aboutsummaryrefslogtreecommitdiff
path: root/clang/lib/CodeGen
diff options
context:
space:
mode:
Diffstat (limited to 'clang/lib/CodeGen')
-rw-r--r--clang/lib/CodeGen/BackendUtil.cpp5
-rw-r--r--clang/lib/CodeGen/CGDebugInfo.cpp64
-rw-r--r--clang/lib/CodeGen/CGDebugInfo.h1
-rw-r--r--clang/lib/CodeGen/CGStmt.cpp3
-rw-r--r--clang/lib/CodeGen/CGStmtOpenMP.cpp42
-rw-r--r--clang/lib/CodeGen/CodeGenFunction.h1
-rw-r--r--clang/lib/CodeGen/TargetBuiltins/AMDGPU.cpp27
7 files changed, 110 insertions, 33 deletions
diff --git a/clang/lib/CodeGen/BackendUtil.cpp b/clang/lib/CodeGen/BackendUtil.cpp
index 57db20f7..64f1917 100644
--- a/clang/lib/CodeGen/BackendUtil.cpp
+++ b/clang/lib/CodeGen/BackendUtil.cpp
@@ -1090,8 +1090,9 @@ void EmitAssemblyHelper::RunOptimizationPipeline(
if (std::optional<GCOVOptions> Options =
getGCOVOptions(CodeGenOpts, LangOpts))
PB.registerPipelineStartEPCallback(
- [Options](ModulePassManager &MPM, OptimizationLevel Level) {
- MPM.addPass(GCOVProfilerPass(*Options));
+ [this, Options](ModulePassManager &MPM, OptimizationLevel Level) {
+ MPM.addPass(
+ GCOVProfilerPass(*Options, CI.getVirtualFileSystemPtr()));
});
if (std::optional<InstrProfOptions> Options =
getInstrProfOptions(CodeGenOpts, LangOpts))
diff --git a/clang/lib/CodeGen/CGDebugInfo.cpp b/clang/lib/CodeGen/CGDebugInfo.cpp
index 12c7d48..fee6bc0 100644
--- a/clang/lib/CodeGen/CGDebugInfo.cpp
+++ b/clang/lib/CodeGen/CGDebugInfo.cpp
@@ -26,6 +26,7 @@
#include "clang/AST/DeclObjC.h"
#include "clang/AST/DeclTemplate.h"
#include "clang/AST/Expr.h"
+#include "clang/AST/LambdaCapture.h"
#include "clang/AST/RecordLayout.h"
#include "clang/AST/RecursiveASTVisitor.h"
#include "clang/AST/VTableBuilder.h"
@@ -1903,46 +1904,61 @@ CGDebugInfo::createInlinedSubprogram(StringRef FuncName,
return SP;
}
+llvm::StringRef
+CGDebugInfo::GetLambdaCaptureName(const LambdaCapture &Capture) {
+ if (Capture.capturesThis())
+ return CGM.getCodeGenOpts().EmitCodeView ? "__this" : "this";
+
+ assert(Capture.capturesVariable());
+
+ const ValueDecl *CaptureDecl = Capture.getCapturedVar();
+ assert(CaptureDecl && "Expected valid decl for captured variable.");
+
+ return CaptureDecl->getName();
+}
+
void CGDebugInfo::CollectRecordLambdaFields(
const CXXRecordDecl *CXXDecl, SmallVectorImpl<llvm::Metadata *> &elements,
llvm::DIType *RecordTy) {
// For C++11 Lambdas a Field will be the same as a Capture, but the Capture
// has the name and the location of the variable so we should iterate over
// both concurrently.
