r310511 - [OPENMP] Emit non-debug version of outlined functions with original
Alexey Bataev via cfe-commits
cfe-commits at lists.llvm.org
Wed Aug 9 12:38:53 PDT 2017
Author: abataev
Date: Wed Aug 9 12:38:53 2017
New Revision: 310511
URL: http://llvm.org/viewvc/llvm-project?rev=310511&view=rev
Log:
[OPENMP] Emit non-debug version of outlined functions with original
name.
If the host code is compiled with the debug info, while the target
without, there is a problem that the compiler is unable to find the
debug wrapper. Patch fixes this problem by emitting special name for the
debug version of the code.
Modified:
cfe/trunk/lib/CodeGen/CGStmtOpenMP.cpp
cfe/trunk/test/OpenMP/distribute_codegen.cpp
cfe/trunk/test/OpenMP/parallel_codegen.cpp
cfe/trunk/test/OpenMP/parallel_for_simd_codegen.cpp
cfe/trunk/test/OpenMP/target_parallel_debug_codegen.cpp
Modified: cfe/trunk/lib/CodeGen/CGStmtOpenMP.cpp
URL: http://llvm.org/viewvc/llvm-project/cfe/trunk/lib/CodeGen/CGStmtOpenMP.cpp?rev=310511&r1=310510&r2=310511&view=diff
==============================================================================
--- cfe/trunk/lib/CodeGen/CGStmtOpenMP.cpp (original)
+++ cfe/trunk/lib/CodeGen/CGStmtOpenMP.cpp Wed Aug 9 12:38:53 2017
@@ -247,7 +247,7 @@ namespace {
/// true if cast to/from UIntPtr is required for variables captured by
/// value.
const bool UIntPtrCastRequired = true;
- /// true if only casted argumefnts must be registered as local args or VLA
+ /// true if only casted arguments must be registered as local args or VLA
/// sizes.
const bool RegisterCastedArgsOnly = false;
/// Name of the generated function.
@@ -261,7 +261,7 @@ namespace {
};
}
-static std::pair<llvm::Function *, bool> emitOutlinedFunctionPrologue(
+static llvm::Function *emitOutlinedFunctionPrologue(
CodeGenFunction &CGF, FunctionArgList &Args,
llvm::MapVector<const Decl *, std::pair<const VarDecl *, Address>>
&LocalAddrs,
@@ -277,7 +277,6 @@ static std::pair<llvm::Function *, bool>
CodeGenModule &CGM = CGF.CGM;
ASTContext &Ctx = CGM.getContext();
FunctionArgList TargetArgs;
- bool HasUIntPtrArgs = false;
Args.append(CD->param_begin(),
std::next(CD->param_begin(), CD->getContextParamPosition()));
TargetArgs.append(
@@ -296,7 +295,6 @@ static std::pair<llvm::Function *, bool>
// outlined function.
