASS_SUM:%.*]], ptr [[TMP0]], i32 0, i32 0 // CHECK-NEXT: [[TMP1:%.*]] = load i32, ptr [[VAL]], align 4 // CHECK-NEXT: [[VAL2:%.*]] = getelementptr inbounds nuw [[CLASS_SUM]], ptr [[THIS1]], i32 0, i32 0 @@ -293,7 +294,7 @@ int main(void) { // CHECK-NEXT: [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8 // CHECK-NEXT: [[VAL:%.*]] = getelementptr inbounds nuw [[CLASS_SUM]], ptr [[THIS1]], i32 0, i32 0 // CHECK-NEXT: [[TMP0:%.*]] = load i32, ptr [[VAL]], align 4 -// CHECK-NEXT: [[TMP1:%.*]] = load ptr, ptr [[RHS_ADDR]], align 8 +// CHECK-NEXT: [[TMP1:%.*]] = load ptr, ptr [[RHS_ADDR]], align 8, !nonnull [[META5]], !align [[META6]] // CHECK-NEXT: [[VAL2:%.*]] = getelementptr inbounds nuw [[CLASS_SUM]], ptr [[TMP1]], i32 0, i32 0 // CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr [[VAL2]], align 4 // CHECK-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP0]], [[TMP2]] @@ -353,9 +354,9 @@ int main(void) { // CHECK-NEXT: store i32 [[N]], ptr [[N_ADDR]], align 4 // CHECK-NEXT: store ptr [[V]], ptr [[V_ADDR]], align 8 // CHECK-NEXT: store ptr [[SUM_V]], ptr [[SUM_V_ADDR]], align 8 -// CHECK-NEXT: [[TMP1:%.*]] = load ptr, ptr [[SUM_V_ADDR]], align 8 +// CHECK-NEXT: [[TMP1:%.*]] = load ptr, ptr [[SUM_V_ADDR]], align 8, !nonnull [[META5]], !align [[META6]] // CHECK-NEXT: store i32 0, ptr [[TMP1]], align 4 -// CHECK-NEXT: [[TMP2:%.*]] = load ptr, ptr [[SUM_V_ADDR]], align 8 +// CHECK-NEXT: [[TMP2:%.*]] = load ptr, ptr [[SUM_V_ADDR]], align 8, !nonnull [[META5]], !align [[META6]] // CHECK-NEXT: store ptr [[TMP2]], ptr [[TMP]], align 8 // CHECK-NEXT: [[TMP3:%.*]] = load i32, ptr [[N_ADDR]], align 4 // CHECK-NEXT: store i32 [[TMP3]], ptr [[DOTCAPTURE_EXPR_]], align 4 @@ -374,7 +375,7 @@ int main(void) { // CHECK-NEXT: store i32 [[TMP6]], ptr [[DOTOMP_UB]], align 4 // CHECK-NEXT: store i32 1, ptr [[DOTOMP_STRIDE]], align 4 // CHECK-NEXT: store i32 0, ptr [[DOTOMP_IS_LAST]], align 4 -// CHECK-NEXT: [[TMP7:%.*]] = load ptr, ptr [[TMP]], align 8 +// CHECK-NEXT: [[TMP7:%.*]] = load ptr, ptr [[TMP]], align 8, !nonnull [[META5]], !align [[META6]] // CHECK-NEXT: store i32 0, ptr [[SUM_V4]], align 4 // CHECK-NEXT: store ptr [[SUM_V4]], ptr [[_TMP5]], align 8 // CHECK-NEXT: call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP0]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1) @@ -409,7 +410,7 @@ int main(void) { // CHECK-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP17]] to i64 // CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[TMP16]], i64 [[IDXPROM]] // CHECK-NEXT: [[TMP18:%.*]] = load i32, ptr [[ARRAYIDX]], align 4 -// CHECK-NEXT: [[TMP19:%.*]] = load ptr, ptr [[_TMP5]], align 8 +// CHECK-NEXT: [[TMP19:%.*]] = load ptr, ptr [[_TMP5]], align 8, !nonnull [[META5]], !align [[META6]] // CHECK-NEXT: [[TMP20:%.*]] = load i32, ptr [[TMP19]], align 4 // CHECK-NEXT: [[ADD9:%.*]] = add nsw i32 [[TMP20]], [[TMP18]] // CHECK-NEXT: store i32 [[ADD9]], ptr [[TMP19]], align 4 @@ -507,13 +508,13 @@ int main(void) { // CHECK-NEXT: store ptr [[V]], ptr [[V_ADDR]], align 8 // CHECK-NEXT: store ptr [[SUM_V]], ptr [[SUM_V_ADDR]], align 8 // CHECK-NEXT: store ptr [[PROD_V]], ptr [[PROD_V_ADDR]], align 8 -// CHECK-NEXT: [[TMP1:%.*]] = load ptr, ptr [[SUM_V_ADDR]], align 8 +// CHECK-NEXT: [[TMP1:%.*]] = load ptr, ptr [[SUM_V_ADDR]], align 8, !nonnull [[META5]], !align [[META6]] // CHECK-NEXT: store i32 0, ptr [[TMP1]], align 4 -// CHECK-NEXT: [[TMP2:%.*]] = load ptr, ptr [[PROD_V_ADDR]], align 8 +// CHECK-NEXT: [[TMP2:%.*]] = load ptr, ptr [[PROD_V_ADDR]], align 8, !nonnull [[META5]], !align [[META6]] // CHECK-NEXT: store i32 1, ptr [[TMP2]], align 4 -// CHECK-NEXT: [[TMP3:%.*]] = load ptr, ptr [[SUM_V_ADDR]], align 8 +// CHECK-NEXT: [[TMP3:%.*]] = load ptr, ptr [[SUM_V_ADDR]], align 8, !nonnull [[META5]], !align [[META6]] // CHECK-NEXT: store ptr [[TMP3]], ptr [[TMP]], align 8 -// CHECK-NEXT: [[TMP4:%.*]] = load ptr, ptr [[PROD_V_ADDR]], align 8 +// CHECK-NEXT: [[TMP4:%.*]] = load ptr, ptr [[PROD_V_ADDR]], align 8, !nonnull [[META5]], !align [[META6]] // CHECK-NEXT: store ptr [[TMP4]], ptr [[_TMP1]], align 8 // CHECK-NEXT: [[TMP5:%.*]] = load i32, ptr [[N_ADDR]], align 4 // CHECK-NEXT: store i32 [[TMP5]], ptr [[DOTCAPTURE_EXPR_]], align 4 @@ -532,10 +533,10 @@ int main(void) { // CHECK-NEXT: store i32 [[TMP8]], ptr [[DOTOMP_UB]], align 4 // CHECK-NEXT: store i32 1, ptr [[DOTOMP_STRIDE]], align 4 // CHECK-NEXT: store i32 0, ptr [[DOTOMP_IS_LAST]], align 4 -// CHECK-NEXT: [[TMP9:%.*]] = load ptr, ptr [[TMP]], align 8 +// CHECK-NEXT: [[TMP9:%.*]] = load ptr, ptr [[TMP]], align 8, !nonnull [[META5]], !align [[META6]] // CHECK-NEXT: store i32 0, ptr [[SUM_V5]], align 4 // CHECK-NEXT: store ptr [[SUM_V5]], ptr [[_TMP6]], align 8 -// CHECK-NEXT: [[TMP10:%.*]] = load ptr, ptr [[_TMP1]], align 8 +// CHECK-NEXT: [[TMP10:%.*]] = load ptr, ptr [[_TMP1]], align 8, !nonnull [[META5]], !align [[META6]] // CHECK-NEXT: store i32 1, ptr [[PROD_V7]], align 4 // CHECK-NEXT: store ptr [[PROD_V7]], ptr [[_TMP8]], align 8 // CHECK-NEXT: call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP0]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1) @@ -570,7 +571,7 @@ int main(void) { // CHECK-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP20]] to i64 // CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[TMP19]], i64 [[IDXPROM]] // CHECK-NEXT: [[TMP21:%.*]] = load i32, ptr [[ARRAYIDX]], align 4 -// CHECK-NEXT: [[TMP22:%.*]] = load ptr, ptr [[_TMP6]], align 8 +// CHECK-NEXT: [[TMP22:%.*]] = load ptr, ptr [[_TMP6]], align 8, !nonnull [[META5]], !align [[META6]] // CHECK-NEXT: [[TMP23:%.*]] = load i32, ptr [[TMP22]], align 4 // CHECK-NEXT: [[ADD12:%.*]] = add nsw i32 [[TMP23]], [[TMP21]] // CHECK-NEXT: store i32 [[ADD12]], ptr [[TMP22]], align 4 @@ -579,7 +580,7 @@ int main(void) { // CHECK-NEXT: [[IDXPROM13:%.*]] = sext i32 [[TMP25]] to i64 // CHECK-NEXT: [[ARRAYIDX14:%.*]] = getelementptr inbounds i32, ptr [[TMP24]], i64 [[IDXPROM13]] // CHECK-NEXT: [[TMP26:%.*]] = load i32, ptr [[ARRAYIDX14]], align 4 -// CHECK-NEXT: [[TMP27:%.*]] = load ptr, ptr [[_TMP8]], align 8 +// CHECK-NEXT: [[TMP27:%.*]] = load ptr, ptr [[_TMP8]], align 8, !nonnull [[META5]], !align [[META6]] // CHECK-NEXT: [[TMP28:%.*]] = load i32, ptr [[TMP27]], align 4 // CHECK-NEXT: [[MUL15:%.*]] = mul nsw i32 [[TMP28]], [[TMP26]] // CHECK-NEXT: store i32 [[MUL15]], ptr [[TMP27]], align 4 @@ -698,7 +699,7 @@ int main(void) { // CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr [[I]], align 4 // CHECK-NEXT: [[INC:%.*]] = add nsw i32 [[TMP4]], 1 // CHECK-NEXT: store i32 [[INC]], ptr [[I]], align 4 -// CHECK-NEXT: br label [[FOR_COND]], !llvm.loop [[LOOP7:![0-9]+]] +// CHECK-NEXT: br label [[FOR_COND]], !llvm.loop [[LOOP9:![0-9]+]] // CHECK: for.end: // CHECK-NEXT: call void @__kmpc_push_num_threads(ptr @[[GLOB3]], i32 [[TMP0]], i32 4) // CHECK-NEXT: call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 1, ptr @main.omp_outlined, ptr [[V]]) @@ -707,4 +708,41 @@ int main(void) { // CHECK-NEXT: call void @__kmpc_push_num_threads(ptr @[[GLOB3]], i32 [[TMP0]], i32 4) // CHECK-NEXT: call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 3, ptr @main.omp_outlined.1, ptr [[V]], ptr [[SUM_V_EXT]], ptr [[PROD_V_EXT]]) // CHECK-NEXT: ret i32 0 - +// +// +// CHECK-LABEL: define {{[^@]+}}@main.omp_outlined +// CHECK-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(40) [[V:%.*]]) #[[ATTR2]] { +// CHECK-NEXT: entry: +// CHECK-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8 +// CHECK-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8 +// CHECK-NEXT: [[V_ADDR:%.