- const ASTRecordLayout &layout = CGM.getContext().getASTRecordLayout(CXXDecl);
RecordDecl::field_iterator Field = CXXDecl->field_begin();
unsigned fieldno = 0;
for (CXXRecordDecl::capture_const_iterator I = CXXDecl->captures_begin(),
E = CXXDecl->captures_end();
I != E; ++I, ++Field, ++fieldno) {
- const LambdaCapture &C = *I;
- if (C.capturesVariable()) {
- SourceLocation Loc = C.getLocation();
- assert(!Field->isBitField() && "lambdas don't have bitfield members!");
- ValueDecl *V = C.getCapturedVar();
- StringRef VName = V->getName();
- llvm::DIFile *VUnit = getOrCreateFile(Loc);
- auto Align = getDeclAlignIfRequired(V, CGM.getContext());
- llvm::DIType *FieldType = createFieldType(
- VName, Field->getType(), Loc, Field->getAccess(),
- layout.getFieldOffset(fieldno), Align, VUnit, RecordTy, CXXDecl);
- elements.push_back(FieldType);
- } else if (C.capturesThis()) {
+ const LambdaCapture &Capture = *I;
+ const uint64_t FieldOffset =
+ CGM.getContext().getASTRecordLayout(CXXDecl).getFieldOffset(fieldno);
+
+ assert(!Field->isBitField() && "lambdas don't have bitfield members!");
+
+ SourceLocation Loc;
+ uint32_t Align = 0;
+
+ if (Capture.capturesThis()) {
// TODO: Need to handle 'this' in some way by probably renaming the
// this of the lambda class and having a field member of 'this' or
// by using AT_object_pointer for the function and having that be
// used as 'this' for semantic references.
- FieldDecl *f = *Field;
- llvm::DIFile *VUnit = getOrCreateFile(f->getLocation());
- QualType type = f->getType();
- StringRef ThisName =
- CGM.getCodeGenOpts().EmitCodeView ? "__this" : "this";
- llvm::DIType *fieldType = createFieldType(
- ThisName, type, f->getLocation(), f->getAccess(),
- layout.getFieldOffset(fieldno), VUnit, RecordTy, CXXDecl);
-
- elements.push_back(fieldType);
+ Loc = Field->getLocation();
+ } else if (Capture.capturesVariable()) {
+ Loc = Capture.getLocation();
+
+ const ValueDecl *CaptureDecl = Capture.getCapturedVar();
+ assert(CaptureDecl && "Expected valid decl for captured variable.");
+
+ Align = getDeclAlignIfRequired(CaptureDecl, CGM.getContext());
+ } else {
+ continue;
}
+
+ llvm::DIFile *VUnit = getOrCreateFile(Loc);
+
+ elements.push_back(createFieldType(
+ GetLambdaCaptureName(Capture), Field->getType(), Loc,
+ Field->getAccess(), FieldOffset, Align, VUnit, RecordTy, CXXDecl));
}
}
diff --git a/clang/lib/CodeGen/CGDebugInfo.h b/clang/lib/CodeGen/CGDebugInfo.h
index f860773..78c3eb9 100644
--- a/clang/lib/CodeGen/CGDebugInfo.h
+++ b/clang/lib/CodeGen/CGDebugInfo.h
@@ -397,6 +397,7 @@ private:
void CollectRecordFields(const RecordDecl *Decl, llvm::DIFile *F,
SmallVectorImpl<llvm::Metadata *> &E,
llvm::DICompositeType *RecordTy);
+ llvm::StringRef GetLambdaCaptureName(const LambdaCapture &Capture);
/// If the C++ class has vtable info then insert appropriate debug
/// info entry in EltTys vector.
