shylie/llvm-project
1
0
Fork 0
Code Issues Pull Requests Actions 6 Packages Projects Releases Wiki Activity

OMPIRBuilder for Interop directive

Browse Source 
Implements the OMPIRBuilder portion for the
Interop directive.

Reviewed By: jdoerfert

Differential Revision: https://reviews.llvm.org/D105876
This commit is contained in:
Sri Hari Krishna Narayanan 2022-01-27 14:53:00 -05:00 committed by Shilei Tian
parent 70e1cc6792
commit 5aa24558cf
8 changed files with 295 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
clang/lib/CodeGen/CGStmt.cpp
View File

@ -385,7 +385,7 @@ void CodeGenFunction::EmitStmt(const Stmt *S, ArrayRef<const Attr *> Attrs) {
cast<OMPTargetTeamsDistributeSimdDirective>(*S));
break;
case Stmt::OMPInteropDirectiveClass:
llvm_unreachable("Interop directive not supported yet.");
EmitOMPInteropDirective(cast<OMPInteropDirective>(*S));
break;
case Stmt::OMPDispatchDirectiveClass:
llvm_unreachable("Dispatch directive not supported yet.");

54
clang/lib/CodeGen/CGStmtOpenMP.cpp
View File

@ -6568,6 +6568,60 @@ void CodeGenFunction::EmitOMPTeamsDistributeParallelForSimdDirective(
[](CodeGenFunction &) { return nullptr; });
}
void CodeGenFunction::EmitOMPInteropDirective(const OMPInteropDirective &S) {
llvm::OpenMPIRBuilder &OMPBuilder = CGM.getOpenMPRuntime().getOMPBuilder();
llvm::Value *Device = nullptr;
if (const auto *C = S.getSingleClause<OMPDeviceClause>())
Device = EmitScalarExpr(C->getDevice());
llvm::Value *NumDependences = nullptr;
llvm::Value *DependenceAddress = nullptr;
if (const auto *DC = S.getSingleClause<OMPDependClause>()) {
OMPTaskDataTy::DependData Dependencies(DC->getDependencyKind(),
DC->getModifier());
Dependencies.DepExprs.append(DC->varlist_begin(), DC->varlist_end());
std::pair<llvm::Value *, Address> DependencePair =
CGM.getOpenMPRuntime().emitDependClause(*this, Dependencies,
DC->getBeginLoc());
NumDependences = DependencePair.first;
DependenceAddress = Builder.CreatePointerCast(
DependencePair.second.getPointer(), CGM.Int8PtrTy);
}
assert(!(S.hasClausesOfKind<OMPNowaitClause>() &&
!(S.getSingleClause<OMPInitClause>() ||
S.getSingleClause<OMPDestroyClause>() ||
S.getSingleClause<OMPUseClause>())) &&
"OMPNowaitClause clause is used separately in OMPInteropDirective.");
if (const auto *C = S.getSingleClause<OMPInitClause>()) {
llvm::Value *InteropvarPtr =
EmitLValue(C->getInteropVar()).getPointer(*this);
llvm::omp::OMPInteropType InteropType = llvm::omp::OMPInteropType::Unknown;
if (C->getIsTarget()) {
InteropType = llvm::omp::OMPInteropType::Target;
} else {
assert(C->getIsTargetSync() && "Expected interop-type target/targetsync");
InteropType = llvm::omp::OMPInteropType::TargetSync;
}
OMPBuilder.createOMPInteropInit(Builder, InteropvarPtr, InteropType, Device,
NumDependences, DependenceAddress,
S.hasClausesOfKind<OMPNowaitClause>());
} else if (const auto *C = S.getSingleClause<OMPDestroyClause>()) {
llvm::Value *InteropvarPtr =
EmitLValue(C->getInteropVar()).getPointer(*this);
OMPBuilder.createOMPInteropDestroy(Builder, InteropvarPtr, Device,
NumDependences, DependenceAddress,
S.hasClausesOfKind<OMPNowaitClause>());
} else if (const auto *C = S.getSingleClause<OMPUseClause>()) {
llvm::Value *InteropvarPtr =
EmitLValue(C->getInteropVar()).getPointer(*this);
OMPBuilder.createOMPInteropUse(Builder, InteropvarPtr, Device,
NumDependences, DependenceAddress,
S.hasClausesOfKind<OMPNowaitClause>());
}
}
static void emitTargetTeamsDistributeParallelForRegion(
CodeGenFunction &CGF, const OMPTargetTeamsDistributeParallelForDirective &S,
PrePostActionTy &Action) {