if ((I->capturesVariableByCopy() && !ArgType->isAnyPointerType()) ||
I->capturesVariableArrayType()) {
- HasUIntPtrArgs = true;
if (FO.UIntPtrCastRequired)
ArgType = Ctx.getUIntPtrType();
}
@@ -432,7 +430,7 @@ static std::pair<llvm::Function *, bool>
++I;
}
- return {F, HasUIntPtrArgs};
+ return F;
}
llvm::Function *
@@ -448,12 +446,15 @@ CodeGenFunction::GenerateOpenMPCapturedS
FunctionArgList Args;
llvm::MapVector<const Decl *, std::pair<const VarDecl *, Address>> LocalAddrs;
llvm::DenseMap<const Decl *, std::pair<const Expr *, llvm::Value *>> VLASizes;
+ SmallString<256> Buffer;
+ llvm::raw_svector_ostream Out(Buffer);
+ Out << CapturedStmtInfo->getHelperName();
+ if (NeedWrapperFunction)
+ Out << "_debug__";
FunctionOptions FO(&S, !NeedWrapperFunction, /*RegisterCastedArgsOnly=*/false,
- CapturedStmtInfo->getHelperName());
- llvm::Function *F;
- bool HasUIntPtrArgs;
- std::tie(F, HasUIntPtrArgs) = emitOutlinedFunctionPrologue(
- *this, Args, LocalAddrs, VLASizes, CXXThisValue, FO);
+ Out.str());
+ llvm::Function *F = emitOutlinedFunctionPrologue(*this, Args, LocalAddrs,
+ VLASizes, CXXThisValue, FO);
for (const auto &LocalAddrPair : LocalAddrs) {
if (LocalAddrPair.second.first) {
setAddrOfLocalVar(LocalAddrPair.second.first,
@@ -465,14 +466,12 @@ CodeGenFunction::GenerateOpenMPCapturedS
PGO.assignRegionCounters(GlobalDecl(CD), F);
CapturedStmtInfo->EmitBody(*this, CD->getBody());
FinishFunction(CD->getBodyRBrace());
- if (!NeedWrapperFunction || !HasUIntPtrArgs)
+ if (!NeedWrapperFunction)
return F;
- SmallString<256> Buffer;
- llvm::raw_svector_ostream Out(Buffer);
- Out << "__nondebug_wrapper_" << CapturedStmtInfo->getHelperName();
FunctionOptions WrapperFO(&S, /*UIntPtrCastRequired=*/true,
- /*RegisterCastedArgsOnly=*/true, Out.str());
+ /*RegisterCastedArgsOnly=*/true,
+ CapturedStmtInfo->getHelperName());
CodeGenFunction WrapperCGF(CGM, /*suppressNewContext=*/true);
WrapperCGF.disableDebugInfo();
Args.clear();
@@ -480,7 +479,7 @@ CodeGenFunction::GenerateOpenMPCapturedS
VLASizes.clear();
llvm::Function *WrapperF =
emitOutlinedFunctionPrologue(WrapperCGF, Args, LocalAddrs, VLASizes,
- WrapperCGF.CXXThisValue, WrapperFO).first;
+ WrapperCGF.CXXThisValue, WrapperFO);
LValueBaseInfo BaseInfo(AlignmentSource::Decl, false);
llvm::SmallVector<llvm::Value *, 4> CallArgs;
for (const auto *Arg : Args) {
Modified: cfe/trunk/test/OpenMP/distribute_codegen.cpp
URL: http://llvm.org/viewvc/llvm-project/cfe/trunk/test/OpenMP/distribute_codegen.cpp?rev=310511&r1=310510&r2=310511&view=diff
==============================================================================
--- cfe/trunk/test/OpenMP/distribute_codegen.cpp (original)
+++ cfe/trunk/test/OpenMP/distribute_codegen.cpp Wed Aug 9 12:38:53 2017
@@ -34,7 +34,7 @@ void without_schedule_clause(float *a, f
}
}
