Files
swift-mirror/include/swift/AST/IRGenOptions.h
Adrian Prantl 4a23512068 Add an option to disable DWARF skeleton CU breadcrumbs for Clang
module imports. This is useful when building redistributable static
archives, since any pointers into the CLang module cache won't be
portable.

When using this option the Clang type fallback path in LLDB will be
less useful since DWARF type information from those modules will not
be available unless another object file compiled without the option
imported the same modules.

rdar://problem/48827784
2019-04-03 15:45:11 -07:00

290 lines
9.3 KiB
C++

//===--- IRGenOptions.h - Swift Language IR Generation Options --*- C++ -*-===//
//
// This source file is part of the Swift.org open source project
//
// Copyright (c) 2014 - 2017 Apple Inc. and the Swift project authors
// Licensed under Apache License v2.0 with Runtime Library Exception
//
// See https://swift.org/LICENSE.txt for license information
// See https://swift.org/CONTRIBUTORS.txt for the list of Swift project authors
//
//===----------------------------------------------------------------------===//
//
// This file defines the options which control the generation of IR for
// swift files.
//
//===----------------------------------------------------------------------===//
#ifndef SWIFT_AST_IRGENOPTIONS_H
#define SWIFT_AST_IRGENOPTIONS_H
#include "swift/AST/LinkLibrary.h"
#include "swift/Basic/PathRemapper.h"
#include "swift/Basic/Sanitizers.h"
#include "swift/Basic/OptionSet.h"
#include "swift/Basic/OptimizationMode.h"
// FIXME: This include is just for llvm::SanitizerCoverageOptions. We should
// split the header upstream so we don't include so much.
#include "llvm/Transforms/Instrumentation.h"
#include <string>
#include <vector>
namespace swift {
enum class IRGenOutputKind : unsigned {
/// Just generate an LLVM module and return it.
Module,
/// Generate an LLVM module and write it out as LLVM assembly.
LLVMAssembly,
/// Generate an LLVM module and write it out as LLVM bitcode.
LLVMBitcode,
/// Generate an LLVM module and compile it to assembly.
NativeAssembly,
/// Generate an LLVM module, compile it, and assemble into an object file.
ObjectFile
};
enum class IRGenDebugInfoLevel : unsigned {
None, ///< No debug info.
LineTables, ///< Line tables only.
ASTTypes, ///< Line tables + AST type references.
DwarfTypes, ///< Line tables + AST type references + DWARF types.
Normal = ASTTypes ///< The setting LLDB prefers.
};
enum class IRGenDebugInfoFormat : unsigned {
None,
DWARF,
CodeView
};
enum class IRGenEmbedMode : unsigned {
None,
EmbedMarker,
EmbedBitcode
};
/// The set of options supported by IR generation.
class IRGenOptions {
public:
std::string ModuleName;
/// The compilation directory for the debug info.
std::string DebugCompilationDir;
/// The DWARF version of debug info.
unsigned DWARFVersion;
/// The command line string that is to be stored in the debug info.
std::string DebugFlags;
/// List of -Xcc -D macro definitions.
std::vector<std::string> ClangDefines;
/// The libraries and frameworks specified on the command line.
SmallVector<LinkLibrary, 4> LinkLibraries;
/// If non-empty, the (unmangled) name of a dummy symbol to emit that can be
/// used to force-load this module.
std::string ForceLoadSymbolName;
/// The kind of compilation we should do.
IRGenOutputKind OutputKind : 3;
/// Should we spend time verifying that the IR we produce is
/// well-formed?
unsigned Verify : 1;
OptimizationMode OptMode;
/// Which sanitizer is turned on.
OptionSet<SanitizerKind> Sanitizers;
/// Path prefixes that should be rewritten in debug info.
PathRemapper DebugPrefixMap;
/// What level of debug info to generate.
IRGenDebugInfoLevel DebugInfoLevel : 2;
/// What type of debug info to generate.
IRGenDebugInfoFormat DebugInfoFormat : 2;
/// Whether to leave DWARF breadcrumbs pointing to imported Clang modules.
unsigned DisableClangModuleSkeletonCUs : 1;
/// Whether we're generating IR for the JIT.
unsigned UseJIT : 1;
/// Whether we're generating code for the integrated REPL.
unsigned IntegratedREPL : 1;
/// Whether we should run LLVM optimizations after IRGen.
unsigned DisableLLVMOptzns : 1;
/// Whether we should run swift specific LLVM optimizations after IRGen.
unsigned DisableSwiftSpecificLLVMOptzns : 1;
/// Whether we should run LLVM SLP vectorizer.
unsigned DisableLLVMSLPVectorizer : 1;
/// Disable frame pointer elimination?
unsigned DisableFPElim : 1;
/// Special codegen for playgrounds.
unsigned Playground : 1;
/// Emit runtime calls to check the end of the lifetime of stack promoted
/// objects.
unsigned EmitStackPromotionChecks : 1;
/// The maximum number of bytes used on a stack frame for stack promotion
/// (includes alloc_stack allocations).
unsigned StackPromotionSizeLimit = 1024;