*]] = alloca ptr, align 8 +// CHECK-NEXT: [[S_V:%.*]] = alloca i32, align 4 +// CHECK-NEXT: store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8 +// CHECK-NEXT: store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8 +// CHECK-NEXT: store ptr [[V]], ptr [[V_ADDR]], align 8 +// CHECK-NEXT: [[TMP0:%.*]] = load ptr, ptr [[V_ADDR]], align 8, !nonnull [[META5]], !align [[META6]] +// CHECK-NEXT: [[ARRAYDECAY:%.*]] = getelementptr inbounds [10 x i32], ptr [[TMP0]], i64 0, i64 0 +// CHECK-NEXT: call void @_Z6do_rediPiRi(i32 noundef 10, ptr noundef [[ARRAYDECAY]], ptr noundef nonnull align 4 dereferenceable(4) [[S_V]]) +// CHECK-NEXT: ret void +// +// +// CHECK-LABEL: define {{[^@]+}}@main.omp_outlined.1 +// CHECK-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(40) [[V:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[SUM_V_EXT:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[PROD_V_EXT:%.*]]) #[[ATTR2]] { +// CHECK-NEXT: entry: +// CHECK-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8 +// CHECK-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8 +// CHECK-NEXT: [[V_ADDR:%.*]] = alloca ptr, align 8 +// CHECK-NEXT: [[SUM_V_EXT_ADDR:%.*]] = alloca ptr, align 8 +// CHECK-NEXT: [[PROD_V_EXT_ADDR:%.*]] = alloca ptr, align 8 +// CHECK-NEXT: store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8 +// CHECK-NEXT: store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8 +// CHECK-NEXT: store ptr [[V]], ptr [[V_ADDR]], align 8 +// CHECK-NEXT: store ptr [[SUM_V_EXT]], ptr [[SUM_V_EXT_ADDR]], align 8 +// CHECK-NEXT: store ptr [[PROD_V_EXT]], ptr [[PROD_V_EXT_ADDR]], align 8 +// CHECK-NEXT: [[TMP0:%.*]] = load ptr, ptr [[V_ADDR]], align 8, !nonnull [[META5]], !align [[META6]] +// CHECK-NEXT: [[TMP1:%.*]] = load ptr, ptr [[SUM_V_EXT_ADDR]], align 8, !nonnull [[META5]], !align [[META6]] +// CHECK-NEXT: [[TMP2:%.*]] = load ptr, ptr [[PROD_V_EXT_ADDR]], align 8, !nonnull [[META5]], !align [[META6]] +// CHECK-NEXT: [[ARRAYDECAY:%.*]] = getelementptr inbounds [10 x i32], ptr [[TMP0]], i64 0, i64 0 +// CHECK-NEXT: call void @_Z15do_red_extendediPiRiS0_(i32 noundef 10, ptr noundef [[ARRAYDECAY]], ptr noundef nonnull align 4 dereferenceable(4) [[TMP1]], ptr noundef nonnull align 4 dereferenceable(4) [[TMP2]]) +// CHECK-NEXT: ret void +// diff --git a/clang/test/OpenMP/loop_collapse_codegen.cpp b/clang/test/OpenMP/loop_collapse_codegen.cpp new file mode 100644 index 0000000000000..dbe42e934076d --- /dev/null +++ b/clang/test/OpenMP/loop_collapse_codegen.cpp @@ -0,0 +1,857 @@ +// RUN: %clang_cc1 -verify -fopenmp -x c++ -triple x86_64-unknown-unknown \ +// RUN: -emit-llvm %s -o - | FileCheck %s + +// expected-no-diagnostics + +// CHECK-LABEL: define internal void @_Z17triangulat_loop_1v.omp_outlined( +// CHECK-SAME: ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1:[0-9]+]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8 +// CHECK-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8 +// CHECK-NEXT: [[DOTOMP_IV:%.*]] = alloca i64, align 8 +// CHECK-NEXT: [[TMP:%.*]] = alloca i32, align 4 +// CHECK-NEXT: [[_TMP1:%.*]] = alloca i32, align 4 +// CHECK-NEXT: [[_TMP2:%.*]] = alloca i32, align 4 +// CHECK-NEXT: [[DOTLB_MIN:%.*]] = alloca i32, align 4 +// CHECK-NEXT: [[DOTLB_MAX:%.*]] = alloca i32, align 4 +// CHECK-NEXT: [[DOTMIN_LESS_MAX:%.*]] = alloca i8, align 1 +// CHECK-NEXT: [[DOTLOWER:%.*]] = alloca i32, align 4 +// CHECK-NEXT: [[DOTLB_MIN4:%.*]] = alloca i32, align 4 +// CHECK-NEXT: [[DOTLB_MAX7:%.*]] = alloca i32, align 4 +// CHECK-NEXT: [[DOTMIN_LESS_MAX13:%.*]] = alloca i8, align 1 +// CHECK-NEXT: [[DOTLOWER16:%.*]] = alloca i32, align 4 +// CHECK-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i64, align 8 +// CHECK-NEXT: [[DOTOMP_LB:%.*]] = alloca i64, align 8 +// CHECK-NEXT: [[DOTOMP_UB:%.*]] = alloca i64, align 8 +// CHECK-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8 +// CHECK-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4 +// CHECK-NEXT: [[I:%.*]] = alloca i32, align 4 +// CHECK-NEXT: [[J:%.*]] = alloca i32, align 4 +// CHECK-NEXT: [[K:%.*]] = alloca i32, align 4 +// CHECK-NEXT: store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8 +// CHECK-NEXT: store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8 +// CHECK-NEXT: store i32 0, ptr [[TMP]], align 4 +// CHECK-NEXT: [[TMP0:%.*]] = load i32, ptr [[TMP]], align 4 +// CHECK-NEXT: [[ADD:%.*]] = add i32 [[TMP0]], 1 +// CHECK-NEXT: store i32 [[ADD]], ptr [[DOTLB_MIN]], align 4 +// CHECK-NEXT: store i32 9, ptr [[TMP]], align 4 +// CHECK-NEXT: [[TMP1:%.*]] = load i32, ptr [[TMP]], align 4 +// CHECK-NEXT: [[ADD3:%.*]] = add i32 [[TMP1]], 1 +// CHECK-NEXT: store i32 [[ADD3]], ptr [[DOTLB_MAX]], align 4 +// CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr [[DOTLB_MIN]], align 4 +// CHECK-NEXT: [[TMP3:%.*]] = load i32, ptr [[DOTLB_MAX]], align 4 +// CHECK-NEXT: [[CMP:%.*]] = icmp ult i32 [[TMP2]], [[TMP3]] +// CHECK-NEXT: [[STOREDV:%.*]] = zext i1 [[CMP]] to i8 +// CHECK-NEXT: store i8 [[STOREDV]], ptr [[DOTMIN_LESS_MAX]], align 1 +// CHECK-NEXT: [[TMP4:%.*]] = load i8, ptr [[DOTMIN_LESS_MAX]], align 1 +// CHECK-NEXT: [[LOADEDV:%.*]] = trunc i8 [[TMP4]] to i1 +// CHECK-NEXT: br i1 [[LOADEDV]], label %[[COND_TRUE:.*]], label %[[COND_FALSE:.*]] +// CHECK: [[COND_TRUE]]: +// CHECK-NEXT: [[TMP5:%.*]] = load i32, ptr [[DOTLB_MIN]], align 4 +// CHECK-NEXT: br label %[[COND_END:.*]] +// CHECK: [[COND_FALSE]]: +// CHECK-NEXT: [[TMP6:%.*]] = load i32, ptr [[DOTLB_MAX]], align 4 +// CHECK-NEXT: br label %[[COND_END]] +// CHECK: [[COND_END]]: +// CHECK-NEXT: [[COND:%.*]] = phi i32 [ [[TMP5]], %[[COND_TRUE]] ], [ [[TMP6]], %[[COND_FALSE]] ] +// CHECK-NEXT: store i32 [[COND]], ptr [[TMP]], align 4 +// CHECK-NEXT: store i32 [[COND]], ptr [[DOTLOWER]], align 4 +// CHECK-NEXT: [[TMP7:%.*]] = load i32, ptr [[TMP]], align 4 +// CHECK-NEXT: [[ADD5:%.*]] = add i32 [[TMP7]], 1 +// CHECK-NEXT: store i32 [[ADD5]], ptr [[_TMP1]], align 4 +// CHECK-NEXT: [[TMP8:%.*]] = load i32, ptr [[_TMP1]], align 4 +// CHECK-NEXT: [[ADD6:%.*]] = add i32 [[TMP8]], 1 +// CHECK-NEXT: store i32 [[ADD6]], ptr [[DOTLB_MIN4]], align 4 +// CHECK-NEXT: [[TMP9:%.*]] = load i32, ptr [[TMP]], align 4 +// CHECK-NEXT: [[ADD8:%.*]] = add i32 [[TMP9]], 1 +// CHECK-NEXT: [[TMP10:%.*]] = load i32, ptr [[TMP]], align 4 +// CHECK-NEXT: [[ADD9:%.*]] = add i32 [[TMP10]], 1 +// CHECK-NEXT: [[SUB:%.*]] = sub i32 10, [[ADD9]] +// CHECK-NEXT: [[ADD10:%.*]] = add i32 [[SUB]], 1 +// CHECK-NEXT: [[MUL:%.*]] = mul i32 [[ADD10]], 1 +// CHECK-NEXT: [[ADD11:%.*]] = add i32 [[ADD8]], [[MUL]] +// CHECK-NEXT: store i32 [[ADD11]], ptr [[_TMP1]], align 4 +// CHECK-NEXT: [[TMP11:%.*]] = load i32, ptr [[_TMP1]], align 4 +// CHECK-NEXT: [[ADD12:%.*]] = add i32 [[TMP11]], 1 +// CHECK-NEXT: store i32 [[ADD12]], ptr [[DOTLB_MAX7]], align 4 +// CHECK-NEXT: [[TMP12:%.*]] = load i32, ptr [[DOTLB_MIN4]], align 4 +// CHECK-NEXT: [[TMP13:%.*]] = load i32, ptr [[DOTLB_MAX7]], align 4 +// CHECK-NEXT: [[CMP14:%.*]] = icmp ult i32 [[TMP12]], [[TMP13]] +// CHECK-NEXT: [[STOREDV15:%.*]] = zext i1 [[CMP14]] to i8 +// CHECK-NEXT: store i8 [[STOREDV15]], ptr [[DOTMIN_LESS_MAX13]], align 1 +// CHECK-NEXT: [[TMP14:%.*]] = load i8, ptr [[DOTMIN_LESS_MAX13]], align 1 +// CHECK-NEXT: [[LOADEDV17:%.*]] = trunc i8 [[TMP14]] to i1 +// CHECK-NEXT: br i1 [[LOADEDV17]], label %[[COND_TRUE18:.*]], label %[[COND_FALSE19:.*]] +// CHECK: [[COND_TRUE18]]: +// CHECK-NEXT: [[TMP15:%.*]] = load i32, ptr [[DOTLB_MIN4]], align 4 +// CHECK-NEXT: br label %[[COND_END20:.