1
clang/lib/CodeGen/CodeGenFunction.h
View File

@ -3562,6 +3562,7 @@ public:
void EmitOMPTargetTeamsDistributeSimdDirective(
const OMPTargetTeamsDistributeSimdDirective &S);
void EmitOMPGenericLoopDirective(const OMPGenericLoopDirective &S);
void EmitOMPInteropDirective(const OMPInteropDirective &S);
/// Emit device code for the target directive.
static void EmitOMPTargetDeviceFunction(CodeGenModule &CGM,

96
clang/test/OpenMP/interop_irbuilder.cpp Normal file
View File

@ -0,0 +1,96 @@
// NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --include-generated-funcs
// RUN: %clang_cc1 -verify -triple x86_64-unknown-linux -fopenmp -x c++ -emit-llvm %s -o - | FileCheck %s
// expected-no-diagnostics
typedef void *omp_interop_t;
void test1() {
int device_id = 4;
int D0, D1;
omp_interop_t interop;
#pragma omp interop init(target \
: interop)
#pragma omp interop init(targetsync \
: interop)
#pragma omp interop init(target \
: interop) device(device_id)
#pragma omp interop init(targetsync \
: interop) device(device_id)
#pragma omp interop use(interop) depend(in \
: D0, D1) nowait
#pragma omp interop destroy(interop) depend(in \
: D0, D1)
}
// CHECK-LABEL: @_Z5test1v(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[DEVICE_ID:%.*]] = alloca i32, align 4
// CHECK-NEXT: [[D0:%.*]] = alloca i32, align 4
// CHECK-NEXT: [[D1:%.*]] = alloca i32, align 4
// CHECK-NEXT: [[INTEROP:%.*]] = alloca i8*, align 8
// CHECK-NEXT: [[DOTDEP_ARR_ADDR:%.*]] = alloca [2 x %struct.kmp_depend_info], align 8
// CHECK-NEXT: [[DEP_COUNTER_ADDR:%.*]] = alloca i64, align 8
// CHECK-NEXT: [[DOTDEP_ARR_ADDR5:%.*]] = alloca [2 x %struct.kmp_depend_info], align 8
// CHECK-NEXT: [[DEP_COUNTER_ADDR6:%.*]] = alloca i64, align 8
// CHECK-NEXT: store i32 4, i32* [[DEVICE_ID]], align 4
// CHECK-NEXT: [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1:[0-9]+]])
// CHECK-NEXT: call void @__tgt_interop_init(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM]], i8** [[INTEROP]], i64 1, i32 -1, i32 0, i8* null, i32 0)
// CHECK-NEXT: [[OMP_GLOBAL_THREAD_NUM1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// CHECK-NEXT: call void @__tgt_interop_init(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM1]], i8** [[INTEROP]], i64 2, i32 -1, i32 0, i8* null, i32 0)
// CHECK-NEXT: [[TMP0:%.*]] = load i32, i32* [[DEVICE_ID]], align 4
// CHECK-NEXT: [[OMP_GLOBAL_THREAD_NUM2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// CHECK-NEXT: call void @__tgt_interop_init(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM2]], i8** [[INTEROP]], i64 1, i32 [[TMP0]], i32 0, i8* null, i32 0)
// CHECK-NEXT: [[TMP1:%.*]] = load i32, i32* [[DEVICE_ID]], align 4
// CHECK-NEXT: [[OMP_GLOBAL_THREAD_NUM3:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// CHECK-NEXT: call void @__tgt_interop_init(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM3]], i8** [[INTEROP]], i64 2, i32 [[TMP1]], i32 0, i8* null, i32 0)
// CHECK-NEXT: [[TMP2:%.*]] = getelementptr inbounds [2 x %struct.kmp_depend_info], [2 x %struct.kmp_depend_info]* [[DOTDEP_ARR_ADDR]], i64 0, i64 0
// CHECK-NEXT: [[TMP3:%.*]] = getelementptr [[STRUCT_KMP_DEPEND_INFO:%.*]], %struct.kmp_depend_info* [[TMP2]], i64 0
// CHECK-NEXT: [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_KMP_DEPEND_INFO]], %struct.kmp_depend_info* [[TMP3]], i32 0, i32 0
// CHECK-NEXT: [[TMP5:%.*]] = ptrtoint i32* [[D0]] to i64
// CHECK-NEXT: store i64 [[TMP5]], i64* [[TMP4]], align 8