-// CHECK: define {{.*}}void @.omp_outlined.(i32* noalias [[GBL_TIDP:%.+]], i32* noalias [[BND_TID:%.+]], float** dereferenceable({{[0-9]+}}) [[APTR:%.+]], float** dereferenceable({{[0-9]+}}) [[BPTR:%.+]], float** dereferenceable({{[0-9]+}}) [[CPTR:%.+]], float** dereferenceable({{[0-9]+}}) [[DPTR:%.+]])
+// CHECK: define {{.*}}void @{{.+}}(i32* noalias [[GBL_TIDP:%.+]], i32* noalias [[BND_TID:%.+]], float** dereferenceable({{[0-9]+}}) [[APTR:%.+]], float** dereferenceable({{[0-9]+}}) [[BPTR:%.+]], float** dereferenceable({{[0-9]+}}) [[CPTR:%.+]], float** dereferenceable({{[0-9]+}}) [[DPTR:%.+]])
// CHECK: [[TID_ADDR:%.+]] = alloca i32*
// CHECK: [[IV:%.+iv]] = alloca i32
// CHECK: [[LB:%.+lb]] = alloca i32
Modified: cfe/trunk/test/OpenMP/parallel_codegen.cpp
URL: http://llvm.org/viewvc/llvm-project/cfe/trunk/test/OpenMP/parallel_codegen.cpp?rev=310511&r1=310510&r2=310511&view=diff
==============================================================================
--- cfe/trunk/test/OpenMP/parallel_codegen.cpp (original)
+++ cfe/trunk/test/OpenMP/parallel_codegen.cpp Wed Aug 9 12:38:53 2017
@@ -109,7 +109,7 @@ int main (int argc, char **argv) {
// CHECK: call {{.*}}void @{{.+terminate.*|abort}}(
// CHECK-NEXT: unreachable
// CHECK-NEXT: }
-// CHECK-DEBUG: define internal void [[OMP_OUTLINED]](i32* noalias %.global_tid., i32* noalias %.bound_tid., i8*** dereferenceable({{4|8}}) %argc)
+// CHECK-DEBUG: define internal void [[OMP_OUTLINED_DEBUG:@.+]](i32* noalias %.global_tid., i32* noalias %.bound_tid., i8*** dereferenceable({{4|8}}) %argc)
// CHECK-DEBUG: store i8*** %argc, i8**** [[ARGC_PTR_ADDR:%.+]],
// CHECK-DEBUG: [[ARGC_REF:%.+]] = load i8***, i8**** [[ARGC_PTR_ADDR]]
// CHECK-DEBUG-NEXT: [[ARGC:%.+]] = load i8**, i8*** [[ARGC_REF]]
@@ -120,7 +120,9 @@ int main (int argc, char **argv) {
// CHECK-DEBUG-NEXT: }
// CHECK: define linkonce_odr {{.*}}void [[FOO1]](i8** %argc)
-// CHECK-DEBUG: define linkonce_odr void [[FOO1]](i8** %argc)
+// CHECK-DEBUG-DAG: define linkonce_odr void [[FOO1]](i8** %argc)
+// CHECK-DEBUG-DAG: define internal void [[OMP_OUTLINED]](i32* noalias %.global_tid., i32* noalias %.bound_tid., i8*** dereferenceable({{4|8}}) %argc)
+// CHECK-DEBUG-DAG: call void [[OMP_OUTLINED_DEBUG]]
// CHECK: attributes #[[FN_ATTRS]] = {{.+}} nounwind
// CHECK-DEBUG: attributes #[[FN_ATTRS]] = {{.+}} nounwind
Modified: cfe/trunk/test/OpenMP/parallel_for_simd_codegen.cpp
URL: http://llvm.org/viewvc/llvm-project/cfe/trunk/test/OpenMP/parallel_for_simd_codegen.cpp?rev=310511&r1=310510&r2=310511&view=diff
==============================================================================
--- cfe/trunk/test/OpenMP/parallel_for_simd_codegen.cpp (original)
+++ cfe/trunk/test/OpenMP/parallel_for_simd_codegen.cpp Wed Aug 9 12:38:53 2017
@@ -114,6 +114,7 @@ void simple(float *a, float *b, float *c
int lin = 12;
#pragma omp parallel for simd linear(lin : get_val()), linear(g_ptr)
+// CHECK: alloca i32,
// Init linear private var.