diff --git a/clang/lib/CodeGen/CGStmt.cpp b/clang/lib/CodeGen/CGStmt.cpp
index e62bc76..92636f2 100644
--- a/clang/lib/CodeGen/CGStmt.cpp
+++ b/clang/lib/CodeGen/CGStmt.cpp
@@ -234,6 +234,9 @@ void CodeGenFunction::EmitStmt(const Stmt *S, ArrayRef<const Attr *> Attrs) {
case Stmt::OMPInterchangeDirectiveClass:
EmitOMPInterchangeDirective(cast<OMPInterchangeDirective>(*S));
break;
+ case Stmt::OMPFuseDirectiveClass:
+ EmitOMPFuseDirective(cast<OMPFuseDirective>(*S));
+ break;
case Stmt::OMPForDirectiveClass:
EmitOMPForDirective(cast<OMPForDirective>(*S));
break;
diff --git a/clang/lib/CodeGen/CGStmtOpenMP.cpp b/clang/lib/CodeGen/CGStmtOpenMP.cpp
index ba9c7c6..efc06a2 100644
--- a/clang/lib/CodeGen/CGStmtOpenMP.cpp
+++ b/clang/lib/CodeGen/CGStmtOpenMP.cpp
@@ -201,6 +201,24 @@ class OMPLoopScope : public CodeGenFunction::RunCleanupsScope {
} else {
llvm_unreachable("Unknown loop-based directive kind.");
}
+ doEmitPreinits(PreInits);
+ PreCondVars.restore(CGF);
+ }
+
+ void
+ emitPreInitStmt(CodeGenFunction &CGF,
+ const OMPCanonicalLoopSequenceTransformationDirective &S) {
+ const Stmt *PreInits;
+ if (const auto *Fuse = dyn_cast<OMPFuseDirective>(&S)) {
+ PreInits = Fuse->getPreInits();
+ } else {
+ llvm_unreachable(
+ "Unknown canonical loop sequence transform directive kind.");
+ }
+ doEmitPreinits(PreInits);
+ }
+
+ void doEmitPreinits(const Stmt *PreInits) {
if (PreInits) {
// CompoundStmts and DeclStmts are used as lists of PreInit statements and
// declarations. Since declarations must be visible in the the following
@@ -222,7 +240,6 @@ class OMPLoopScope : public CodeGenFunction::RunCleanupsScope {
CGF.EmitStmt(S);
}
}
- PreCondVars.restore(CGF);
}
public:
@@ -230,6 +247,11 @@ public:
: CodeGenFunction::RunCleanupsScope(CGF) {
emitPreInitStmt(CGF, S);
}
+ OMPLoopScope(CodeGenFunction &CGF,
+ const OMPCanonicalLoopSequenceTransformationDirective &S)
+ : CodeGenFunction::RunCleanupsScope(CGF) {
+ emitPreInitStmt(CGF, S);
+ }
};
class OMPSimdLexicalScope : public CodeGenFunction::LexicalScope {
@@ -1929,6 +1951,15 @@ public:
CGSI = new CodeGenFunction::CGCapturedStmtInfo(CR_OpenMP);
CapInfoRAII = new CodeGenFunction::CGCapturedStmtRAII(CGF, CGSI);
}
+ if (const auto *Dir =
+ dyn_cast<OMPCanonicalLoopSequenceTransformationDirective>(S)) {
+ // For simplicity we reuse the loop scope similarly to what we do with
+ // OMPCanonicalLoopNestTransformationDirective do by being a subclass
+ // of OMPLoopBasedDirective.
+ Scope = new OMPLoopScope(CGF, *Dir);
+ CGSI = new CodeGenFunction::CGCapturedStmtInfo(CR_OpenMP);
+ CapInfoRAII = new CodeGenFunction::CGCapturedStmtRAII(CGF, CGSI);
+ }
}
~OMPTransformDirectiveScopeRAII() {
if (!Scope)
@@ -1956,8 +1987,7 @@ static void emitBody(CodeGenFunction &CGF, const Stmt *S, const Stmt *NextLoop,
return;
}
if (SimplifiedS == NextLoop) {
- if (auto *Dir =
- dyn_cast<OMPCanonicalLoopNestTransformationDirective>(SimplifiedS))
+ if (auto *Dir = dyn_cast<OMPLoopTransformationDirective>(SimplifiedS))
SimplifiedS = Dir->getTransformedStmt();
if (const auto *CanonLoop = dyn_cast<OMPCanonicalLoop>(SimplifiedS))
SimplifiedS = CanonLoop->getLoopStmt();
@@ -2952,6 +2982,12 @@ void CodeGenFunction::EmitOMPInterchangeDirective(
EmitStmt(S.getTransformedStmt());
}
+void CodeGenFunction::EmitOMPFuseDirective(const OMPFuseDirective &S) {
+ // Emit the de-sugared statement
+ OMPTransformDirectiveScopeRAII FuseScope(*this, &S);
+ EmitStmt(S.getTransformedStmt());
+}
+
void CodeGenFunction::EmitOMPUnrollDirective(const OMPUnrollDirective &S) {