// CHECK-NEXT: [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_KMP_DEPEND_INFO]], %struct.kmp_depend_info* [[TMP3]], i32 0, i32 1
// CHECK-NEXT: store i64 4, i64* [[TMP6]], align 8
// CHECK-NEXT: [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_KMP_DEPEND_INFO]], %struct.kmp_depend_info* [[TMP3]], i32 0, i32 2
// CHECK-NEXT: store i8 1, i8* [[TMP7]], align 8
// CHECK-NEXT: [[TMP8:%.*]] = getelementptr [[STRUCT_KMP_DEPEND_INFO]], %struct.kmp_depend_info* [[TMP2]], i64 1
// CHECK-NEXT: [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_KMP_DEPEND_INFO]], %struct.kmp_depend_info* [[TMP8]], i32 0, i32 0
// CHECK-NEXT: [[TMP10:%.*]] = ptrtoint i32* [[D1]] to i64
// CHECK-NEXT: store i64 [[TMP10]], i64* [[TMP9]], align 8
// CHECK-NEXT: [[TMP11:%.*]] = getelementptr inbounds [[STRUCT_KMP_DEPEND_INFO]], %struct.kmp_depend_info* [[TMP8]], i32 0, i32 1
// CHECK-NEXT: store i64 4, i64* [[TMP11]], align 8
// CHECK-NEXT: [[TMP12:%.*]] = getelementptr inbounds [[STRUCT_KMP_DEPEND_INFO]], %struct.kmp_depend_info* [[TMP8]], i32 0, i32 2
// CHECK-NEXT: store i8 1, i8* [[TMP12]], align 8
// CHECK-NEXT: store i64 2, i64* [[DEP_COUNTER_ADDR]], align 8
// CHECK-NEXT: [[TMP13:%.*]] = bitcast %struct.kmp_depend_info* [[TMP2]] to i8*
// CHECK-NEXT: [[OMP_GLOBAL_THREAD_NUM4:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// CHECK-NEXT: call void @__tgt_interop_use(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM4]], i8** [[INTEROP]], i32 -1, i32 2, i8* [[TMP13]], i32 1)
// CHECK-NEXT: [[TMP14:%.*]] = getelementptr inbounds [2 x %struct.kmp_depend_info], [2 x %struct.kmp_depend_info]* [[DOTDEP_ARR_ADDR5]], i64 0, i64 0
// CHECK-NEXT: [[TMP15:%.*]] = getelementptr [[STRUCT_KMP_DEPEND_INFO]], %struct.kmp_depend_info* [[TMP14]], i64 0
// CHECK-NEXT: [[TMP16:%.*]] = getelementptr inbounds [[STRUCT_KMP_DEPEND_INFO]], %struct.kmp_depend_info* [[TMP15]], i32 0, i32 0
// CHECK-NEXT: [[TMP17:%.*]] = ptrtoint i32* [[D0]] to i64
// CHECK-NEXT: store i64 [[TMP17]], i64* [[TMP16]], align 8
// CHECK-NEXT: [[TMP18:%.*]] = getelementptr inbounds [[STRUCT_KMP_DEPEND_INFO]], %struct.kmp_depend_info* [[TMP15]], i32 0, i32 1
// CHECK-NEXT: store i64 4, i64* [[TMP18]], align 8
// CHECK-NEXT: [[TMP19:%.*]] = getelementptr inbounds [[STRUCT_KMP_DEPEND_INFO]], %struct.kmp_depend_info* [[TMP15]], i32 0, i32 2
// CHECK-NEXT: store i8 1, i8* [[TMP19]], align 8
// CHECK-NEXT: [[TMP20:%.*]] = getelementptr [[STRUCT_KMP_DEPEND_INFO]], %struct.kmp_depend_info* [[TMP14]], i64 1
// CHECK-NEXT: [[TMP21:%.*]] = getelementptr inbounds [[STRUCT_KMP_DEPEND_INFO]], %struct.kmp_depend_info* [[TMP20]], i32 0, i32 0
// CHECK-NEXT: [[TMP22:%.*]] = ptrtoint i32* [[D1]] to i64
// CHECK-NEXT: store i64 [[TMP22]], i64* [[TMP21]], align 8
// CHECK-NEXT: [[TMP23:%.*]] = getelementptr inbounds [[STRUCT_KMP_DEPEND_INFO]], %struct.kmp_depend_info* [[TMP20]], i32 0, i32 1
// CHECK-NEXT: store i64 4, i64* [[TMP23]], align 8
// CHECK-NEXT: [[TMP24:%.*]] = getelementptr inbounds [[STRUCT_KMP_DEPEND_INFO]], %struct.kmp_depend_info* [[TMP20]], i32 0, i32 2
// CHECK-NEXT: store i8 1, i8* [[TMP24]], align 8
// CHECK-NEXT: store i64 2, i64* [[DEP_COUNTER_ADDR6]], align 8
// CHECK-NEXT: [[TMP25:%.*]] = bitcast %struct.kmp_depend_info* [[TMP14]] to i8*
// CHECK-NEXT: [[OMP_GLOBAL_THREAD_NUM7:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// CHECK-NEXT: call void @__tgt_interop_destroy(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM7]], i8** [[INTEROP]], i32 -1, i32 2, i8* [[TMP25]], i32 0)
// CHECK-NEXT: ret void
//

