[llvm] [OpenMP][OMPIRBuilder] Add support to omp target parallel (PR #67000)
Dominik Adamski via llvm-commits
llvm-commits at lists.llvm.org
Fri Sep 22 11:08:34 PDT 2023
https://github.com/DominikAdamski updated https://github.com/llvm/llvm-project/pull/67000
>From 13edd33c2b532a2f71a78392da52e6f5f57f51fc Mon Sep 17 00:00:00 2001
From: Dominik Adamski <dominik.adamski at gmail.com>
Date: Mon, 11 Sep 2023 05:31:37 -0400
Subject: [PATCH] [OpenMP][OMPIRBuilder] Add support to omp target parallel
Added support for LLVM IR code generation which is used for handling
omp target parallel code. The call for __kmpc_parallel_51 is generated
and the parallel region is outlined to separate function.
The proper setup of kmpc_target_init mode is not included in the commit.
It is assumed that the SPMD mode for target init is properly set by other
codegen functions.
---
llvm/lib/Frontend/OpenMP/OMPIRBuilder.cpp | 273 ++++++++++++------
llvm/lib/Transforms/IPO/OpenMPOpt.cpp | 1 +
.../Frontend/OpenMPIRBuilderTest.cpp | 121 +++++++-
3 files changed, 310 insertions(+), 85 deletions(-)
diff --git a/llvm/lib/Frontend/OpenMP/OMPIRBuilder.cpp b/llvm/lib/Frontend/OpenMP/OMPIRBuilder.cpp
index 1ace7d5b97ffc96..2fdf6994738d432 100644
--- a/llvm/lib/Frontend/OpenMP/OMPIRBuilder.cpp
+++ b/llvm/lib/Frontend/OpenMP/OMPIRBuilder.cpp
@@ -602,6 +602,8 @@ void OpenMPIRBuilder::finalize(Function *Fn) {
Function *OuterFn = OI.getFunction();
CodeExtractorAnalysisCache CEAC(*OuterFn);
+ unsigned AllocaAddressSpace =
+ Config.isTargetDevice() ? 0 : M.getDataLayout().getAllocaAddrSpace();
CodeExtractor Extractor(Blocks, /* DominatorTree */ nullptr,
/* AggregateArgs */ true,
/* BlockFrequencyInfo */ nullptr,
@@ -610,7 +612,7 @@ void OpenMPIRBuilder::finalize(Function *Fn) {
/* AllowVarArgs */ true,
/* AllowAlloca */ true,
/* AllocaBlock*/ OI.OuterAllocaBB,
- /* Suffix */ ".omp_par");
+ /* Suffix */ ".omp_par", AllocaAddressSpace);
LLVM_DEBUG(dbgs() << "Before outlining: " << *OuterFn << "\n");
LLVM_DEBUG(dbgs() << "Entry " << OI.EntryBB->getName()
@@ -1090,6 +1092,174 @@ void OpenMPIRBuilder::emitCancelationCheckImpl(Value *CancelFlag,
Builder.SetInsertPoint(NonCancellationBlock, NonCancellationBlock->begin());
}
+static void targetParallelCallback(OpenMPIRBuilder *OMPIRBuilder,
+ Function &OutlinedFn, Function *OuterFn,
+ Value *Ident, Value *IfCondition,
+ Value *NumThreads, Instruction *PrivTID,
+ AllocaInst *PrivTIDAddr, Value *ThreadID,
+ SmallVector<Instruction *, 4> ToBeDeleted) {
+ // Add some known attributes.
+ Module &M = OMPIRBuilder->M;
+ IRBuilder<> &Builder = OMPIRBuilder->Builder;
+ OutlinedFn.addParamAttr(0, Attribute::NoAlias);
+ OutlinedFn.addParamAttr(1, Attribute::NoAlias);
+ OutlinedFn.addParamAttr(0, Attribute::NoUndef);
+ OutlinedFn.addParamAttr(1, Attribute::NoUndef);
+ OutlinedFn.addFnAttr(Attribute::NoUnwind);
+
+ assert(OutlinedFn.arg_size() >= 2 &&
+ "Expected at least tid and bounded tid as arguments");
+ unsigned NumCapturedVars = OutlinedFn.arg_size() - /* tid & bounded tid */ 2;
+
+ CallInst *CI = cast<CallInst>(OutlinedFn.user_back());
+ assert(CI && "Expected call instruction to outlined function");
+ CI->getParent()->setName("omp_parallel");
+ // Replace direct call to the outlined function by the call to
+ // __kmpc_parallel_51
+ Builder.SetInsertPoint(CI);
+
+ // Build call __kmpc_parallel_51
+ auto PtrTy = Type::getInt8PtrTy(M.getContext());
+ Value *Void = ConstantPointerNull::get(PtrTy);
+ // Add alloca for kernel args. Put this instruction at the beginning
+ // of the function.