*]] +// CHECK: [[COND_FALSE19]]: +// CHECK-NEXT: [[TMP16:%.*]] = load i32, ptr [[DOTLB_MAX7]], align 4 +// CHECK-NEXT: br label %[[COND_END20]] +// CHECK: [[COND_END20]]: +// CHECK-NEXT: [[COND21:%.*]] = phi i32 [ [[TMP15]], %[[COND_TRUE18]] ], [ [[TMP16]], %[[COND_FALSE19]] ] +// CHECK-NEXT: store i32 [[COND21]], ptr [[_TMP1]], align 4 +// CHECK-NEXT: store i32 [[COND21]], ptr [[DOTLOWER16]], align 4 +// CHECK-NEXT: [[TMP17:%.*]] = load i32, ptr [[DOTLOWER]], align 4 +// CHECK-NEXT: [[SUB22:%.*]] = sub i32 10, [[TMP17]] +// CHECK-NEXT: [[ADD23:%.*]] = add i32 [[SUB22]], 1 +// CHECK-NEXT: [[CONV:%.*]] = zext i32 [[ADD23]] to i64 +// CHECK-NEXT: [[MUL24:%.*]] = mul nsw i64 10, [[CONV]] +// CHECK-NEXT: [[TMP18:%.*]] = load i32, ptr [[DOTLOWER16]], align 4 +// CHECK-NEXT: [[SUB25:%.*]] = sub i32 10, [[TMP18]] +// CHECK-NEXT: [[ADD26:%.*]] = add i32 [[SUB25]], 1 +// CHECK-NEXT: [[CONV27:%.*]] = zext i32 [[ADD26]] to i64 +// CHECK-NEXT: [[MUL28:%.*]] = mul nsw i64 [[MUL24]], [[CONV27]] +// CHECK-NEXT: [[SUB29:%.*]] = sub nsw i64 [[MUL28]], 1 +// CHECK-NEXT: store i64 [[SUB29]], ptr [[DOTCAPTURE_EXPR_]], align 8 +// CHECK-NEXT: store i64 0, ptr [[DOTOMP_LB]], align 8 +// CHECK-NEXT: [[TMP19:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_]], align 8 +// CHECK-NEXT: store i64 [[TMP19]], ptr [[DOTOMP_UB]], align 8 +// CHECK-NEXT: store i64 1, ptr [[DOTOMP_STRIDE]], align 8 +// CHECK-NEXT: store i32 0, ptr [[DOTOMP_IS_LAST]], align 4 +// CHECK-NEXT: [[TMP20:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8 +// CHECK-NEXT: [[TMP21:%.*]] = load i32, ptr [[TMP20]], align 4 +// CHECK-NEXT: call void @__kmpc_for_static_init_8(ptr @[[GLOB1:[0-9]+]], i32 [[TMP21]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i64 1, i64 1) +// CHECK-NEXT: [[TMP22:%.*]] = load i64, ptr [[DOTOMP_UB]], align 8 +// CHECK-NEXT: [[TMP23:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_]], align 8 +// CHECK-NEXT: [[CMP30:%.*]] = icmp sgt i64 [[TMP22]], [[TMP23]] +// CHECK-NEXT: br i1 [[CMP30]], label %[[COND_TRUE31:.*]], label %[[COND_FALSE32:.*]] +// CHECK: [[COND_TRUE31]]: +// CHECK-NEXT: [[TMP24:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_]], align 8 +// CHECK-NEXT: br label %[[COND_END33:.*]] +// CHECK: [[COND_FALSE32]]: +// CHECK-NEXT: [[TMP25:%.*]] = load i64, ptr [[DOTOMP_UB]], align 8 +// CHECK-NEXT: br label %[[COND_END33]] +// CHECK: [[COND_END33]]: +// CHECK-NEXT: [[COND34:%.*]] = phi i64 [ [[TMP24]], %[[COND_TRUE31]] ], [ [[TMP25]], %[[COND_FALSE32]] ] +// CHECK-NEXT: store i64 [[COND34]], ptr [[DOTOMP_UB]], align 8 +// CHECK-NEXT: [[TMP26:%.*]] = load i64, ptr [[DOTOMP_LB]], align 8 +// CHECK-NEXT: store i64 [[TMP26]], ptr [[DOTOMP_IV]], align 8 +// CHECK-NEXT: br label %[[OMP_INNER_FOR_COND:.*]] +// CHECK: [[OMP_INNER_FOR_COND]]: +// CHECK-NEXT: [[TMP27:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8 +// CHECK-NEXT: [[TMP28:%.*]] = load i64, ptr [[DOTOMP_UB]], align 8 +// CHECK-NEXT: [[CMP35:%.*]] = icmp sle i64 [[TMP27]], [[TMP28]] +// CHECK-NEXT: br i1 [[CMP35]], label %[[OMP_INNER_FOR_BODY:.*]], label %[[OMP_INNER_FOR_END:.*]] +// CHECK: [[OMP_INNER_FOR_BODY]]: +// CHECK-NEXT: [[TMP29:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8 +// CHECK-NEXT: [[TMP30:%.*]] = load i32, ptr [[DOTLOWER]], align 4 +// CHECK-NEXT: [[SUB36:%.*]] = sub i32 10, [[TMP30]] +// CHECK-NEXT: [[ADD37:%.*]] = add i32 [[SUB36]], 1 +// CHECK-NEXT: [[MUL38:%.*]] = mul i32 1, [[ADD37]] +// CHECK-NEXT: [[TMP31:%.*]] = load i32, ptr [[DOTLOWER16]], align 4 +// CHECK-NEXT: [[SUB39:%.*]] = sub i32 10, [[TMP31]] +// CHECK-NEXT: [[ADD40:%.*]] = add i32 [[SUB39]], 1 +// CHECK-NEXT: [[MUL41:%.*]] = mul i32 [[MUL38]], [[ADD40]] +// CHECK-NEXT: [[CONV42:%.*]] = zext i32 [[MUL41]] to i64 +// CHECK-NEXT: [[DIV:%.*]] = sdiv i64 [[TMP29]], [[CONV42]] +// CHECK-NEXT: [[MUL43:%.*]] = mul nsw i64 [[DIV]], 1 +// CHECK-NEXT: [[ADD44:%.*]] = add nsw i64 0, [[MUL43]] +// CHECK-NEXT: [[CONV45:%.*]] = trunc i64 [[ADD44]] to i32 +// CHECK-NEXT: store i32 [[CONV45]], ptr [[I]], align 4 +// CHECK-NEXT: [[TMP32:%.*]] = load i32, ptr [[I]], align 4 +// CHECK-NEXT: [[ADD46:%.*]] = add i32 [[TMP32]], 1 +// CHECK-NEXT: [[CONV47:%.*]] = zext i32 [[ADD46]] to i64 +// CHECK-NEXT: [[TMP33:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8 +// CHECK-NEXT: [[TMP34:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8 +// CHECK-NEXT: [[TMP35:%.*]] = load i32, ptr [[DOTLOWER]], align 4 +// CHECK-NEXT: [[SUB48:%.*]] = sub i32 10, [[TMP35]] +// CHECK-NEXT: [[ADD49:%.*]] = add i32 [[SUB48]], 1 +// CHECK-NEXT: [[MUL50:%.*]] = mul i32 1, [[ADD49]] +// CHECK-NEXT: [[TMP36:%.*]] = load i32, ptr [[DOTLOWER16]], align 4 +// CHECK-NEXT: [[SUB51:%.*]] = sub i32 10, [[TMP36]] +// CHECK-NEXT: [[ADD52:%.*]] = add i32 [[SUB51]], 1 +// CHECK-NEXT: [[MUL53:%.*]] = mul i32 [[MUL50]], [[ADD52]] +// CHECK-NEXT: [[CONV54:%.*]] = zext i32 [[MUL53]] to i64 +// CHECK-NEXT: [[DIV55:%.*]] = sdiv i64 [[TMP34]], [[CONV54]] +// CHECK-NEXT: [[TMP37:%.*]] = load i32, ptr [[DOTLOWER]], align 4 +// CHECK-NEXT: [[SUB56:%.*]] = sub i32 10, [[TMP37]] +// CHECK-NEXT: [[ADD57:%.*]] = add i32 [[SUB56]], 1 +// CHECK-NEXT: [[MUL58:%.*]] = mul i32 1, [[ADD57]] +// CHECK-NEXT: [[TMP38:%.*]] = load i32, ptr [[DOTLOWER16]], align 4 +// CHECK-NEXT: [[SUB59:%.*]] = sub i32 10, [[TMP38]] +// CHECK-NEXT: [[ADD60:%.*]] = add i32 [[SUB59]], 1 +// CHECK-NEXT: [[MUL61:%.*]] = mul i32 [[MUL58]], [[ADD60]] +// CHECK-NEXT: [[CONV62:%.*]] = zext i32 [[MUL61]] to i64 +// CHECK-NEXT: [[MUL63:%.*]] = mul nsw i64 [[DIV55]], [[CONV62]] +// CHECK-NEXT: [[SUB64:%.*]] = sub nsw i64 [[TMP33]], [[MUL63]] +// CHECK-NEXT: [[TMP39:%.*]] = load i32, ptr [[DOTLOWER16]], align 4 +// CHECK-NEXT: [[SUB65:%.*]] = sub i32 10, [[TMP39]] +// CHECK-NEXT: [[ADD66:%.*]] = add i32 [[SUB65]], 1 +// CHECK-NEXT: [[MUL67:%.*]] = mul i32 1, [[ADD66]] +// CHECK-NEXT: [[CONV68:%.*]] = zext i32 [[MUL67]] to i64 +// CHECK-NEXT: [[DIV69:%.*]] = sdiv i64 [[SUB64]], [[CONV68]] +// CHECK-NEXT: [[MUL70:%.*]] = mul nsw i64 [[DIV69]], 1 +// CHECK-NEXT: [[ADD71:%.*]] = add nsw i64 [[CONV47]], [[MUL70]] +// CHECK-NEXT: [[CONV72:%.*]] = trunc i64 [[ADD71]] to i32 +// CHECK-NEXT: store i32 [[CONV72]], ptr [[J]], align 4 +// CHECK-NEXT: [[TMP40:%.*]] = load i32, ptr [[J]], align 4 +// CHECK-NEXT: [[ADD73:%.*]] = add i32 [[TMP40]], 1 +// CHECK-NEXT: [[CONV74:%.*]] = zext i32 [[ADD73]] to i64 +// CHECK-NEXT: [[TMP41:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8 +// CHECK-NEXT: [[TMP42:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8 +// CHECK-NEXT: [[TMP43:%.*]] = load i32, ptr [[DOTLOWER]], align 4 +// CHECK-NEXT: [[SUB75:%.*]] = sub i32 10, [[TMP43]] +// CHECK-NEXT: [[ADD76:%.*]] = add i32 [[SUB75]], 1 +// CHECK-NEXT: [[MUL77:%.*]] = mul i32 1, [[ADD76]] +// CHECK-NEXT: [[TMP44:%.*]] = load i32, ptr [[DOTLOWER16]], align 4 +// CHECK-NEXT: [[SUB78:%.*]] = sub i32 10, [[TMP44]] +// CHECK-NEXT: [[ADD79:%.*]] = add i32 [[SUB78]], 1 +// CHECK-NEXT: [[MUL80:%.*]] = mul i32 [[MUL77]], [[ADD79]] +// CHECK-NEXT: [[CONV81:%.*]] = zext i32 [[MUL80]] to i64 +// CHECK-NEXT: [[DIV82:%.*]] = sdiv i64 [[TMP42]], [[CONV81]] +// CHECK-NEXT: [[TMP45:%.*]] = load i32, ptr [[DOTLOWER]], align 4 +// CHECK-NEXT: [[SUB83:%.*]] = sub i32 10, [[TMP45]] +// CHECK-NEXT: [[ADD84:%.*]] = add i32 [[SUB83]], 1 +// CHECK-NEXT: [[MUL85:%.*]] = mul i32 1, [[ADD84]] +// CHECK-NEXT: [[TMP46:%.*]] = load i32, ptr [[DOTLOWER16]], align 4 +// CHECK-NEXT: [[SUB86:%.*]] = sub i32 10, [[TMP46]] +// CHECK-NEXT: [[ADD87:%.*]] = add i32 [[SUB86]], 1 +// CHECK-NEXT: [[MUL88:%.