bool UseOMPIRBuilder = CGM.getLangOpts().OpenMPIRBuilder;
diff --git a/clang/lib/CodeGen/CodeGenFunction.h b/clang/lib/CodeGen/CodeGenFunction.h
index 727487b..f0565c1 100644
--- a/clang/lib/CodeGen/CodeGenFunction.h
+++ b/clang/lib/CodeGen/CodeGenFunction.h
@@ -3861,6 +3861,7 @@ public:
void EmitOMPUnrollDirective(const OMPUnrollDirective &S);
void EmitOMPReverseDirective(const OMPReverseDirective &S);
void EmitOMPInterchangeDirective(const OMPInterchangeDirective &S);
+ void EmitOMPFuseDirective(const OMPFuseDirective &S);
void EmitOMPForDirective(const OMPForDirective &S);
void EmitOMPForSimdDirective(const OMPForSimdDirective &S);
void EmitOMPScopeDirective(const OMPScopeDirective &S);
diff --git a/clang/lib/CodeGen/TargetBuiltins/AMDGPU.cpp b/clang/lib/CodeGen/TargetBuiltins/AMDGPU.cpp
index 07cf08c..6596ec0 100644
--- a/clang/lib/CodeGen/TargetBuiltins/AMDGPU.cpp
+++ b/clang/lib/CodeGen/TargetBuiltins/AMDGPU.cpp
@@ -192,9 +192,17 @@ static Value *emitFPIntBuiltin(CodeGenFunction &CGF,
return CGF.Builder.CreateCall(F, {Src0, Src1});
}
+static inline StringRef mapScopeToSPIRV(StringRef AMDGCNScope) {
+ if (AMDGCNScope == "agent")
+ return "device";
+ if (AMDGCNScope == "wavefront")
+ return "subgroup";
+ return AMDGCNScope;
+}
+
// For processing memory ordering and memory scope arguments of various
// amdgcn builtins.
-// \p Order takes a C++11 comptabile memory-ordering specifier and converts
+// \p Order takes a C++11 compatible memory-ordering specifier and converts
// it into LLVM's memory ordering specifier using atomic C ABI, and writes
// to \p AO. \p Scope takes a const char * and converts it into AMDGCN
// specific SyncScopeID and writes it to \p SSID.
@@ -227,6 +235,8 @@ void CodeGenFunction::ProcessOrderScopeAMDGCN(Value *Order, Value *Scope,
// Some of the atomic builtins take the scope as a string name.
StringRef scp;
if (llvm::getConstantStringInfo(Scope, scp)) {
+ if (getTarget().getTriple().isSPIRV())
+ scp = mapScopeToSPIRV(scp);
SSID = getLLVMContext().getOrInsertSyncScopeID(scp);
return;
}
@@ -238,13 +248,19 @@ void CodeGenFunction::ProcessOrderScopeAMDGCN(Value *Order, Value *Scope,
SSID = llvm::SyncScope::System;
break;
case 1: // __MEMORY_SCOPE_DEVICE
- SSID = getLLVMContext().getOrInsertSyncScopeID("agent");
+ if (getTarget().getTriple().isSPIRV())
+ SSID = getLLVMContext().getOrInsertSyncScopeID("device");
+ else
+ SSID = getLLVMContext().getOrInsertSyncScopeID("agent");
break;
case 2: // __MEMORY_SCOPE_WRKGRP
SSID = getLLVMContext().getOrInsertSyncScopeID("workgroup");
break;
case 3: // __MEMORY_SCOPE_WVFRNT
- SSID = getLLVMContext().getOrInsertSyncScopeID("wavefront");
+ if (getTarget().getTriple().isSPIRV())
+ SSID = getLLVMContext().getOrInsertSyncScopeID("subgroup");
+ else
+ SSID = getLLVMContext().getOrInsertSyncScopeID("wavefront");
break;
case 4: // __MEMORY_SCOPE_SINGLE
SSID = llvm::SyncScope::SingleThread;
@@ -1510,7 +1526,10 @@ Value *CodeGenFunction::EmitAMDGPUBuiltinExpr(unsigned BuiltinID,
//
// The global/flat cases need to use agent scope to consistently produce
// the native instruction instead of a cmpxchg expansion.
- SSID = getLLVMContext().getOrInsertSyncScopeID("agent");
+ if (getTarget().getTriple().isSPIRV())
+ SSID = getLLVMContext().getOrInsertSyncScopeID("device");
+ else
+ SSID = getLLVMContext().getOrInsertSyncScopeID("agent");
AO = AtomicOrdering::Monotonic;
// The v2bf16 builtin uses i16 instead of a natural bfloat type.
generated by cgit v1.1 at 2025-09-30 14:32:35 +0000