+ OpenMPIRBuilder ::InsertPointTy CurrentIP = Builder.saveIP();
+ Builder.SetInsertPoint(&OuterFn->front(),
+ OuterFn->front().getFirstInsertionPt());
+ AllocaInst *ArgsAlloca =
+ Builder.CreateAlloca(ArrayType::get(PtrTy, NumCapturedVars));
+ Value *Args =
+ Builder.CreatePointerCast(ArgsAlloca, Type::getInt8PtrTy(M.getContext()));
+ Builder.restoreIP(CurrentIP);
+ // Store captured vars which are used by kmpc_parallel_51
+ if (NumCapturedVars) {
+ for (unsigned Idx = 0; Idx < NumCapturedVars; Idx++) {
+ Value *V = *(CI->arg_begin() + 2 + Idx);
+ Value *StoreAddress = Builder.CreateConstInBoundsGEP2_64(
+ ArrayType::get(PtrTy, NumCapturedVars), Args, 0, Idx);
+ Builder.CreateStore(V, StoreAddress);
+ }
+ }
+ Value *Cond = IfCondition ? Builder.CreateSExtOrTrunc(
+ IfCondition, Type::getInt32Ty(M.getContext()))
+ : Builder.getInt32(1);
+ Value *Parallel51CallArgs[] = {
+ /* identifier*/ Ident,
+ /* global thread num*/ ThreadID,
+ /* if expression */ Cond, NumThreads ? NumThreads : Builder.getInt32(-1),
+ /* Proc bind */ Builder.getInt32(-1),
+ /* outlined function */
+ Builder.CreateBitCast(&OutlinedFn, OMPIRBuilder->ParallelTaskPtr), Void,
+ Args, Builder.getInt64(NumCapturedVars)};
+
+ SmallVector<Value *, 16> RealArgs;
+ RealArgs.append(std::begin(Parallel51CallArgs), std::end(Parallel51CallArgs));
+ FunctionCallee RTLFn =
+ OMPIRBuilder->getOrCreateRuntimeFunctionPtr(OMPRTL___kmpc_parallel_51);
+
+ Builder.CreateCall(RTLFn, RealArgs);
+
+ LLVM_DEBUG(dbgs() << "With kmpc_parallel_51 placed: "
+ << *Builder.GetInsertBlock()->getParent() << "\n");
+
+ // Initialize the local TID stack location with the argument value.
+ Builder.SetInsertPoint(PrivTID);
+ Function::arg_iterator OutlinedAI = OutlinedFn.arg_begin();
+ Builder.CreateStore(Builder.CreateLoad(OMPIRBuilder->Int32, OutlinedAI),
+ PrivTIDAddr);
+
+ ToBeDeleted.insert(ToBeDeleted.begin(), CI);
+
+ for (Instruction *I : ToBeDeleted) {
+ I->eraseFromParent();
+ }
+}
+static void hostParallelCallback(OpenMPIRBuilder *OMPIRBuilder,
+ Function &OutlinedFn, Function *OuterFn,
+ Value *Ident, Value *IfCondition,
+ Instruction *PrivTID, AllocaInst *PrivTIDAddr,
+ SmallVector<Instruction *, 4> ToBeDeleted) {
+ Module &M = OMPIRBuilder->M;
+ IRBuilder<> &Builder = OMPIRBuilder->Builder;
+ FunctionCallee RTLFn;
+ if (IfCondition) {
+ RTLFn =
+ OMPIRBuilder->getOrCreateRuntimeFunctionPtr(OMPRTL___kmpc_fork_call_if);
+ } else {
+ RTLFn =
+ OMPIRBuilder->getOrCreateRuntimeFunctionPtr(OMPRTL___kmpc_fork_call);
+ }
+ if (auto *F = dyn_cast<llvm::Function>(RTLFn.getCallee())) {
+ if (!F->hasMetadata(llvm::LLVMContext::MD_callback)) {
+ llvm::LLVMContext &Ctx = F->getContext();
+ MDBuilder MDB(Ctx);
+ // Annotate the callback behavior of the __kmpc_fork_call:
+ // - The callback callee is argument number 2 (microtask).