*]] = mul i32 [[MUL85]], [[ADD87]] +// CHECK-NEXT: [[CONV89:%.*]] = zext i32 [[MUL88]] to i64 +// CHECK-NEXT: [[MUL90:%.*]] = mul nsw i64 [[DIV82]], [[CONV89]] +// CHECK-NEXT: [[SUB91:%.*]] = sub nsw i64 [[TMP41]], [[MUL90]] +// CHECK-NEXT: [[TMP47:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8 +// CHECK-NEXT: [[TMP48:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8 +// CHECK-NEXT: [[TMP49:%.*]] = load i32, ptr [[DOTLOWER]], align 4 +// CHECK-NEXT: [[SUB92:%.*]] = sub i32 10, [[TMP49]] +// CHECK-NEXT: [[ADD93:%.*]] = add i32 [[SUB92]], 1 +// CHECK-NEXT: [[MUL94:%.*]] = mul i32 1, [[ADD93]] +// CHECK-NEXT: [[TMP50:%.*]] = load i32, ptr [[DOTLOWER16]], align 4 +// CHECK-NEXT: [[SUB95:%.*]] = sub i32 10, [[TMP50]] +// CHECK-NEXT: [[ADD96:%.*]] = add i32 [[SUB95]], 1 +// CHECK-NEXT: [[MUL97:%.*]] = mul i32 [[MUL94]], [[ADD96]] +// CHECK-NEXT: [[CONV98:%.*]] = zext i32 [[MUL97]] to i64 +// CHECK-NEXT: [[DIV99:%.*]] = sdiv i64 [[TMP48]], [[CONV98]] +// CHECK-NEXT: [[TMP51:%.*]] = load i32, ptr [[DOTLOWER]], align 4 +// CHECK-NEXT: [[SUB100:%.*]] = sub i32 10, [[TMP51]] +// CHECK-NEXT: [[ADD101:%.*]] = add i32 [[SUB100]], 1 +// CHECK-NEXT: [[MUL102:%.*]] = mul i32 1, [[ADD101]] +// CHECK-NEXT: [[TMP52:%.*]] = load i32, ptr [[DOTLOWER16]], align 4 +// CHECK-NEXT: [[SUB103:%.*]] = sub i32 10, [[TMP52]] +// CHECK-NEXT: [[ADD104:%.*]] = add i32 [[SUB103]], 1 +// CHECK-NEXT: [[MUL105:%.*]] = mul i32 [[MUL102]], [[ADD104]] +// CHECK-NEXT: [[CONV106:%.*]] = zext i32 [[MUL105]] to i64 +// CHECK-NEXT: [[MUL107:%.*]] = mul nsw i64 [[DIV99]], [[CONV106]] +// CHECK-NEXT: [[SUB108:%.*]] = sub nsw i64 [[TMP47]], [[MUL107]] +// CHECK-NEXT: [[TMP53:%.*]] = load i32, ptr [[DOTLOWER16]], align 4 +// CHECK-NEXT: [[SUB109:%.*]] = sub i32 10, [[TMP53]] +// CHECK-NEXT: [[ADD110:%.*]] = add i32 [[SUB109]], 1 +// CHECK-NEXT: [[MUL111:%.*]] = mul i32 1, [[ADD110]] +// CHECK-NEXT: [[CONV112:%.*]] = zext i32 [[MUL111]] to i64 +// CHECK-NEXT: [[DIV113:%.*]] = sdiv i64 [[SUB108]], [[CONV112]] +// CHECK-NEXT: [[TMP54:%.*]] = load i32, ptr [[DOTLOWER16]], align 4 +// CHECK-NEXT: [[SUB114:%.*]] = sub i32 10, [[TMP54]] +// CHECK-NEXT: [[ADD115:%.*]] = add i32 [[SUB114]], 1 +// CHECK-NEXT: [[MUL116:%.*]] = mul i32 1, [[ADD115]] +// CHECK-NEXT: [[CONV117:%.*]] = zext i32 [[MUL116]] to i64 +// CHECK-NEXT: [[MUL118:%.*]] = mul nsw i64 [[DIV113]], [[CONV117]] +// CHECK-NEXT: [[SUB119:%.*]] = sub nsw i64 [[SUB91]], [[MUL118]] +// CHECK-NEXT: [[MUL120:%.*]] = mul nsw i64 [[SUB119]], 1 +// CHECK-NEXT: [[ADD121:%.*]] = add nsw i64 [[CONV74]], [[MUL120]] +// CHECK-NEXT: [[CONV122:%.*]] = trunc i64 [[ADD121]] to i32 +// CHECK-NEXT: store i32 [[CONV122]], ptr [[K]], align 4 +// CHECK-NEXT: [[TMP55:%.*]] = load i32, ptr [[J]], align 4 +// CHECK-NEXT: [[CMP123:%.*]] = icmp ult i32 [[TMP55]], 10 +// CHECK-NEXT: br i1 [[CMP123]], label %[[OMP_BODY_NEXT:.*]], label %[[OMP_BODY_CONTINUE:.*]] +// CHECK: [[OMP_BODY_NEXT]]: +// CHECK-NEXT: [[TMP56:%.*]] = load i32, ptr [[K]], align 4 +// CHECK-NEXT: [[CMP124:%.*]] = icmp ult i32 [[TMP56]], 10 +// CHECK-NEXT: br i1 [[CMP124]], label %[[OMP_BODY_NEXT125:.*]], label %[[OMP_BODY_CONTINUE]] +// CHECK: [[OMP_BODY_NEXT125]]: +// CHECK-NEXT: br label %[[OMP_BODY_CONTINUE]] +// CHECK: [[OMP_BODY_CONTINUE]]: +// CHECK-NEXT: br label %[[OMP_INNER_FOR_INC:.*]] +// CHECK: [[OMP_INNER_FOR_INC]]: +// CHECK-NEXT: [[TMP57:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8 +// CHECK-NEXT: [[ADD126:%.*]] = add nsw i64 [[TMP57]], 1 +// CHECK-NEXT: store i64 [[ADD126]], ptr [[DOTOMP_IV]], align 8 +// CHECK-NEXT: br label %[[OMP_INNER_FOR_COND]] +// CHECK: [[OMP_INNER_FOR_END]]: +// CHECK-NEXT: br label %[[OMP_LOOP_EXIT:.*]] +// CHECK: [[OMP_LOOP_EXIT]]: +// CHECK-NEXT: [[TMP58:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8 +// CHECK-NEXT: [[TMP59:%.*]] = load i32, ptr [[TMP58]], align 4 +// CHECK-NEXT: call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP59]]) +// CHECK-NEXT: ret void +void triangulat_loop_1() { +#pragma omp parallel for collapse(3) + for (unsigned int i = 0; i < 10; ++i) + for (unsigned int j = i + 1; j < 10; ++j) + for (unsigned int k = j + 1; k < 10; ++k) + ; +} + +// CHECK-LABEL: define internal void @_Z17triangulat_loop_2v.omp_outlined( +// CHECK-SAME: ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8 +// CHECK-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8 +// CHECK-NEXT: [[DOTOMP_IV:%.*]] = alloca i64, align 8 +// CHECK-NEXT: [[TMP:%.*]] = alloca i32, align 4 +// CHECK-NEXT: [[_TMP1:%.*]] = alloca i32, align 4 +// CHECK-NEXT: [[_TMP2:%.*]] = alloca i32, align 4 +// CHECK-NEXT: [[DOTLB_MIN:%.*]] = alloca i32, align 4 +// CHECK-NEXT: [[DOTLB_MAX:%.*]] = alloca i32, align 4 +// CHECK-NEXT: [[DOTMIN_LESS_MAX:%.*]] = alloca i8, align 1 +// CHECK-NEXT: [[DOTLOWER:%.*]] = alloca i32, align 4 +// CHECK-NEXT: [[DOTLB_MIN4:%.*]] = alloca i32, align 4 +// CHECK-NEXT: [[DOTLB_MAX7:%.*]] = alloca i32, align 4 +// CHECK-NEXT: [[DOTMIN_LESS_MAX13:%.*]] = alloca i8, align 1 +// CHECK-NEXT: [[DOTLOWER16:%.*]] = alloca i32, align 4 +// CHECK-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i64, align 8 +// CHECK-NEXT: [[DOTOMP_LB:%.*]] = alloca i64, align 8 +// CHECK-NEXT: [[DOTOMP_UB:%.*]] = alloca i64, align 8 +// CHECK-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8 +// CHECK-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4 +// CHECK-NEXT: [[I:%.*]] = alloca i32, align 4 +// CHECK-NEXT: [[J:%.*]] = alloca i32, align 4 +// CHECK-NEXT: [[K:%.*]] = alloca i32, align 4 +// CHECK-NEXT: store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8 +// CHECK-NEXT: store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8 +// CHECK-NEXT: store i32 0, ptr [[TMP]], align 4 +// CHECK-NEXT: [[TMP0:%.*]] = load i32, ptr [[TMP]], align 4 +// CHECK-NEXT: [[ADD:%.*]] = add i32 [[TMP0]], 1 +// CHECK-NEXT: store i32 [[ADD]], ptr [[DOTLB_MIN]], align 4 +// CHECK-NEXT: store i32 9, ptr [[TMP]], align 4 +// CHECK-NEXT: [[TMP1:%.*]] = load i32, ptr [[TMP]], align 4 +// CHECK-NEXT: [[ADD3:%.*]] = add i32 [[TMP1]], 1 +// CHECK-NEXT: store i32 [[ADD3]], ptr [[DOTLB_MAX]], align 4 +// CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr [[DOTLB_MIN]], align 4 +// CHECK-NEXT: [[TMP3:%.*]] = load i32, ptr [[DOTLB_MAX]], align 4 +// CHECK-NEXT: [[CMP:%.*]] = icmp slt i32 [[TMP2]], [[TMP3]] +// CHECK-NEXT: [[STOREDV:%.*]] = zext i1 [[CMP]] to i8 +// CHECK-NEXT: store i8 [[STOREDV]], ptr [[DOTMIN_LESS_MAX]], align 1 +// CHECK-NEXT: [[TMP4:%.*]] = load i8, ptr [[DOTMIN_LESS_MAX]], align 1 +// CHECK-NEXT: [[LOADEDV:%.*]] = trunc i8 [[TMP4]] to i1 +// CHECK-NEXT: br i1 [[LOADEDV]], label %[[COND_TRUE:.*]], label %[[COND_FALSE:.*]] +// CHECK: [[COND_TRUE]]: +// CHECK-NEXT: [[TMP5:%.*]] = load i32, ptr [[DOTLB_MIN]], align 4 +// CHECK-NEXT: br label %[[COND_END:.*]] +// CHECK: [[COND_FALSE]]: +// CHECK-NEXT: [[TMP6:%.*]] = load i32, ptr [[DOTLB_MAX]], align 4 +// CHECK-NEXT: br label %[[COND_END]] +// CHECK: [[COND_END]]: +// CHECK-NEXT: [[COND:%.*]] = phi i32 [ [[TMP5]], %[[COND_TRUE]] ], [ [[TMP6]], %[[COND_FALSE]] ] +// CHECK-NEXT: store i32 [[COND]], ptr [[TMP]], align 4 +// CHECK-NEXT: store i32 [[COND]], ptr [[DOTLOWER]], align 4 +// CHECK-NEXT: [[TMP7:%.*]] = load i32, ptr [[TMP]], align 4 +// CHECK-NEXT: [[ADD5:%.*]] = add i32 [[TMP7]], 1 +// CHECK-NEXT: store i32 [[ADD5]], ptr [[_TMP1]], align 4 +// CHECK-NEXT: [[TMP8:%.*]] = load i32, ptr [[_TMP1]], align 4 +// CHECK-NEXT: [[ADD6:%.*]] = add nsw i32 [[TMP8]], 1 +// CHECK-NEXT: store i32 [[ADD6]], ptr [[DOTLB_MIN4]], align 4 +// CHECK-NEXT: [[TMP9:%.