3
llvm/include/llvm/Frontend/OpenMP/OMPConstants.h
View File

@ -113,6 +113,9 @@ enum class AddressSpace : unsigned {
Local = 5,
};
/// \note This needs to be kept in sync with interop.h enum kmp_interop_type_t.:
enum class OMPInteropType { Unknown, Target, TargetSync };
} // end namespace omp
} // end namespace llvm

49
llvm/include/llvm/Frontend/OpenMP/OMPIRBuilder.h
View File

@ -1003,6 +1003,55 @@ public:
llvm::ConstantInt *Size,
const llvm::Twine &Name = Twine(""));
/// Create a runtime call for __tgt_interop_init
///
/// \param Loc The insert and source location description.
/// \param InteropVar variable to be allocated
/// \param InteropType type of interop operation
/// \param Device devide to which offloading will occur
/// \param NumDependences number of dependence variables
/// \param DependenceAddress pointer to dependence variables
/// \param HaveNowaitClause does nowait clause exist
///
/// \returns CallInst to the __tgt_interop_init call
CallInst *createOMPInteropInit(const LocationDescription &Loc,
Value *InteropVar,
omp::OMPInteropType InteropType, Value *Device,
Value *NumDependences,
Value *DependenceAddress,
bool HaveNowaitClause);
/// Create a runtime call for __tgt_interop_destroy
///
/// \param Loc The insert and source location description.
/// \param InteropVar variable to be allocated
/// \param Device devide to which offloading will occur
/// \param NumDependences number of dependence variables
/// \param DependenceAddress pointer to dependence variables
/// \param HaveNowaitClause does nowait clause exist
///
/// \returns CallInst to the __tgt_interop_destroy call
CallInst *createOMPInteropDestroy(const LocationDescription &Loc,
Value *InteropVar, Value *Device,
Value *NumDependences,
Value *DependenceAddress,
bool HaveNowaitClause);
/// Create a runtime call for __tgt_interop_use
///
/// \param Loc The insert and source location description.
/// \param InteropVar variable to be allocated
/// \param Device devide to which offloading will occur
/// \param NumDependences number of dependence variables
/// \param DependenceAddress pointer to dependence variables
/// \param HaveNowaitClause does nowait clause exist
///
/// \returns CallInst to the __tgt_interop_use call
CallInst *createOMPInteropUse(const LocationDescription &Loc,
Value *InteropVar, Value *Device,
Value *NumDependences, Value *DependenceAddress,
bool HaveNowaitClause);
/// The `omp target` interface
///
/// For more information about the usage of this interface,

7
llvm/include/llvm/Frontend/OpenMP/OMPKinds.def
View File

@ -386,6 +386,13 @@ __OMP_RTL(__kmpc_aligned_alloc, false, VoidPtr, /* Int */ Int32, SizeTy, SizeTy,
VoidPtr)
__OMP_RTL(__kmpc_free, false, Void, /* Int */ Int32, VoidPtr, VoidPtr)
__OMP_RTL(__tgt_interop_init, false, Void, IdentPtr, Int32, VoidPtrPtr, Int64,
Int32, Int32, VoidPtr, Int32)
__OMP_RTL(__tgt_interop_destroy, false, Void, IdentPtr, Int32, VoidPtrPtr,
Int32, Int32, VoidPtr, Int32)
__OMP_RTL(__tgt_interop_use, false, Void, IdentPtr, Int32, VoidPtrPtr, Int32,
Int32, VoidPtr, Int32)
__OMP_RTL(__kmpc_init_allocator, false, /* omp_allocator_handle_t */ VoidPtr,
/* Int */ Int32, /* omp_memespace_handle_t */ VoidPtr,
/* Int */ Int32, /* omp_alloctrait_t */ VoidPtr)