+ // - The first two arguments of the callback callee are unknown (-1).
+ // - All variadic arguments to the __kmpc_fork_call are passed to the
+ // callback callee.
+ F->addMetadata(
+ llvm::LLVMContext::MD_callback,
+ *llvm::MDNode::get(
+ Ctx, {MDB.createCallbackEncoding(2, {-1, -1},
+ /* VarArgsArePassed */ true)}));
+ }
+ }
+ // Add some known attributes.
+ OutlinedFn.addParamAttr(0, Attribute::NoAlias);
+ OutlinedFn.addParamAttr(1, Attribute::NoAlias);
+ OutlinedFn.addFnAttr(Attribute::NoUnwind);
+
+ assert(OutlinedFn.arg_size() >= 2 &&
+ "Expected at least tid and bounded tid as arguments");
+ unsigned NumCapturedVars = OutlinedFn.arg_size() - /* tid & bounded tid */ 2;
+
+ CallInst *CI = cast<CallInst>(OutlinedFn.user_back());
+ CI->getParent()->setName("omp_parallel");
+ Builder.SetInsertPoint(CI);
+
+ // Build call __kmpc_fork_call[_if](Ident, n, microtask, var1, .., varn);
+ Value *ForkCallArgs[] = {
+ Ident, Builder.getInt32(NumCapturedVars),
+ Builder.CreateBitCast(&OutlinedFn, OMPIRBuilder->ParallelTaskPtr)};
+
+ SmallVector<Value *, 16> RealArgs;
+ RealArgs.append(std::begin(ForkCallArgs), std::end(ForkCallArgs));
+ if (IfCondition) {
+ Value *Cond = Builder.CreateSExtOrTrunc(IfCondition,
+ Type::getInt32Ty(M.getContext()));
+ RealArgs.push_back(Cond);
+ }
+ RealArgs.append(CI->arg_begin() + /* tid & bound tid */ 2, CI->arg_end());
+
+ // __kmpc_fork_call_if always expects a void ptr as the last argument
+ // If there are no arguments, pass a null pointer.
+ auto PtrTy = Type::getInt8PtrTy(M.getContext());
+ if (IfCondition && NumCapturedVars == 0) {
+ Value *Void = ConstantPointerNull::get(PtrTy);
+ RealArgs.push_back(Void);
+ }
+ if (IfCondition && RealArgs.back()->getType() != PtrTy)
+ RealArgs.back() = Builder.CreateBitCast(RealArgs.back(), PtrTy);
+
+ Builder.CreateCall(RTLFn, RealArgs);
+
+ LLVM_DEBUG(dbgs() << "With fork_call placed: "
+ << *Builder.GetInsertBlock()->getParent() << "\n");
+
+ // Initialize the local TID stack location with the argument value.
+ Builder.SetInsertPoint(PrivTID);
+ Function::arg_iterator OutlinedAI = OutlinedFn.arg_begin();
+ Builder.CreateStore(Builder.CreateLoad(OMPIRBuilder->Int32, OutlinedAI),
+ PrivTIDAddr);
+
+ ToBeDeleted.insert(ToBeDeleted.begin(), CI);
+
+ for (Instruction *I : ToBeDeleted) {
+ I->eraseFromParent();
+ }
+}
+
IRBuilder<>::InsertPoint OpenMPIRBuilder::createParallel(
const LocationDescription &Loc, InsertPointTy OuterAllocaIP,
BodyGenCallbackTy BodyGenCB, PrivatizeCallbackTy PrivCB,
@@ -1137,8 +1307,12 @@ IRBuilder<>::InsertPoint OpenMPIRBuilder::createParallel(
// Change the location to the outer alloca insertion point to create and
// initialize the allocas we pass into the parallel region.