*]] = load i32, ptr [[TMP]], align 4 +// CHECK-NEXT: [[ADD8:%.*]] = add i32 [[TMP9]], 1 +// CHECK-NEXT: [[TMP10:%.*]] = load i32, ptr [[TMP]], align 4 +// CHECK-NEXT: [[ADD9:%.*]] = add i32 [[TMP10]], 1 +// CHECK-NEXT: [[SUB:%.*]] = sub i32 10, [[ADD9]] +// CHECK-NEXT: [[ADD10:%.*]] = add i32 [[SUB]], 1 +// CHECK-NEXT: [[MUL:%.*]] = mul i32 [[ADD10]], 2 +// CHECK-NEXT: [[ADD11:%.*]] = add i32 [[ADD8]], [[MUL]] +// CHECK-NEXT: store i32 [[ADD11]], ptr [[_TMP1]], align 4 +// CHECK-NEXT: [[TMP11:%.*]] = load i32, ptr [[_TMP1]], align 4 +// CHECK-NEXT: [[ADD12:%.*]] = add nsw i32 [[TMP11]], 1 +// CHECK-NEXT: store i32 [[ADD12]], ptr [[DOTLB_MAX7]], align 4 +// CHECK-NEXT: [[TMP12:%.*]] = load i32, ptr [[DOTLB_MIN4]], align 4 +// CHECK-NEXT: [[TMP13:%.*]] = load i32, ptr [[DOTLB_MAX7]], align 4 +// CHECK-NEXT: [[CMP14:%.*]] = icmp ult i32 [[TMP12]], [[TMP13]] +// CHECK-NEXT: [[STOREDV15:%.*]] = zext i1 [[CMP14]] to i8 +// CHECK-NEXT: store i8 [[STOREDV15]], ptr [[DOTMIN_LESS_MAX13]], align 1 +// CHECK-NEXT: [[TMP14:%.*]] = load i8, ptr [[DOTMIN_LESS_MAX13]], align 1 +// CHECK-NEXT: [[LOADEDV17:%.*]] = trunc i8 [[TMP14]] to i1 +// CHECK-NEXT: br i1 [[LOADEDV17]], label %[[COND_TRUE18:.*]], label %[[COND_FALSE19:.*]] +// CHECK: [[COND_TRUE18]]: +// CHECK-NEXT: [[TMP15:%.*]] = load i32, ptr [[DOTLB_MIN4]], align 4 +// CHECK-NEXT: br label %[[COND_END20:.*]] +// CHECK: [[COND_FALSE19]]: +// CHECK-NEXT: [[TMP16:%.*]] = load i32, ptr [[DOTLB_MAX7]], align 4 +// CHECK-NEXT: br label %[[COND_END20]] +// CHECK: [[COND_END20]]: +// CHECK-NEXT: [[COND21:%.*]] = phi i32 [ [[TMP15]], %[[COND_TRUE18]] ], [ [[TMP16]], %[[COND_FALSE19]] ] +// CHECK-NEXT: store i32 [[COND21]], ptr [[_TMP1]], align 4 +// CHECK-NEXT: store i32 [[COND21]], ptr [[DOTLOWER16]], align 4 +// CHECK-NEXT: [[TMP17:%.*]] = load i32, ptr [[DOTLOWER]], align 4 +// CHECK-NEXT: [[SUB22:%.*]] = sub i32 10, [[TMP17]] +// CHECK-NEXT: [[ADD23:%.*]] = add i32 [[SUB22]], 1 +// CHECK-NEXT: [[CONV:%.*]] = zext i32 [[ADD23]] to i64 +// CHECK-NEXT: [[MUL24:%.*]] = mul nsw i64 10, [[CONV]] +// CHECK-NEXT: [[TMP18:%.*]] = load i32, ptr [[DOTLOWER16]], align 4 +// CHECK-NEXT: [[SUB25:%.*]] = sub i32 10, [[TMP18]] +// CHECK-NEXT: [[ADD26:%.*]] = add i32 [[SUB25]], 1 +// CHECK-NEXT: [[CONV27:%.*]] = zext i32 [[ADD26]] to i64 +// CHECK-NEXT: [[MUL28:%.*]] = mul nsw i64 [[MUL24]], [[CONV27]] +// CHECK-NEXT: [[SUB29:%.*]] = sub nsw i64 [[MUL28]], 1 +// CHECK-NEXT: store i64 [[SUB29]], ptr [[DOTCAPTURE_EXPR_]], align 8 +// CHECK-NEXT: store i64 0, ptr [[DOTOMP_LB]], align 8 +// CHECK-NEXT: [[TMP19:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_]], align 8 +// CHECK-NEXT: store i64 [[TMP19]], ptr [[DOTOMP_UB]], align 8 +// CHECK-NEXT: store i64 1, ptr [[DOTOMP_STRIDE]], align 8 +// CHECK-NEXT: store i32 0, ptr [[DOTOMP_IS_LAST]], align 4 +// CHECK-NEXT: [[TMP20:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8 +// CHECK-NEXT: [[TMP21:%.*]] = load i32, ptr [[TMP20]], align 4 +// CHECK-NEXT: call void @__kmpc_for_static_init_8(ptr @[[GLOB1]], i32 [[TMP21]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i64 1, i64 1) +// CHECK-NEXT: [[TMP22:%.*]] = load i64, ptr [[DOTOMP_UB]], align 8 +// CHECK-NEXT: [[TMP23:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_]], align 8 +// CHECK-NEXT: [[CMP30:%.*]] = icmp sgt i64 [[TMP22]], [[TMP23]] +// CHECK-NEXT: br i1 [[CMP30]], label %[[COND_TRUE31:.*]], label %[[COND_FALSE32:.*]] +// CHECK: [[COND_TRUE31]]: +// CHECK-NEXT: [[TMP24:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_]], align 8 +// CHECK-NEXT: br label %[[COND_END33:.*]] +// CHECK: [[COND_FALSE32]]: +// CHECK-NEXT: [[TMP25:%.*]] = load i64, ptr [[DOTOMP_UB]], align 8 +// CHECK-NEXT: br label %[[COND_END33]] +// CHECK: [[COND_END33]]: +// CHECK-NEXT: [[COND34:%.*]] = phi i64 [ [[TMP24]], %[[COND_TRUE31]] ], [ [[TMP25]], %[[COND_FALSE32]] ] +// CHECK-NEXT: store i64 [[COND34]], ptr [[DOTOMP_UB]], align 8 +// CHECK-NEXT: [[TMP26:%.*]] = load i64, ptr [[DOTOMP_LB]], align 8 +// CHECK-NEXT: store i64 [[TMP26]], ptr [[DOTOMP_IV]], align 8 +// CHECK-NEXT: br label %[[OMP_INNER_FOR_COND:.*]] +// CHECK: [[OMP_INNER_FOR_COND]]: +// CHECK-NEXT: [[TMP27:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8 +// CHECK-NEXT: [[TMP28:%.*]] = load i64, ptr [[DOTOMP_UB]], align 8 +// CHECK-NEXT: [[CMP35:%.*]] = icmp sle i64 [[TMP27]], [[TMP28]] +// CHECK-NEXT: br i1 [[CMP35]], label %[[OMP_INNER_FOR_BODY:.*]], label %[[OMP_INNER_FOR_END:.*]] +// CHECK: [[OMP_INNER_FOR_BODY]]: +// CHECK-NEXT: [[TMP29:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8 +// CHECK-NEXT: [[TMP30:%.*]] = load i32, ptr [[DOTLOWER]], align 4 +// CHECK-NEXT: [[SUB36:%.*]] = sub i32 10, [[TMP30]] +// CHECK-NEXT: [[ADD37:%.*]] = add i32 [[SUB36]], 1 +// CHECK-NEXT: [[MUL38:%.*]] = mul i32 1, [[ADD37]] +// CHECK-NEXT: [[TMP31:%.*]] = load i32, ptr [[DOTLOWER16]], align 4 +// CHECK-NEXT: [[SUB39:%.*]] = sub i32 10, [[TMP31]] +// CHECK-NEXT: [[ADD40:%.*]] = add i32 [[SUB39]], 1 +// CHECK-NEXT: [[MUL41:%.*]] = mul i32 [[MUL38]], [[ADD40]] +// CHECK-NEXT: [[CONV42:%.*]] = zext i32 [[MUL41]] to i64 +// CHECK-NEXT: [[DIV:%.*]] = sdiv i64 [[TMP29]], [[CONV42]] +// CHECK-NEXT: [[MUL43:%.*]] = mul nsw i64 [[DIV]], 1 +// CHECK-NEXT: [[ADD44:%.*]] = add nsw i64 0, [[MUL43]] +// CHECK-NEXT: [[CONV45:%.*]] = trunc i64 [[ADD44]] to i32 +// CHECK-NEXT: store i32 [[CONV45]], ptr [[I]], align 4 +// CHECK-NEXT: [[TMP32:%.*]] = load i32, ptr [[I]], align 4 +// CHECK-NEXT: [[ADD46:%.*]] = add i32 [[TMP32]], 1 +// CHECK-NEXT: [[CONV47:%.*]] = sext i32 [[ADD46]] to i64 +// CHECK-NEXT: [[TMP33:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8 +// CHECK-NEXT: [[TMP34:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8 +// CHECK-NEXT: [[TMP35:%.*]] = load i32, ptr [[DOTLOWER]], align 4 +// CHECK-NEXT: [[SUB48:%.*]] = sub i32 10, [[TMP35]] +// CHECK-NEXT: [[ADD49:%.*]] = add i32 [[SUB48]], 1 +// CHECK-NEXT: [[MUL50:%.*]] = mul i32 1, [[ADD49]] +// CHECK-NEXT: [[TMP36:%.*]] = load i32, ptr [[DOTLOWER16]], align 4 +// CHECK-NEXT: [[SUB51:%.*]] = sub i32 10, [[TMP36]] +// CHECK-NEXT: [[ADD52:%.*]] = add i32 [[SUB51]], 1 +// CHECK-NEXT: [[MUL53:%.*]] = mul i32 [[MUL50]], [[ADD52]] +// CHECK-NEXT: [[CONV54:%.*]] = zext i32 [[MUL53]] to i64 +// CHECK-NEXT: [[DIV55:%.*]] = sdiv i64 [[TMP34]], [[CONV54]] +// CHECK-NEXT: [[TMP37:%.*]] = load i32, ptr [[DOTLOWER]], align 4 +// CHECK-NEXT: [[SUB56:%.*]] = sub i32 10, [[TMP37]] +// CHECK-NEXT: [[ADD57:%.*]] = add i32 [[SUB56]], 1 +// CHECK-NEXT: [[MUL58:%.*]] = mul i32 1, [[ADD57]] +// CHECK-NEXT: [[TMP38:%.*]] = load i32, ptr [[DOTLOWER16]], align 4 +// CHECK-NEXT: [[SUB59:%.*]] = sub i32 10, [[TMP38]] +// CHECK-NEXT: [[ADD60:%.*]] = add i32 [[SUB59]], 1 +// CHECK-NEXT: [[MUL61:%.*]] = mul i32 [[MUL58]], [[ADD60]] +// CHECK-NEXT: [[CONV62:%.*]] = zext i32 [[MUL61]] to i64 +// CHECK-NEXT: [[MUL63:%.*]] = mul nsw i64 [[DIV55]], [[CONV62]] +// CHECK-NEXT: [[SUB64:%.*]] = sub nsw i64 [[TMP33]], [[MUL63]] +// CHECK-NEXT: [[TMP39:%.*]] = load i32, ptr [[DOTLOWER16]], align 4 +// CHECK-NEXT: [[SUB65:%.*]] = sub i32 10, [[TMP39]] +// CHECK-NEXT: [[ADD66:%.*]] = add i32 [[SUB65]], 1 +// CHECK-NEXT: [[MUL67:%.*]] = mul i32 1, [[ADD66]] +// CHECK-NEXT: [[CONV68:%.*]] = zext i32 [[MUL67]] to i64 +// CHECK-NEXT: [[DIV69:%.*]] = sdiv i64 [[SUB64]], [[CONV68]] +// CHECK-NEXT: [[MUL70:%.*]] = mul nsw i64 [[DIV69]], 2 +// CHECK-NEXT: [[ADD71:%.*]] = add nsw i64 [[CONV47]], [[MUL70]] +// CHECK-NEXT: [[CONV72:%.