// CHECK: [[LIN_VAR:%.+]] = load i32*, i32** %
// CHECK: [[LIN_LOAD:%.+]] = load i32, i32* [[LIN_VAR]]
Modified: cfe/trunk/test/OpenMP/target_parallel_debug_codegen.cpp
URL: http://llvm.org/viewvc/llvm-project/cfe/trunk/test/OpenMP/target_parallel_debug_codegen.cpp?rev=310511&r1=310510&r2=310511&view=diff
==============================================================================
--- cfe/trunk/test/OpenMP/target_parallel_debug_codegen.cpp (original)
+++ cfe/trunk/test/OpenMP/target_parallel_debug_codegen.cpp Wed Aug 9 12:38:53 2017
@@ -58,7 +58,7 @@ int main() {
// CHECK: addrspacecast [10 x [10 x [10 x i32]]]* %{{.+}} to [10 x [10 x [10 x i32]]] addrspace(1)*
// CHECK: call void [[DEBUG_PARALLEL]](i32* {{[^,]+}}, i32* {{[^,]+}}, [10 x [10 x [10 x i32]]] addrspace(1)* {{[^,]+}}, i32 {{[^,]+}}, [10 x [10 x i32]]* {{[^)]+}})
-// CHECK: define void @__nondebug_wrapper___omp_offloading_{{[^(]+}}([10 x [10 x [10 x i32]]]* dereferenceable{{[^,]+}}, i64 {{[^,]+}}, [10 x [10 x i32]]* dereferenceable{{[^)]+}})
+// CHECK: define void @__omp_offloading_{{[^(]+}}([10 x [10 x [10 x i32]]]* dereferenceable{{[^,]+}}, i64 {{[^,]+}}, [10 x [10 x i32]]* dereferenceable{{[^)]+}})
// CHECK: addrspacecast [10 x [10 x [10 x i32]]]* %{{.+}} to [10 x [10 x [10 x i32]]] addrspace(1)*
// CHECK: call void @__omp_offloading_{{[^(]+}}([10 x [10 x [10 x i32]]] addrspace(1)* {{[^,]+}}, i32 {{[^,]+}}, [10 x [10 x i32]]* {{[^)]+}})
@@ -76,22 +76,31 @@ int main() {
// CHECK: addrspacecast [10 x [10 x i32]]* %{{.+}} to [10 x [10 x i32]] addrspace(1)*
// CHECK: call void [[DEBUG_PARALLEL]](i32* {{[^,]+}}, i32* {{[^,]+}}, [10 x [10 x [10 x i32]]] addrspace(1)* {{[^,]+}}, i32 {{[^,]+}}, [10 x [10 x i32]] addrspace(1)* {{[^)]+}})
-// CHECK: define void @__nondebug_wrapper___omp_offloading_{{[^(]+}}([10 x [10 x [10 x i32]]]* dereferenceable{{[^,]+}}, i64 {{[^,]+}}, [10 x [10 x i32]]* dereferenceable{{[^)]+}})
+// CHECK: define void @__omp_offloading_{{[^(]+}}([10 x [10 x [10 x i32]]]* dereferenceable{{[^,]+}}, i64 {{[^,]+}}, [10 x [10 x i32]]* dereferenceable{{[^)]+}})
// CHECK: addrspacecast [10 x [10 x [10 x i32]]]* %{{.+}} to [10 x [10 x [10 x i32]]] addrspace(1)*
// CHECK: addrspacecast [10 x [10 x i32]]* %{{.+}} to [10 x [10 x i32]] addrspace(1)*
// CHECK: call void @__omp_offloading_{{[^(]+}}([10 x [10 x [10 x i32]]] addrspace(1)* {{[^,]+}}, i32 {{[^,]+}}, [10 x [10 x i32]] addrspace(1)* {{[^)]+}})
-// CHECK: define void @__omp_offloading_{{[^(]+}}([10 x [10 x [10 x i32]]] addrspace(1)* noalias {{[^,]+}}, i32 addrspace(1)* noalias {{[^,]+}}, [10 x [10 x i32]] addrspace(1)* noalias {{[^)]+}})
+// CHECK: define internal void @__omp_offloading_{{[^(]+}}([10 x [10 x [10 x i32]]] addrspace(1)* noalias {{[^,]+}}, i32 addrspace(1)* noalias {{[^,]+}}, [10 x [10 x i32]] addrspace(1)* noalias {{[^)]+}})