Builder.restoreIP(OuterAllocaIP);
- AllocaInst *TIDAddr = Builder.CreateAlloca(Int32, nullptr, "tid.addr");
- AllocaInst *ZeroAddr = Builder.CreateAlloca(Int32, nullptr, "zero.addr");
+ unsigned AllocaAddressSpace =
+ Config.isTargetDevice() ? 0 : M.getDataLayout().getAllocaAddrSpace();
+ AllocaInst *TIDAddr =
+ Builder.CreateAlloca(Int32, AllocaAddressSpace, nullptr, "tid.addr");
+ AllocaInst *ZeroAddr =
+ Builder.CreateAlloca(Int32, AllocaAddressSpace, nullptr, "zero.addr");
// We only need TIDAddr and ZeroAddr for modeling purposes to get the
// associated arguments in the outlined function, so we delete them later.
@@ -1211,87 +1385,22 @@ IRBuilder<>::InsertPoint OpenMPIRBuilder::createParallel(
BodyGenCB(InnerAllocaIP, CodeGenIP);
LLVM_DEBUG(dbgs() << "After body codegen: " << *OuterFn << "\n");
- FunctionCallee RTLFn;
- if (IfCondition)
- RTLFn = getOrCreateRuntimeFunctionPtr(OMPRTL___kmpc_fork_call_if);
- else
- RTLFn = getOrCreateRuntimeFunctionPtr(OMPRTL___kmpc_fork_call);
-
- if (auto *F = dyn_cast<llvm::Function>(RTLFn.getCallee())) {
- if (!F->hasMetadata(llvm::LLVMContext::MD_callback)) {
- llvm::LLVMContext &Ctx = F->getContext();
- MDBuilder MDB(Ctx);
- // Annotate the callback behavior of the __kmpc_fork_call:
- // - The callback callee is argument number 2 (microtask).
- // - The first two arguments of the callback callee are unknown (-1).
- // - All variadic arguments to the __kmpc_fork_call are passed to the
- // callback callee.
- F->addMetadata(
- llvm::LLVMContext::MD_callback,
- *llvm::MDNode::get(
- Ctx, {MDB.createCallbackEncoding(2, {-1, -1},
- /* VarArgsArePassed */ true)}));
- }
- }
OutlineInfo OI;
- OI.PostOutlineCB = [=](Function &OutlinedFn) {
- // Add some known attributes.
- OutlinedFn.addParamAttr(0, Attribute::NoAlias);
- OutlinedFn.addParamAttr(1, Attribute::NoAlias);
- OutlinedFn.addFnAttr(Attribute::NoUnwind);
- OutlinedFn.addFnAttr(Attribute::NoRecurse);
-
- assert(OutlinedFn.arg_size() >= 2 &&
- "Expected at least tid and bounded tid as arguments");
- unsigned NumCapturedVars =
- OutlinedFn.arg_size() - /* tid & bounded tid */ 2;
-
- CallInst *CI = cast<CallInst>(OutlinedFn.user_back());
- CI->getParent()->setName("omp_parallel");
- Builder.SetInsertPoint(CI);
-
- // Build call __kmpc_fork_call[_if](Ident, n, microtask, var1, .., varn);
- Value *ForkCallArgs[] = {
- Ident, Builder.getInt32(NumCapturedVars),
- Builder.CreateBitCast(&OutlinedFn, ParallelTaskPtr)};
-
- SmallVector<Value *, 16> RealArgs;
- RealArgs.append(std::begin(ForkCallArgs), std::end(ForkCallArgs));
- if (IfCondition) {
- Value *Cond = Builder.CreateSExtOrTrunc(IfCondition,
- Type::getInt32Ty(M.getContext()));
- RealArgs.push_back(Cond);
- }
- RealArgs.append(CI->arg_begin() + /* tid & bound tid */ 2, CI->arg_end());
-
- // __kmpc_fork_call_if always expects a void ptr as the last argument
- // If there are no arguments, pass a null pointer.
- auto PtrTy = Type::getInt8PtrTy(M.getContext());
- if (IfCondition && NumCapturedVars == 0) {
- llvm::Value *Void = ConstantPointerNull::get(PtrTy);
- RealArgs.push_back(Void);
- }
- if (IfCondition && RealArgs.back()->getType() != PtrTy)
- RealArgs.back() = Builder.CreateBitCast(RealArgs.back(), PtrTy);
-
- Builder.CreateCall(RTLFn, RealArgs);
-
- LLVM_DEBUG(dbgs() << "With fork_call placed: "
- << *Builder.GetInsertBlock()->getParent() << "\n");
-
- InsertPointTy ExitIP(PRegExitBB, PRegExitBB->end());
-
- // Initialize the local TID stack location with the argument value.