*]] = trunc i64 [[ADD71]] to i32 +// CHECK-NEXT: store i32 [[CONV72]], ptr [[J]], align 4 +// CHECK-NEXT: [[TMP40:%.*]] = load i32, ptr [[J]], align 4 +// CHECK-NEXT: [[ADD73:%.*]] = add nsw i32 [[TMP40]], 1 +// CHECK-NEXT: [[CONV74:%.*]] = zext i32 [[ADD73]] to i64 +// CHECK-NEXT: [[TMP41:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8 +// CHECK-NEXT: [[TMP42:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8 +// CHECK-NEXT: [[TMP43:%.*]] = load i32, ptr [[DOTLOWER]], align 4 +// CHECK-NEXT: [[SUB75:%.*]] = sub i32 10, [[TMP43]] +// CHECK-NEXT: [[ADD76:%.*]] = add i32 [[SUB75]], 1 +// CHECK-NEXT: [[MUL77:%.*]] = mul i32 1, [[ADD76]] +// CHECK-NEXT: [[TMP44:%.*]] = load i32, ptr [[DOTLOWER16]], align 4 +// CHECK-NEXT: [[SUB78:%.*]] = sub i32 10, [[TMP44]] +// CHECK-NEXT: [[ADD79:%.*]] = add i32 [[SUB78]], 1 +// CHECK-NEXT: [[MUL80:%.*]] = mul i32 [[MUL77]], [[ADD79]] +// CHECK-NEXT: [[CONV81:%.*]] = zext i32 [[MUL80]] to i64 +// CHECK-NEXT: [[DIV82:%.*]] = sdiv i64 [[TMP42]], [[CONV81]] +// CHECK-NEXT: [[TMP45:%.*]] = load i32, ptr [[DOTLOWER]], align 4 +// CHECK-NEXT: [[SUB83:%.*]] = sub i32 10, [[TMP45]] +// CHECK-NEXT: [[ADD84:%.*]] = add i32 [[SUB83]], 1 +// CHECK-NEXT: [[MUL85:%.*]] = mul i32 1, [[ADD84]] +// CHECK-NEXT: [[TMP46:%.*]] = load i32, ptr [[DOTLOWER16]], align 4 +// CHECK-NEXT: [[SUB86:%.*]] = sub i32 10, [[TMP46]] +// CHECK-NEXT: [[ADD87:%.*]] = add i32 [[SUB86]], 1 +// CHECK-NEXT: [[MUL88:%.*]] = mul i32 [[MUL85]], [[ADD87]] +// CHECK-NEXT: [[CONV89:%.*]] = zext i32 [[MUL88]] to i64 +// CHECK-NEXT: [[MUL90:%.*]] = mul nsw i64 [[DIV82]], [[CONV89]] +// CHECK-NEXT: [[SUB91:%.*]] = sub nsw i64 [[TMP41]], [[MUL90]] +// CHECK-NEXT: [[TMP47:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8 +// CHECK-NEXT: [[TMP48:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8 +// CHECK-NEXT: [[TMP49:%.*]] = load i32, ptr [[DOTLOWER]], align 4 +// CHECK-NEXT: [[SUB92:%.*]] = sub i32 10, [[TMP49]] +// CHECK-NEXT: [[ADD93:%.*]] = add i32 [[SUB92]], 1 +// CHECK-NEXT: [[MUL94:%.*]] = mul i32 1, [[ADD93]] +// CHECK-NEXT: [[TMP50:%.*]] = load i32, ptr [[DOTLOWER16]], align 4 +// CHECK-NEXT: [[SUB95:%.*]] = sub i32 10, [[TMP50]] +// CHECK-NEXT: [[ADD96:%.*]] = add i32 [[SUB95]], 1 +// CHECK-NEXT: [[MUL97:%.*]] = mul i32 [[MUL94]], [[ADD96]] +// CHECK-NEXT: [[CONV98:%.*]] = zext i32 [[MUL97]] to i64 +// CHECK-NEXT: [[DIV99:%.*]] = sdiv i64 [[TMP48]], [[CONV98]] +// CHECK-NEXT: [[TMP51:%.*]] = load i32, ptr [[DOTLOWER]], align 4 +// CHECK-NEXT: [[SUB100:%.*]] = sub i32 10, [[TMP51]] +// CHECK-NEXT: [[ADD101:%.*]] = add i32 [[SUB100]], 1 +// CHECK-NEXT: [[MUL102:%.*]] = mul i32 1, [[ADD101]] +// CHECK-NEXT: [[TMP52:%.*]] = load i32, ptr [[DOTLOWER16]], align 4 +// CHECK-NEXT: [[SUB103:%.*]] = sub i32 10, [[TMP52]] +// CHECK-NEXT: [[ADD104:%.*]] = add i32 [[SUB103]], 1 +// CHECK-NEXT: [[MUL105:%.*]] = mul i32 [[MUL102]], [[ADD104]] +// CHECK-NEXT: [[CONV106:%.*]] = zext i32 [[MUL105]] to i64 +// CHECK-NEXT: [[MUL107:%.*]] = mul nsw i64 [[DIV99]], [[CONV106]] +// CHECK-NEXT: [[SUB108:%.*]] = sub nsw i64 [[TMP47]], [[MUL107]] +// CHECK-NEXT: [[TMP53:%.*]] = load i32, ptr [[DOTLOWER16]], align 4 +// CHECK-NEXT: [[SUB109:%.*]] = sub i32 10, [[TMP53]] +// CHECK-NEXT: [[ADD110:%.*]] = add i32 [[SUB109]], 1 +// CHECK-NEXT: [[MUL111:%.*]] = mul i32 1, [[ADD110]] +// CHECK-NEXT: [[CONV112:%.*]] = zext i32 [[MUL111]] to i64 +// CHECK-NEXT: [[DIV113:%.*]] = sdiv i64 [[SUB108]], [[CONV112]] +// CHECK-NEXT: [[TMP54:%.*]] = load i32, ptr [[DOTLOWER16]], align 4 +// CHECK-NEXT: [[SUB114:%.*]] = sub i32 10, [[TMP54]] +// CHECK-NEXT: [[ADD115:%.*]] = add i32 [[SUB114]], 1 +// CHECK-NEXT: [[MUL116:%.*]] = mul i32 1, [[ADD115]] +// CHECK-NEXT: [[CONV117:%.*]] = zext i32 [[MUL116]] to i64 +// CHECK-NEXT: [[MUL118:%.*]] = mul nsw i64 [[DIV113]], [[CONV117]] +// CHECK-NEXT: [[SUB119:%.*]] = sub nsw i64 [[SUB91]], [[MUL118]] +// CHECK-NEXT: [[MUL120:%.*]] = mul nsw i64 [[SUB119]], 1 +// CHECK-NEXT: [[ADD121:%.*]] = add nsw i64 [[CONV74]], [[MUL120]] +// CHECK-NEXT: [[CONV122:%.*]] = trunc i64 [[ADD121]] to i32 +// CHECK-NEXT: store i32 [[CONV122]], ptr [[K]], align 4 +// CHECK-NEXT: [[TMP55:%.*]] = load i32, ptr [[J]], align 4 +// CHECK-NEXT: [[CMP123:%.*]] = icmp slt i32 [[TMP55]], 10 +// CHECK-NEXT: br i1 [[CMP123]], label %[[OMP_BODY_NEXT:.*]], label %[[OMP_BODY_CONTINUE:.*]] +// CHECK: [[OMP_BODY_NEXT]]: +// CHECK-NEXT: [[TMP56:%.*]] = load i32, ptr [[K]], align 4 +// CHECK-NEXT: [[CMP124:%.*]] = icmp ult i32 [[TMP56]], 10 +// CHECK-NEXT: br i1 [[CMP124]], label %[[OMP_BODY_NEXT125:.*]], label %[[OMP_BODY_CONTINUE]] +// CHECK: [[OMP_BODY_NEXT125]]: +// CHECK-NEXT: br label %[[OMP_BODY_CONTINUE]] +// CHECK: [[OMP_BODY_CONTINUE]]: +// CHECK-NEXT: br label %[[OMP_INNER_FOR_INC:.*]] +// CHECK: [[OMP_INNER_FOR_INC]]: +// CHECK-NEXT: [[TMP57:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8 +// CHECK-NEXT: [[ADD126:%.*]] = add nsw i64 [[TMP57]], 1 +// CHECK-NEXT: store i64 [[ADD126]], ptr [[DOTOMP_IV]], align 8 +// CHECK-NEXT: br label %[[OMP_INNER_FOR_COND]] +// CHECK: [[OMP_INNER_FOR_END]]: +// CHECK-NEXT: br label %[[OMP_LOOP_EXIT:.*]] +// CHECK: [[OMP_LOOP_EXIT]]: +// CHECK-NEXT: [[TMP58:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8 +// CHECK-NEXT: [[TMP59:%.*]] = load i32, ptr [[TMP58]], align 4 +// CHECK-NEXT: call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP59]]) +// CHECK-NEXT: ret void +void triangulat_loop_2() { +#pragma omp parallel for collapse(3) + for (unsigned int i = 0; i < 10; ++i) + for (int j = i + 1; j < 10; j += 2) + for (unsigned int k = j + 1; k < 10; ++k) + ; +} + +// CHECK-LABEL: define internal void @_Z10mixed_loopv.omp_outlined( +// CHECK-SAME: ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8 +// CHECK-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8 +// CHECK-NEXT: [[DOTOMP_IV:%.*]] = alloca i64, align 8 +// CHECK-NEXT: [[TMP:%.*]] = alloca i32, align 4 +// CHECK-NEXT: [[_TMP1:%.*]] = alloca i32, align 4 +// CHECK-NEXT: [[_TMP2:%.*]] = alloca i32, align 4 +// CHECK-NEXT: [[DOTLB_MIN:%.*]] = alloca i32, align 4 +// CHECK-NEXT: [[DOTLB_MAX:%.*]] = alloca i32, align 4 +// CHECK-NEXT: [[DOTMIN_LESS_MAX:%.*]] = alloca i8, align 1 +// CHECK-NEXT: [[DOTLOWER:%.*]] = alloca i32, align 4 +// CHECK-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i64, align 8 +// CHECK-NEXT: [[DOTOMP_LB:%.*]] = alloca i64, align 8 +// CHECK-NEXT: [[DOTOMP_UB:%.*]] = alloca i64, align 8 +// CHECK-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8 +// CHECK-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4 +// CHECK-NEXT: [[I:%.*]] = alloca i32, align 4 +// CHECK-NEXT: [[J:%.*]] = alloca i32, align 4 +// CHECK-NEXT: [[K:%.*]] = alloca i32, align 4 +// CHECK-NEXT: store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8 +// CHECK-NEXT: store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8 +// CHECK-NEXT: store i32 0, ptr [[_TMP1]], align 4 +// CHECK-NEXT: [[TMP0:%.*]] = load i32, ptr [[_TMP1]], align 4 +// CHECK-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP0]], 1 +// CHECK-NEXT: store i32 [[ADD]], ptr [[DOTLB_MIN]], align 4 +// CHECK-NEXT: store i32 9, ptr [[_TMP1]], align 4 +// CHECK-NEXT: [[TMP1:%.*]] = load i32, ptr [[_TMP1]], align 4 +// CHECK-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP1]], 1 +// CHECK-NEXT: store i32 [[ADD3]], ptr [[DOTLB_MAX]], align 4 +// CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr [[DOTLB_MIN]], align 4 +// CHECK-NEXT: [[TMP3:%.*]] = load i32, ptr [[DOTLB_MAX]], align 4 +// CHECK-NEXT: [[CMP:%.*]] = icmp slt i32 [[TMP2]], [[TMP3]] +// CHECK-NEXT: [[STOREDV:%.