84
llvm/lib/Frontend/OpenMP/OMPIRBuilder.cpp
View File

@ -2866,6 +2866,90 @@ CallInst *OpenMPIRBuilder::createOMPFree(const LocationDescription &Loc,
return Builder.CreateCall(Fn, Args, Name);
}
CallInst *OpenMPIRBuilder::createOMPInteropInit(
const LocationDescription &Loc, Value *InteropVar,
omp::OMPInteropType InteropType, Value *Device, Value *NumDependences,
Value *DependenceAddress, bool HaveNowaitClause) {
IRBuilder<>::InsertPointGuard IPG(Builder);
Builder.restoreIP(Loc.IP);
uint32_t SrcLocStrSize;
Constant *SrcLocStr = getOrCreateSrcLocStr(Loc, SrcLocStrSize);
Value *Ident = getOrCreateIdent(SrcLocStr, SrcLocStrSize);
Value *ThreadId = getOrCreateThreadID(Ident);
if (Device == NULL)
Device = ConstantInt::get(Int32, -1);
Constant *InteropTypeVal = ConstantInt::get(Int64, (int)InteropType);
if (NumDependences == nullptr) {
NumDependences = ConstantInt::get(Int32, 0);
PointerType *PointerTypeVar = Type::getInt8PtrTy(M.getContext());
DependenceAddress = ConstantPointerNull::get(PointerTypeVar);
}
Value *HaveNowaitClauseVal = ConstantInt::get(Int32, HaveNowaitClause);
Value *Args[] = {
Ident, ThreadId, InteropVar, InteropTypeVal,
Device, NumDependences, DependenceAddress, HaveNowaitClauseVal};
Function *Fn = getOrCreateRuntimeFunctionPtr(OMPRTL___tgt_interop_init);
return Builder.CreateCall(Fn, Args);
}
CallInst *OpenMPIRBuilder::createOMPInteropDestroy(
const LocationDescription &Loc, Value *InteropVar, Value *Device,
Value *NumDependences, Value *DependenceAddress, bool HaveNowaitClause) {
IRBuilder<>::InsertPointGuard IPG(Builder);
Builder.restoreIP(Loc.IP);
uint32_t SrcLocStrSize;
Constant *SrcLocStr = getOrCreateSrcLocStr(Loc, SrcLocStrSize);
Value *Ident = getOrCreateIdent(SrcLocStr, SrcLocStrSize);
Value *ThreadId = getOrCreateThreadID(Ident);
if (Device == NULL)
Device = ConstantInt::get(Int32, -1);
if (NumDependences == nullptr) {
NumDependences = ConstantInt::get(Int32, 0);
PointerType *PointerTypeVar = Type::getInt8PtrTy(M.getContext());
DependenceAddress = ConstantPointerNull::get(PointerTypeVar);
}
Value *HaveNowaitClauseVal = ConstantInt::get(Int32, HaveNowaitClause);
Value *Args[] = {
Ident, ThreadId, InteropVar, Device,
NumDependences, DependenceAddress, HaveNowaitClauseVal};
Function *Fn = getOrCreateRuntimeFunctionPtr(OMPRTL___tgt_interop_destroy);
return Builder.CreateCall(Fn, Args);
}
CallInst *OpenMPIRBuilder::createOMPInteropUse(const LocationDescription &Loc,
Value *InteropVar, Value *Device,
Value *NumDependences,
Value *DependenceAddress,
bool HaveNowaitClause) {
IRBuilder<>::InsertPointGuard IPG(Builder);
Builder.restoreIP(Loc.IP);
uint32_t SrcLocStrSize;
Constant *SrcLocStr = getOrCreateSrcLocStr(Loc, SrcLocStrSize);
Value *Ident = getOrCreateIdent(SrcLocStr, SrcLocStrSize);
Value *ThreadId = getOrCreateThreadID(Ident);
if (Device == NULL)
Device = ConstantInt::get(Int32, -1);
if (NumDependences == nullptr) {
NumDependences = ConstantInt::get(Int32, 0);
PointerType *PointerTypeVar = Type::getInt8PtrTy(M.getContext());
DependenceAddress = ConstantPointerNull::get(PointerTypeVar);
}
Value *HaveNowaitClauseVal = ConstantInt::get(Int32, HaveNowaitClause);
Value *Args[] = {
Ident, ThreadId, InteropVar, Device,
NumDependences, DependenceAddress, HaveNowaitClauseVal};
Function *Fn = getOrCreateRuntimeFunctionPtr(OMPRTL___tgt_interop_use);
return Builder.CreateCall(Fn, Args);
}
CallInst *OpenMPIRBuilder::createCachedThreadPrivate(
const LocationDescription &Loc, llvm::Value *Pointer,
llvm::ConstantInt *Size, const llvm::Twine &Name) {