- Builder.SetInsertPoint(PrivTID);
- Function::arg_iterator OutlinedAI = OutlinedFn.arg_begin();
- Builder.CreateStore(Builder.CreateLoad(Int32, OutlinedAI), PrivTIDAddr);
-
- CI->eraseFromParent();
-
- for (Instruction *I : ToBeDeleted)
- I->eraseFromParent();
- };
+ if (Config.isTargetDevice()) {
+ // Generate OpenMP target specific runtime call
+ OI.PostOutlineCB = [=](Function &OutlinedFn) {
+ targetParallelCallback(this, OutlinedFn, OuterFn, Ident, IfCondition,
+ NumThreads, PrivTID, PrivTIDAddr, ThreadID,
+ ToBeDeleted);
+ };
+ } else {
+ // Generate OpenMP host runtime call
+ OI.PostOutlineCB = [=](Function &OutlinedFn) {
+ hostParallelCallback(this, OutlinedFn, OuterFn, Ident, IfCondition,
+ PrivTID, PrivTIDAddr, ToBeDeleted);
+ };
+ }
// Adjust the finalization stack, verify the adjustment, and call the
// finalize function a last time to finalize values between the pre-fini
diff --git a/llvm/lib/Transforms/IPO/OpenMPOpt.cpp b/llvm/lib/Transforms/IPO/OpenMPOpt.cpp
index f945de52920ccfe..a92343a4028226e 100644
--- a/llvm/lib/Transforms/IPO/OpenMPOpt.cpp
+++ b/llvm/lib/Transforms/IPO/OpenMPOpt.cpp
@@ -520,6 +520,7 @@ struct OMPInformationCache : public InformationCache {
void recollectUses() {
for (int Idx = 0; Idx < RFIs.size(); ++Idx)
recollectUsesForFunction(static_cast<RuntimeFunction>(Idx));
+ OMPBuilder.Config.IsTargetDevice = isOpenMPDevice(OMPBuilder.M);
}
// Helper function to inherit the calling convention of the function callee.
diff --git a/llvm/unittests/Frontend/OpenMPIRBuilderTest.cpp b/llvm/unittests/Frontend/OpenMPIRBuilderTest.cpp
index 2026824416f3e3c..f6ca787cc2604cc 100644
--- a/llvm/unittests/Frontend/OpenMPIRBuilderTest.cpp
+++ b/llvm/unittests/Frontend/OpenMPIRBuilderTest.cpp
@@ -591,9 +591,113 @@ TEST_F(OpenMPIRBuilderTest, DbgLoc) {
EXPECT_EQ(SrcSrc->getAsCString(), ";/src/test.dbg;foo;3;7;;");
}
+TEST_F(OpenMPIRBuilderTest, ParallelSimpleGPU) {
+ using InsertPointTy = OpenMPIRBuilder::InsertPointTy;
+ OpenMPIRBuilder OMPBuilder(*M);
+ OMPBuilder.Config.IsTargetDevice = true;
+ OMPBuilder.initialize();
+ F->setName("func");
+ IRBuilder<> Builder(BB);
+ BasicBlock *EnterBB = BasicBlock::Create(Ctx, "parallel.enter", F);
+ Builder.CreateBr(EnterBB);
+ Builder.SetInsertPoint(EnterBB);
+ OpenMPIRBuilder::LocationDescription Loc({Builder.saveIP(), DL});
+ Loc = OMPBuilder.createTargetInit(Loc, true);
+ OMPBuilder.M.print(errs(), nullptr);
+
+ AllocaInst *PrivAI = nullptr;
+
+ unsigned NumBodiesGenerated = 0;
+ unsigned NumPrivatizedVars = 0;
+ unsigned NumFinalizationPoints = 0;
+
+ auto BodyGenCB = [&](InsertPointTy AllocaIP, InsertPointTy CodeGenIP) {
+ ++NumBodiesGenerated;
+
+ Builder.restoreIP(AllocaIP);
+ PrivAI = Builder.CreateAlloca(F->arg_begin()->getType());
+ Builder.CreateStore(F->arg_begin(), PrivAI);
+
+ Builder.restoreIP(CodeGenIP);
+ Value *PrivLoad =
+ Builder.CreateLoad(PrivAI->getAllocatedType(), PrivAI, "local.use");
+ Value *Cmp = Builder.CreateICmpNE(F->arg_begin(), PrivLoad);
+ Instruction *ThenTerm, *ElseTerm;
+ SplitBlockAndInsertIfThenElse(Cmp, CodeGenIP.getBlock()->getTerminator(),
+ &ThenTerm, &ElseTerm);
+ };
+
+ auto PrivCB = [&](InsertPointTy AllocaIP, InsertPointTy CodeGenIP,
+ Value &Orig, Value &Inner,
+ Value *&ReplacementValue) -> InsertPointTy {
+ ++NumPrivatizedVars;
+
+ if (!isa<AllocaInst>(Orig)) {
+ EXPECT_EQ(&Orig, F->arg_begin());
+ ReplacementValue = &Inner;
+ return CodeGenIP;
+ }
+
+ // Since the original value is an allocation, it has a pointer type and
+ // therefore no additional wrapping should happen.