*]] = zext i1 [[CMP]] to i8 +// CHECK-NEXT: store i8 [[STOREDV]], ptr [[DOTMIN_LESS_MAX]], align 1 +// CHECK-NEXT: [[TMP4:%.*]] = load i8, ptr [[DOTMIN_LESS_MAX]], align 1 +// CHECK-NEXT: [[LOADEDV:%.*]] = trunc i8 [[TMP4]] to i1 +// CHECK-NEXT: br i1 [[LOADEDV]], label %[[COND_TRUE:.*]], label %[[COND_FALSE:.*]] +// CHECK: [[COND_TRUE]]: +// CHECK-NEXT: [[TMP5:%.*]] = load i32, ptr [[DOTLB_MIN]], align 4 +// CHECK-NEXT: br label %[[COND_END:.*]] +// CHECK: [[COND_FALSE]]: +// CHECK-NEXT: [[TMP6:%.*]] = load i32, ptr [[DOTLB_MAX]], align 4 +// CHECK-NEXT: br label %[[COND_END]] +// CHECK: [[COND_END]]: +// CHECK-NEXT: [[COND:%.*]] = phi i32 [ [[TMP5]], %[[COND_TRUE]] ], [ [[TMP6]], %[[COND_FALSE]] ] +// CHECK-NEXT: store i32 [[COND]], ptr [[_TMP1]], align 4 +// CHECK-NEXT: store i32 [[COND]], ptr [[DOTLOWER]], align 4 +// CHECK-NEXT: [[TMP7:%.*]] = load i32, ptr [[DOTLOWER]], align 4 +// CHECK-NEXT: [[SUB:%.*]] = sub i32 10, [[TMP7]] +// CHECK-NEXT: [[ADD4:%.*]] = add i32 [[SUB]], 1 +// CHECK-NEXT: [[CONV:%.*]] = zext i32 [[ADD4]] to i64 +// CHECK-NEXT: [[MUL:%.*]] = mul nsw i64 100, [[CONV]] +// CHECK-NEXT: [[SUB5:%.*]] = sub nsw i64 [[MUL]], 1 +// CHECK-NEXT: store i64 [[SUB5]], ptr [[DOTCAPTURE_EXPR_]], align 8 +// CHECK-NEXT: store i64 0, ptr [[DOTOMP_LB]], align 8 +// CHECK-NEXT: [[TMP8:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_]], align 8 +// CHECK-NEXT: store i64 [[TMP8]], ptr [[DOTOMP_UB]], align 8 +// CHECK-NEXT: store i64 1, ptr [[DOTOMP_STRIDE]], align 8 +// CHECK-NEXT: store i32 0, ptr [[DOTOMP_IS_LAST]], align 4 +// CHECK-NEXT: [[TMP9:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8 +// CHECK-NEXT: [[TMP10:%.*]] = load i32, ptr [[TMP9]], align 4 +// CHECK-NEXT: call void @__kmpc_for_static_init_8(ptr @[[GLOB1]], i32 [[TMP10]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i64 1, i64 1) +// CHECK-NEXT: [[TMP11:%.*]] = load i64, ptr [[DOTOMP_UB]], align 8 +// CHECK-NEXT: [[TMP12:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_]], align 8 +// CHECK-NEXT: [[CMP6:%.*]] = icmp sgt i64 [[TMP11]], [[TMP12]] +// CHECK-NEXT: br i1 [[CMP6]], label %[[COND_TRUE7:.*]], label %[[COND_FALSE8:.*]] +// CHECK: [[COND_TRUE7]]: +// CHECK-NEXT: [[TMP13:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_]], align 8 +// CHECK-NEXT: br label %[[COND_END9:.*]] +// CHECK: [[COND_FALSE8]]: +// CHECK-NEXT: [[TMP14:%.*]] = load i64, ptr [[DOTOMP_UB]], align 8 +// CHECK-NEXT: br label %[[COND_END9]] +// CHECK: [[COND_END9]]: +// CHECK-NEXT: [[COND10:%.*]] = phi i64 [ [[TMP13]], %[[COND_TRUE7]] ], [ [[TMP14]], %[[COND_FALSE8]] ] +// CHECK-NEXT: store i64 [[COND10]], ptr [[DOTOMP_UB]], align 8 +// CHECK-NEXT: [[TMP15:%.*]] = load i64, ptr [[DOTOMP_LB]], align 8 +// CHECK-NEXT: store i64 [[TMP15]], ptr [[DOTOMP_IV]], align 8 +// CHECK-NEXT: br label %[[OMP_INNER_FOR_COND:.*]] +// CHECK: [[OMP_INNER_FOR_COND]]: +// CHECK-NEXT: [[TMP16:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8 +// CHECK-NEXT: [[TMP17:%.*]] = load i64, ptr [[DOTOMP_UB]], align 8 +// CHECK-NEXT: [[CMP11:%.*]] = icmp sle i64 [[TMP16]], [[TMP17]] +// CHECK-NEXT: br i1 [[CMP11]], label %[[OMP_INNER_FOR_BODY:.*]], label %[[OMP_INNER_FOR_END:.*]] +// CHECK: [[OMP_INNER_FOR_BODY]]: +// CHECK-NEXT: [[TMP18:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8 +// CHECK-NEXT: [[TMP19:%.*]] = load i32, ptr [[DOTLOWER]], align 4 +// CHECK-NEXT: [[SUB12:%.*]] = sub i32 10, [[TMP19]] +// CHECK-NEXT: [[ADD13:%.*]] = add i32 [[SUB12]], 1 +// CHECK-NEXT: [[MUL14:%.*]] = mul i32 10, [[ADD13]] +// CHECK-NEXT: [[CONV15:%.*]] = zext i32 [[MUL14]] to i64 +// CHECK-NEXT: [[DIV:%.*]] = sdiv i64 [[TMP18]], [[CONV15]] +// CHECK-NEXT: [[MUL16:%.*]] = mul nsw i64 [[DIV]], 1 +// CHECK-NEXT: [[ADD17:%.*]] = add nsw i64 0, [[MUL16]] +// CHECK-NEXT: [[CONV18:%.*]] = trunc i64 [[ADD17]] to i32 +// CHECK-NEXT: store i32 [[CONV18]], ptr [[I]], align 4 +// CHECK-NEXT: [[TMP20:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8 +// CHECK-NEXT: [[TMP21:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8 +// CHECK-NEXT: [[TMP22:%.*]] = load i32, ptr [[DOTLOWER]], align 4 +// CHECK-NEXT: [[SUB19:%.*]] = sub i32 10, [[TMP22]] +// CHECK-NEXT: [[ADD20:%.*]] = add i32 [[SUB19]], 1 +// CHECK-NEXT: [[MUL21:%.*]] = mul i32 10, [[ADD20]] +// CHECK-NEXT: [[CONV22:%.*]] = zext i32 [[MUL21]] to i64 +// CHECK-NEXT: [[DIV23:%.*]] = sdiv i64 [[TMP21]], [[CONV22]] +// CHECK-NEXT: [[TMP23:%.*]] = load i32, ptr [[DOTLOWER]], align 4 +// CHECK-NEXT: [[SUB24:%.*]] = sub i32 10, [[TMP23]] +// CHECK-NEXT: [[ADD25:%.*]] = add i32 [[SUB24]], 1 +// CHECK-NEXT: [[MUL26:%.*]] = mul i32 10, [[ADD25]] +// CHECK-NEXT: [[CONV27:%.*]] = zext i32 [[MUL26]] to i64 +// CHECK-NEXT: [[MUL28:%.*]] = mul nsw i64 [[DIV23]], [[CONV27]] +// CHECK-NEXT: [[SUB29:%.*]] = sub nsw i64 [[TMP20]], [[MUL28]] +// CHECK-NEXT: [[TMP24:%.*]] = load i32, ptr [[DOTLOWER]], align 4 +// CHECK-NEXT: [[SUB30:%.*]] = sub i32 10, [[TMP24]] +// CHECK-NEXT: [[ADD31:%.*]] = add i32 [[SUB30]], 1 +// CHECK-NEXT: [[MUL32:%.*]] = mul i32 1, [[ADD31]] +// CHECK-NEXT: [[CONV33:%.*]] = zext i32 [[MUL32]] to i64 +// CHECK-NEXT: [[DIV34:%.*]] = sdiv i64 [[SUB29]], [[CONV33]] +// CHECK-NEXT: [[MUL35:%.*]] = mul nsw i64 [[DIV34]], 1 +// CHECK-NEXT: [[ADD36:%.*]] = add nsw i64 0, [[MUL35]] +// CHECK-NEXT: [[CONV37:%.*]] = trunc i64 [[ADD36]] to i32 +// CHECK-NEXT: store i32 [[CONV37]], ptr [[J]], align 4 +// CHECK-NEXT: [[TMP25:%.*]] = load i32, ptr [[J]], align 4 +// CHECK-NEXT: [[ADD38:%.*]] = add nsw i32 [[TMP25]], 1 +// CHECK-NEXT: [[CONV39:%.*]] = sext i32 [[ADD38]] to i64 +// CHECK-NEXT: [[TMP26:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8 +// CHECK-NEXT: [[TMP27:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8 +// CHECK-NEXT: [[TMP28:%.*]] = load i32, ptr [[DOTLOWER]], align 4 +// CHECK-NEXT: [[SUB40:%.*]] = sub i32 10, [[TMP28]] +// CHECK-NEXT: [[ADD41:%.*]] = add i32 [[SUB40]], 1 +// CHECK-NEXT: [[MUL42:%.*]] = mul i32 10, [[ADD41]] +// CHECK-NEXT: [[CONV43:%.*]] = zext i32 [[MUL42]] to i64 +// CHECK-NEXT: [[DIV44:%.*]] = sdiv i64 [[TMP27]], [[CONV43]] +// CHECK-NEXT: [[TMP29:%.*]] = load i32, ptr [[DOTLOWER]], align 4 +// CHECK-NEXT: [[SUB45:%.*]] = sub i32 10, [[TMP29]] +// CHECK-NEXT: [[ADD46:%.*]] = add i32 [[SUB45]], 1 +// CHECK-NEXT: [[MUL47:%.*]] = mul i32 10, [[ADD46]] +// CHECK-NEXT: [[CONV48:%.*]] = zext i32 [[MUL47]] to i64 +// CHECK-NEXT: [[MUL49:%.*]] = mul nsw i64 [[DIV44]], [[CONV48]] +// CHECK-NEXT: [[SUB50:%.*]] = sub nsw i64 [[TMP26]], [[MUL49]] +// CHECK-NEXT: [[TMP30:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8 +// CHECK-NEXT: [[TMP31:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8 +// CHECK-NEXT: [[TMP32:%.*]] = load i32, ptr [[DOTLOWER]], align 4 +// CHECK-NEXT: [[SUB51:%.*]] = sub i32 10, [[TMP32]] +// CHECK-NEXT: [[ADD52:%.*]] = add i32 [[SUB51]], 1 +// CHECK-NEXT: [[MUL53:%.*]] = mul i32 10, [[ADD52]] +// CHECK-NEXT: [[CONV54:%.*]] = zext i32 [[MUL53]] to i64 +// CHECK-NEXT: [[DIV55:%.*]] = sdiv i64 [[TMP31]], [[CONV54]] +// CHECK-NEXT: [[TMP33:%.*]] = load i32, ptr [[DOTLOWER]], align 4 +// CHECK-NEXT: [[SUB56:%.*]] = sub i32 10, [[TMP33]] +// CHECK-NEXT: [[ADD57:%.*]] = add i32 [[SUB56]], 1 +// CHECK-NEXT: [[MUL58:%.*]] = mul i32 10, [[ADD57]] +// CHECK-NEXT: [[CONV59:%.*]] = zext i32 [[MUL58]] to i64 +// CHECK-NEXT: [[MUL60:%.*]] = mul nsw i64 [[DIV55]], [[CONV59]] +// CHECK-NEXT: [[SUB61:%.*]] = sub nsw i64 [[TMP30]], [[MUL60]] +// CHECK-NEXT: [[TMP34:%.*]] = load i32, ptr [[DOTLOWER]], align 4 +// CHECK-NEXT: [[SUB62:%.