// CHECK: addrspacecast [10 x [10 x [10 x i32]]] addrspace(1)* %{{.+}} to [10 x [10 x [10 x i32]]]*
// CHECK: addrspacecast i32 addrspace(1)* %{{.+}} to i32*
// CHECK: addrspacecast [10 x [10 x i32]] addrspace(1)* %{{.+}} to [10 x [10 x i32]]*
-// CHECK: addrspacecast [10 x [10 x [10 x i32]]]* %{{.+}} to [10 x [10 x [10 x i32]]] addrspace(1)*
-// CHECK: addrspacecast i32* %{{.+}} to i32 addrspace(1)*
-// CHECK: addrspacecast [10 x [10 x i32]]* %{{.+}} to [10 x [10 x i32]] addrspace(1)*
-// CHECK: call void [[DEBUG_PARALLEL:@.+]](i32* {{[^,]+}}, i32* {{[^,]+}}, [10 x [10 x [10 x i32]]] addrspace(1)* {{[^,]+}}, i32 addrspace(1)* {{[^,]+}}, [10 x [10 x i32]] addrspace(1)* {{[^)]+}})
+// CHECK: call void [[NONDEBUG_WRAPPER:@.+]](i32* {{[^,]+}}, i32* {{[^,]+}}, [10 x [10 x [10 x i32]]]* {{[^,]+}}, i32* {{[^,]+}}, [10 x [10 x i32]]* {{[^)]+}})
-// CHECK: define internal void [[DEBUG_PARALLEL]](i32* {{[^,]+}}, i32* {{[^,]+}}, [10 x [10 x [10 x i32]]] addrspace(1)* noalias{{[^,]+}}, i32 addrspace(1)* noalias{{[^,]+}}, [10 x [10 x i32]] addrspace(1)* noalias{{[^)]+}})
+// CHECK: define internal void [[DEBUG_PARALLEL:@.+]](i32* {{[^,]+}}, i32* {{[^,]+}}, [10 x [10 x [10 x i32]]] addrspace(1)* noalias{{[^,]+}}, i32 addrspace(1)* noalias{{[^,]+}}, [10 x [10 x i32]] addrspace(1)* noalias{{[^)]+}})
// CHECK: addrspacecast [10 x [10 x [10 x i32]]] addrspace(1)* %{{.+}} to [10 x [10 x [10 x i32]]]*
// CHECK: addrspacecast i32 addrspace(1)* %{{.+}} to i32*
// CHECK: addrspacecast [10 x [10 x i32]] addrspace(1)* %{{.+}} to [10 x [10 x i32]]*
+// CHECK: define internal void [[NONDEBUG_WRAPPER]](i32* {{[^,]+}}, i32* {{[^,]+}}, [10 x [10 x [10 x i32]]]* dereferenceable{{[^,]+}}, i32* dereferenceable{{[^,]+}}, [10 x [10 x i32]]* dereferenceable{{[^)]+}})
+// CHECK: addrspacecast [10 x [10 x [10 x i32]]]* %{{.+}} to [10 x [10 x [10 x i32]]] addrspace(1)*
+// CHECK: addrspacecast i32* %{{.+}} to i32 addrspace(1)*
+// CHECK: addrspacecast [10 x [10 x i32]]* %{{.+}} to [10 x [10 x i32]] addrspace(1)*
+// CHECK: call void [[DEBUG_PARALLEL]](i32* {{[^,]+}}, i32* {{[^,]+}}, [10 x [10 x [10 x i32]]] addrspace(1)* {{[^,]+}}, i32 addrspace(1)* {{[^,]+}}, [10 x [10 x i32]] addrspace(1)* {{[^)]+}})
+
+// CHECK: define void @__omp_offloading_{{[^(]+}}([10 x [10 x [10 x i32]]]* dereferenceable{{[^,]+}}, i32* dereferenceable{{[^,]+}}, [10 x [10 x i32]]* dereferenceable{{[^)]+}})
+// CHECK: addrspacecast [10 x [10 x [10 x i32]]]* %{{.+}} to [10 x [10 x [10 x i32]]] addrspace(1)*
+// CHECK: addrspacecast i32* %{{.+}} to i32 addrspace(1)*
+// CHECK: addrspacecast [10 x [10 x i32]]* %{{.+}} to [10 x [10 x i32]] addrspace(1)*
+// CHECK: call void @__omp_offloading_{{[^(]+}}([10 x [10 x [10 x i32]]] addrspace(1)* {{[^,]+}}, i32 addrspace(1)* {{[^,]+}}, [10 x [10 x i32]] addrspace(1)* {{[^)]+}})
+
More information about the cfe-commits
mailing list