+ EXPECT_EQ(&Orig, &Inner);
+
+ // Trivial copy (=firstprivate).
+ Builder.restoreIP(AllocaIP);
+ Type *VTy = ReplacementValue->getType();
+ Value *V = Builder.CreateLoad(VTy, &Inner, Orig.getName() + ".reload");
+ ReplacementValue = Builder.CreateAlloca(VTy, 0, Orig.getName() + ".copy");
+ Builder.restoreIP(CodeGenIP);
+ Builder.CreateStore(V, ReplacementValue);
+ return CodeGenIP;
+ };
+
+ auto FiniCB = [&](InsertPointTy CodeGenIP) { ++NumFinalizationPoints; };
+
+ IRBuilder<>::InsertPoint AllocaIP(&F->getEntryBlock(),
+ F->getEntryBlock().getFirstInsertionPt());
+ IRBuilder<>::InsertPoint AfterIP =
+ OMPBuilder.createParallel(Loc, AllocaIP, BodyGenCB, PrivCB, FiniCB,
+ nullptr, nullptr, OMP_PROC_BIND_default, false);
+
+ EXPECT_EQ(NumBodiesGenerated, 1U);
+ EXPECT_EQ(NumPrivatizedVars, 1U);
+ EXPECT_EQ(NumFinalizationPoints, 1U);
+
+ Builder.restoreIP(AfterIP);
+ OMPBuilder.createTargetDeinit(Builder);
+ Builder.CreateRetVoid();
+
+ OMPBuilder.finalize();
+ OMPBuilder.M.print(llvm::errs(), nullptr);
+ Function *OutlinedFn = PrivAI->getFunction();
+ EXPECT_FALSE(verifyModule(*M, &errs()));
+ EXPECT_NE(OutlinedFn, F);
+ EXPECT_TRUE(OutlinedFn->hasFnAttribute(Attribute::NoUnwind));
+ EXPECT_TRUE(OutlinedFn->hasParamAttribute(0, Attribute::NoAlias));
+ EXPECT_TRUE(OutlinedFn->hasParamAttribute(1, Attribute::NoAlias));
+
+ EXPECT_TRUE(OutlinedFn->hasInternalLinkage());
+ EXPECT_EQ(OutlinedFn->arg_size(), 3U);
+
+ EXPECT_EQ(&OutlinedFn->getEntryBlock(), PrivAI->getParent());
+ EXPECT_EQ(OutlinedFn->getNumUses(), 1U);
+ User *Usr = OutlinedFn->user_back();
+ ASSERT_TRUE(isa<CallInst>(Usr));
+ CallInst *Parallel51CI = dyn_cast<CallInst>(Usr);
+ ASSERT_NE(Parallel51CI, nullptr);
+
+ EXPECT_EQ(Parallel51CI->getCalledFunction()->getName(), "__kmpc_parallel_51");
+ EXPECT_EQ(Parallel51CI->arg_size(), 9U);
+ EXPECT_EQ(Parallel51CI->getArgOperand(5), OutlinedFn);
+ EXPECT_TRUE(isa<GlobalVariable>(Parallel51CI->getArgOperand(0)));
+ EXPECT_EQ(Parallel51CI, Usr);
+}
+
TEST_F(OpenMPIRBuilderTest, ParallelSimple) {
using InsertPointTy = OpenMPIRBuilder::InsertPointTy;
OpenMPIRBuilder OMPBuilder(*M);
+ OMPBuilder.Config.IsTargetDevice = false;
OMPBuilder.initialize();
F->setName("func");
IRBuilder<> Builder(BB);
@@ -671,7 +775,6 @@ TEST_F(OpenMPIRBuilderTest, ParallelSimple) {
EXPECT_NE(F, OutlinedFn);
EXPECT_FALSE(verifyModule(*M, &errs()));
EXPECT_TRUE(OutlinedFn->hasFnAttribute(Attribute::NoUnwind));
- EXPECT_TRUE(OutlinedFn->hasFnAttribute(Attribute::NoRecurse));
EXPECT_TRUE(OutlinedFn->hasParamAttribute(0, Attribute::NoAlias));
EXPECT_TRUE(OutlinedFn->hasParamAttribute(1, Attribute::NoAlias));