*]] = sub i32 10, [[TMP34]] +// CHECK-NEXT: [[ADD63:%.*]] = add i32 [[SUB62]], 1 +// CHECK-NEXT: [[MUL64:%.*]] = mul i32 1, [[ADD63]] +// CHECK-NEXT: [[CONV65:%.*]] = zext i32 [[MUL64]] to i64 +// CHECK-NEXT: [[DIV66:%.*]] = sdiv i64 [[SUB61]], [[CONV65]] +// CHECK-NEXT: [[TMP35:%.*]] = load i32, ptr [[DOTLOWER]], align 4 +// CHECK-NEXT: [[SUB67:%.*]] = sub i32 10, [[TMP35]] +// CHECK-NEXT: [[ADD68:%.*]] = add i32 [[SUB67]], 1 +// CHECK-NEXT: [[MUL69:%.*]] = mul i32 1, [[ADD68]] +// CHECK-NEXT: [[CONV70:%.*]] = zext i32 [[MUL69]] to i64 +// CHECK-NEXT: [[MUL71:%.*]] = mul nsw i64 [[DIV66]], [[CONV70]] +// CHECK-NEXT: [[SUB72:%.*]] = sub nsw i64 [[SUB50]], [[MUL71]] +// CHECK-NEXT: [[MUL73:%.*]] = mul nsw i64 [[SUB72]], 1 +// CHECK-NEXT: [[ADD74:%.*]] = add nsw i64 [[CONV39]], [[MUL73]] +// CHECK-NEXT: [[CONV75:%.*]] = trunc i64 [[ADD74]] to i32 +// CHECK-NEXT: store i32 [[CONV75]], ptr [[K]], align 4 +// CHECK-NEXT: [[TMP36:%.*]] = load i32, ptr [[K]], align 4 +// CHECK-NEXT: [[CMP76:%.*]] = icmp slt i32 [[TMP36]], 10 +// CHECK-NEXT: br i1 [[CMP76]], label %[[OMP_BODY_NEXT:.*]], label %[[OMP_BODY_CONTINUE:.*]] +// CHECK: [[OMP_BODY_NEXT]]: +// CHECK-NEXT: br label %[[OMP_BODY_CONTINUE]] +// CHECK: [[OMP_BODY_CONTINUE]]: +// CHECK-NEXT: br label %[[OMP_INNER_FOR_INC:.*]] +// CHECK: [[OMP_INNER_FOR_INC]]: +// CHECK-NEXT: [[TMP37:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8 +// CHECK-NEXT: [[ADD77:%.*]] = add nsw i64 [[TMP37]], 1 +// CHECK-NEXT: store i64 [[ADD77]], ptr [[DOTOMP_IV]], align 8 +// CHECK-NEXT: br label %[[OMP_INNER_FOR_COND]] +// CHECK: [[OMP_INNER_FOR_END]]: +// CHECK-NEXT: br label %[[OMP_LOOP_EXIT:.*]] +// CHECK: [[OMP_LOOP_EXIT]]: +// CHECK-NEXT: [[TMP38:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8 +// CHECK-NEXT: [[TMP39:%.*]] = load i32, ptr [[TMP38]], align 4 +// CHECK-NEXT: call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP39]]) +// CHECK-NEXT: ret void +void mixed_loop() { +#pragma omp parallel for collapse(3) + for (int i = 0; i < 10; ++i) + for (int j = 0; j < 10; ++j) + for (int k = j + 1; k < 10; ++k) + ; +} + +// CHECK-LABEL: define internal void @_Z16rectangular_loopv.omp_outlined +// CHECK-SAME: ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8 +// CHECK-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8 +// CHECK-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4 +// CHECK-NEXT: [[TMP:%.*]] = alloca i32, align 4 +// CHECK-NEXT: [[_TMP1:%.*]] = alloca i32, align 4 +// CHECK-NEXT: [[_TMP2:%.*]] = alloca i32, align 4 +// CHECK-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4 +// CHECK-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4 +// CHECK-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4 +// CHECK-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4 +// CHECK-NEXT: [[I:%.*]] = alloca i32, align 4 +// CHECK-NEXT: [[J:%.*]] = alloca i32, align 4 +// CHECK-NEXT: [[K:%.*]] = alloca i32, align 4 +// CHECK-NEXT: store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8 +// CHECK-NEXT: store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8 +// CHECK-NEXT: store i32 0, ptr [[DOTOMP_LB]], align 4 +// CHECK-NEXT: store i32 999, ptr [[DOTOMP_UB]], align 4 +// CHECK-NEXT: store i32 1, ptr [[DOTOMP_STRIDE]], align 4 +// CHECK-NEXT: store i32 0, ptr [[DOTOMP_IS_LAST]], align 4 +// CHECK-NEXT: [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8 +// CHECK-NEXT: [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4 +// CHECK-NEXT: call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP1]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1) +// CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4 +// CHECK-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 999 +// CHECK-NEXT: br i1 [[CMP]], label %[[COND_TRUE:.*]], label %[[COND_FALSE:.*]] +// CHECK: [[COND_TRUE]]: +// CHECK-NEXT: br label %[[COND_END:.*]] +// CHECK: [[COND_FALSE]]: +// CHECK-NEXT: [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4 +// CHECK-NEXT: br label %[[COND_END]] +// CHECK: [[COND_END]]: +// CHECK-NEXT: [[COND:%.*]] = phi i32 [ 999, %[[COND_TRUE]] ], [ [[TMP3]], %[[COND_FALSE]] ] +// CHECK-NEXT: store i32 [[COND]], ptr [[DOTOMP_UB]], align 4 +// CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4 +// CHECK-NEXT: store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4 +// CHECK-NEXT: br label %[[OMP_INNER_FOR_COND:.*]] +// CHECK: [[OMP_INNER_FOR_COND]]: +// CHECK-NEXT: [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4 +// CHECK-NEXT: [[TMP6:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4 +// CHECK-NEXT: [[CMP3:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]] +// CHECK-NEXT: br i1 [[CMP3]], label %[[OMP_INNER_FOR_BODY:.*]], label %[[OMP_INNER_FOR_END:.*]] +// CHECK: [[OMP_INNER_FOR_BODY]]: +// CHECK-NEXT: [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4 +// CHECK-NEXT: [[DIV:%.*]] = sdiv i32 [[TMP7]], 100 +// CHECK-NEXT: [[MUL:%.*]] = mul nsw i32 [[DIV]], 1 +// CHECK-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]] +// CHECK-NEXT: store i32 [[ADD]], ptr [[I]], align 4 +// CHECK-NEXT: [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4 +// CHECK-NEXT: [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4 +// CHECK-NEXT: [[DIV4:%.*]] = sdiv i32 [[TMP9]], 100 +// CHECK-NEXT: [[MUL5:%.*]] = mul nsw i32 [[DIV4]], 100 +// CHECK-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP8]], [[MUL5]] +// CHECK-NEXT: [[DIV6:%.*]] = sdiv i32 [[SUB]], 10 +// CHECK-NEXT: [[MUL7:%.*]] = mul nsw i32 [[DIV6]], 1 +// CHECK-NEXT: [[ADD8:%.*]] = add nsw i32 0, [[MUL7]] +// CHECK-NEXT: store i32 [[ADD8]], ptr [[J]], align 4 +// CHECK-NEXT: [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4 +// CHECK-NEXT: [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4 +// CHECK-NEXT: [[DIV9:%.*]] = sdiv i32 [[TMP11]], 100 +// CHECK-NEXT: [[MUL10:%.*]] = mul nsw i32 [[DIV9]], 100 +// CHECK-NEXT: [[SUB11:%.*]] = sub nsw i32 [[TMP10]], [[MUL10]] +// CHECK-NEXT: [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4 +// CHECK-NEXT: [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4 +// CHECK-NEXT: [[DIV12:%.*]] = sdiv i32 [[TMP13]], 100 +// CHECK-NEXT: [[MUL13:%.*]] = mul nsw i32 [[DIV12]], 100 +// CHECK-NEXT: [[SUB14:%.*]] = sub nsw i32 [[TMP12]], [[MUL13]] +// CHECK-NEXT: [[DIV15:%.*]] = sdiv i32 [[SUB14]], 10 +// CHECK-NEXT: [[MUL16:%.*]] = mul nsw i32 [[DIV15]], 10 +// CHECK-NEXT: [[SUB17:%.*]] = sub nsw i32 [[SUB11]], [[MUL16]] +// CHECK-NEXT: [[MUL18:%.*]] = mul nsw i32 [[SUB17]], 1 +// CHECK-NEXT: [[ADD19:%.*]] = add nsw i32 0, [[MUL18]] +// CHECK-NEXT: store i32 [[ADD19]], ptr [[K]], align 4 +// CHECK-NEXT: br label %[[OMP_BODY_CONTINUE:.*]] +// CHECK: [[OMP_BODY_CONTINUE]]: +// CHECK-NEXT: br label %[[OMP_INNER_FOR_INC:.*]] +// CHECK: [[OMP_INNER_FOR_INC]]: +// CHECK-NEXT: [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4 +// CHECK-NEXT: [[ADD20:%.*]] = add nsw i32 [[TMP14]], 1 +// CHECK-NEXT: store i32 [[ADD20]], ptr [[DOTOMP_IV]], align 4 +// CHECK-NEXT: br label %[[OMP_INNER_FOR_COND]] +// CHECK: [[OMP_INNER_FOR_END]]: +// CHECK-NEXT: br label %[[OMP_LOOP_EXIT:.*]] +// CHECK: [[OMP_LOOP_EXIT]]: +// CHECK-NEXT: call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP1]]) +// CHECK-NEXT: ret void +void rectangular_loop() { +#pragma omp parallel for collapse(3) + for (int i = 0; i < 10; ++i) + for (int j = 0; j < 10; ++j) + for (int k = 0; k < 10; ++k) + ; +} _______________________________________________ cfe-commits mailing list [email protected] https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