/// Emit code to verify that static and runtime type layout are consistent for
/// the given type names.
SmallVector<StringRef, 1> VerifyTypeLayoutNames;
/// Frameworks that we should not autolink against.
SmallVector<std::string, 1> DisableAutolinkFrameworks;
/// Print the LLVM inline tree at the end of the LLVM pass pipeline.
unsigned PrintInlineTree : 1;
/// Whether we should embed the bitcode file.
IRGenEmbedMode EmbedMode : 2;
/// Add names to LLVM values.
unsigned HasValueNamesSetting : 1;
unsigned ValueNames : 1;
/// Emit nominal type field metadata.
unsigned EnableReflectionMetadata : 1;
/// Emit names of struct stored properties and enum cases.
unsigned EnableReflectionNames : 1;
/// Emit mangled names of anonymous context descriptors.
unsigned EnableAnonymousContextMangledNames : 1;
/// Force public linkage for private symbols. Used only by the LLDB
/// expression evaluator.
unsigned ForcePublicLinkage : 1;
/// Force lazy initialization of class metadata
/// Used on Windows to avoid cross-module references.
unsigned LazyInitializeClassMetadata : 1;
unsigned LazyInitializeProtocolConformances : 1;
/// Normally if the -read-legacy-type-info flag is not specified, we look for
/// a file named "legacy-<arch>.yaml" in SearchPathOpts.RuntimeLibraryPath.
/// Passing this flag completely disables this behavior.
unsigned DisableLegacyTypeInfo : 1;
/// The path to load legacy type layouts from.
StringRef ReadLegacyTypeInfoPath;
/// Should we try to build incrementally by not emitting an object file if it
/// has the same IR hash as the module that we are preparing to emit?
///
/// This is a debugging option meant to make it easier to perform compile time
/// measurements on a non-clean build directory.
unsigned UseIncrementalLLVMCodeGen : 1;
/// Enable use of the swiftcall calling convention.
unsigned UseSwiftCall : 1;
/// Instrument code to generate profiling information.
unsigned GenerateProfile : 1;
/// Enable chaining of dynamic replacements.
unsigned EnableDynamicReplacementChaining : 1;
/// Disable round-trip verification of mangled debug types.
unsigned DisableRoundTripDebugTypes : 1;
/// Path to the profdata file to be used for PGO, or the empty string.
std::string UseProfile = "";
/// List of backend command-line options for -embed-bitcode.
std::vector<uint8_t> CmdArgs;
/// Which sanitizer coverage is turned on.
llvm::SanitizerCoverageOptions SanitizeCoverage;
/// The different modes for dumping IRGen type info.
enum class TypeInfoDumpFilter {
All,
Resilient,
Fragile
};
TypeInfoDumpFilter TypeInfoFilter;
IRGenOptions()
: DWARFVersion(2), OutputKind(IRGenOutputKind::LLVMAssembly),
Verify(true), OptMode(OptimizationMode::NotSet),
Sanitizers(OptionSet<SanitizerKind>()),
DebugInfoLevel(IRGenDebugInfoLevel::None),
DebugInfoFormat(IRGenDebugInfoFormat::None),
DisableClangModuleSkeletonCUs(false),
UseJIT(false),
IntegratedREPL(false), DisableLLVMOptzns(false),
DisableSwiftSpecificLLVMOptzns(false), DisableLLVMSLPVectorizer(false),
DisableFPElim(true), Playground(false), EmitStackPromotionChecks(false),
PrintInlineTree(false), EmbedMode(IRGenEmbedMode::None),
HasValueNamesSetting(false), ValueNames(false),
EnableReflectionMetadata(true), EnableReflectionNames(true),
EnableAnonymousContextMangledNames(false), ForcePublicLinkage(false),
LazyInitializeClassMetadata(false),
LazyInitializeProtocolConformances(false), DisableLegacyTypeInfo(false),
UseIncrementalLLVMCodeGen(true), UseSwiftCall(false),
GenerateProfile(false), EnableDynamicReplacementChaining(false),
DisableRoundTripDebugTypes(false), CmdArgs(),
SanitizeCoverage(llvm::SanitizerCoverageOptions()),
TypeInfoFilter(TypeInfoDumpFilter::All) {}
// Get a hash of all options which influence the llvm compilation but are not
// reflected in the llvm module itself.
unsigned getLLVMCodeGenOptionsHash() {
unsigned Hash = (unsigned)OptMode;
Hash = (Hash << 1) | DisableLLVMOptzns;
Hash = (Hash << 1) | DisableSwiftSpecificLLVMOptzns;
return Hash;
}
/// Should LLVM IR value names be emitted and preserved?
bool shouldProvideValueNames() const {
// If the command line contains an explicit request about whether to add
// LLVM value names, honor it. Otherwise, add value names only if the
// final result is textual LLVM assembly.
if (HasValueNamesSetting) {
return ValueNames;
} else {
return OutputKind == IRGenOutputKind::LLVMAssembly;
}
}
bool shouldOptimize() const {
return OptMode > OptimizationMode::NoOptimization;
}
bool optimizeForSize() const {
return OptMode == OptimizationMode::ForSize;
}
/// Return a hash code of any components from these options that should
/// contribute to a Swift Bridging PCH hash.
llvm::hash_code getPCHHashComponents() const {
return llvm::hash_value(0);
}
};
} // end namespace swift
#endif