@@ -699,6 +802,7 @@ TEST_F(OpenMPIRBuilderTest, ParallelSimple) {
TEST_F(OpenMPIRBuilderTest, ParallelNested) {
using InsertPointTy = OpenMPIRBuilder::InsertPointTy;
OpenMPIRBuilder OMPBuilder(*M);
+ OMPBuilder.Config.IsTargetDevice = false;
OMPBuilder.initialize();
F->setName("func");
IRBuilder<> Builder(BB);
@@ -768,7 +872,6 @@ TEST_F(OpenMPIRBuilderTest, ParallelNested) {
continue;
EXPECT_FALSE(verifyModule(*M, &errs()));
EXPECT_TRUE(OutlinedFn.hasFnAttribute(Attribute::NoUnwind));
- EXPECT_TRUE(OutlinedFn.hasFnAttribute(Attribute::NoRecurse));
EXPECT_TRUE(OutlinedFn.hasParamAttribute(0, Attribute::NoAlias));
EXPECT_TRUE(OutlinedFn.hasParamAttribute(1, Attribute::NoAlias));
@@ -793,6 +896,7 @@ TEST_F(OpenMPIRBuilderTest, ParallelNested) {
TEST_F(OpenMPIRBuilderTest, ParallelNested2Inner) {
using InsertPointTy = OpenMPIRBuilder::InsertPointTy;
OpenMPIRBuilder OMPBuilder(*M);
+ OMPBuilder.Config.IsTargetDevice = false;
OMPBuilder.initialize();
F->setName("func");
IRBuilder<> Builder(BB);
@@ -872,7 +976,6 @@ TEST_F(OpenMPIRBuilderTest, ParallelNested2Inner) {
continue;
EXPECT_FALSE(verifyModule(*M, &errs()));
EXPECT_TRUE(OutlinedFn.hasFnAttribute(Attribute::NoUnwind));
- EXPECT_TRUE(OutlinedFn.hasFnAttribute(Attribute::NoRecurse));
EXPECT_TRUE(OutlinedFn.hasParamAttribute(0, Attribute::NoAlias));
EXPECT_TRUE(OutlinedFn.hasParamAttribute(1, Attribute::NoAlias));
@@ -902,6 +1005,7 @@ TEST_F(OpenMPIRBuilderTest, ParallelNested2Inner) {
TEST_F(OpenMPIRBuilderTest, ParallelIfCond) {
using InsertPointTy = OpenMPIRBuilder::InsertPointTy;
OpenMPIRBuilder OMPBuilder(*M);
+ OMPBuilder.Config.IsTargetDevice = false;
OMPBuilder.initialize();
F->setName("func");
IRBuilder<> Builder(BB);
@@ -1006,6 +1110,7 @@ TEST_F(OpenMPIRBuilderTest, ParallelIfCond) {
TEST_F(OpenMPIRBuilderTest, ParallelCancelBarrier) {
using InsertPointTy = OpenMPIRBuilder::InsertPointTy;
OpenMPIRBuilder OMPBuilder(*M);
+ OMPBuilder.Config.IsTargetDevice = false;
OMPBuilder.initialize();
F->setName("func");
IRBuilder<> Builder(BB);
@@ -1119,6 +1224,7 @@ TEST_F(OpenMPIRBuilderTest, ParallelCancelBarrier) {
TEST_F(OpenMPIRBuilderTest, ParallelForwardAsPointers) {
OpenMPIRBuilder OMPBuilder(*M);
+ OMPBuilder.Config.IsTargetDevice = false;
OMPBuilder.initialize();
F->setName("func");
IRBuilder<> Builder(BB);
@@ -4131,6 +4237,7 @@ xorAtomicReduction(OpenMPIRBuilder::InsertPointTy IP, Type *Ty, Value *LHS,
TEST_F(OpenMPIRBuilderTest, CreateReductions) {
using InsertPointTy = OpenMPIRBuilder::InsertPointTy;
OpenMPIRBuilder OMPBuilder(*M);
+ OMPBuilder.Config.IsTargetDevice = false;
OMPBuilder.initialize();
F->setName("func");
IRBuilder<> Builder(BB);
@@ -4363,6 +4470,7 @@ TEST_F(OpenMPIRBuilderTest, CreateReductions) {
TEST_F(OpenMPIRBuilderTest, CreateTwoReductions) {
using InsertPointTy = OpenMPIRBuilder::InsertPointTy;
OpenMPIRBuilder OMPBuilder(*M);
+ OMPBuilder.Config.IsTargetDevice = false;
OMPBuilder.initialize();
F->setName("func");
IRBuilder<> Builder(BB);
@@ -5324,6 +5432,7 @@ TEST_F(OpenMPIRBuilderTest, TargetRegionDevice) {
TEST_F(OpenMPIRBuilderTest, CreateTask) {
using InsertPointTy = OpenMPIRBuilder::InsertPointTy;
OpenMPIRBuilder OMPBuilder(*M);
+ OMPBuilder.Config.IsTargetDevice = false;
OMPBuilder.initialize();
F->setName("func");
IRBuilder<> Builder(BB);
@@ -5449,6 +5558,7 @@ TEST_F(OpenMPIRBuilderTest, CreateTask) {
TEST_F(OpenMPIRBuilderTest, CreateTaskNoArgs) {
using InsertPointTy = OpenMPIRBuilder::InsertPointTy;
OpenMPIRBuilder OMPBuilder(*M);
+ OMPBuilder.Config.IsTargetDevice = false;
OMPBuilder.initialize();
F->setName("func");
IRBuilder<> Builder(BB);
@@ -5471,6 +5581,7 @@ TEST_F(OpenMPIRBuilderTest, CreateTaskNoArgs) {
TEST_F(OpenMPIRBuilderTest, CreateTaskUntied) {
using InsertPointTy = OpenMPIRBuilder::InsertPointTy;
OpenMPIRBuilder OMPBuilder(*M);
+ OMPBuilder.Config.IsTargetDevice = false;
OMPBuilder.initialize();
F->setName("func");
IRBuilder<> Builder(BB);
@@ -5500,6 +5611,7 @@ TEST_F(OpenMPIRBuilderTest, CreateTaskUntied) {
TEST_F(OpenMPIRBuilderTest, CreateTaskDepend) {
using InsertPointTy = OpenMPIRBuilder::InsertPointTy;
OpenMPIRBuilder OMPBuilder(*M);
+ OMPBuilder.Config.IsTargetDevice = false;
OMPBuilder.initialize();
F->setName("func");
IRBuilder<> Builder(BB);
@@ -5573,6 +5685,7 @@ TEST_F(OpenMPIRBuilderTest, CreateTaskDepend) {
TEST_F(OpenMPIRBuilderTest, CreateTaskFinal) {
using InsertPointTy = OpenMPIRBuilder::InsertPointTy;
OpenMPIRBuilder OMPBuilder(*M);
+ OMPBuilder.Config.IsTargetDevice = false;
OMPBuilder.initialize();
F->setName("func");
IRBuilder<> Builder(BB);
@@ -5626,6 +5739,7 @@ TEST_F(OpenMPIRBuilderTest, CreateTaskFinal) {
TEST_F(OpenMPIRBuilderTest, CreateTaskIfCondition) {
using InsertPointTy = OpenMPIRBuilder::InsertPointTy;
OpenMPIRBuilder OMPBuilder(*M);
+ OMPBuilder.Config.IsTargetDevice = false;
OMPBuilder.initialize();
F->setName("func");
IRBuilder<> Builder(BB);
@@ -5785,6 +5899,7 @@ TEST_F(OpenMPIRBuilderTest, CreateTaskgroup) {
TEST_F(OpenMPIRBuilderTest, CreateTaskgroupWithTasks) {
using InsertPointTy = OpenMPIRBuilder::InsertPointTy;
OpenMPIRBuilder OMPBuilder(*M);
+ OMPBuilder.Config.IsTargetDevice = false;
OMPBuilder.initialize();
F->setName("func");
IRBuilder<> Builder(BB);
More information about the llvm-commits
mailing list