1
0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2024-10-21 20:12:56 +02:00
llvm-mirror/include/llvm/IR/DiagnosticInfo.h

1022 lines
43 KiB
C
Raw Normal View History

//===- llvm/IR/DiagnosticInfo.h - Diagnostic Declaration --------*- C++ -*-===//
//
// The LLVM Compiler Infrastructure
//
// This file is distributed under the University of Illinois Open Source
// License. See LICENSE.TXT for details.
//
//===----------------------------------------------------------------------===//
//
// This file declares the different classes involved in low level diagnostics.
//
// Diagnostics reporting is still done as part of the LLVMContext.
//===----------------------------------------------------------------------===//
#ifndef LLVM_IR_DIAGNOSTICINFO_H
#define LLVM_IR_DIAGNOSTICINFO_H
#include "llvm-c/Types.h"
[OptRemark,LDist] RFC: Add hotness attribute Summary: This is the first set of changes implementing the RFC from http://thread.gmane.org/gmane.comp.compilers.llvm.devel/98334 This is a cross-sectional patch; rather than implementing the hotness attribute for all optimization remarks and all passes in a patch set, it implements it for the 'missed-optimization' remark for Loop Distribution. My goal is to shake out the design issues before scaling it up to other types and passes. Hotness is computed as an integer as the multiplication of the block frequency with the function entry count. It's only printed in opt currently since clang prints the diagnostic fields directly. E.g.: remark: /tmp/t.c:3:3: loop not distributed: use -Rpass-analysis=loop-distribute for more info (hotness: 300) A new API added is similar to emitOptimizationRemarkMissed. The difference is that it additionally takes a code region that the diagnostic corresponds to. From this, hotness is computed using BFI. The new API is exposed via an analysis pass so that it can be made dependent on LazyBFI. (Thanks to Hal for the analysis pass idea.) This feature can all be enabled by setDiagnosticHotnessRequested in the LLVM context. If this is off, LazyBFI is not calculated (D22141) so there should be no overhead. A new command-line option is added to turn this on in opt. My plan is to switch all user of emitOptimizationRemark* to use this module instead. Reviewers: hfinkel Subscribers: rcox2, mzolotukhin, llvm-commits Differential Revision: http://reviews.llvm.org/D21771 llvm-svn: 275583
2016-07-15 19:23:20 +02:00
#include "llvm/ADT/Optional.h"
#include "llvm/ADT/SmallVector.h"
#include "llvm/ADT/StringRef.h"
#include "llvm/ADT/Twine.h"
#include "llvm/IR/DebugLoc.h"
#include "llvm/Support/CBindingWrapping.h"
Output optimization remarks in YAML (Re-committed after moving the template specialization under the yaml namespace. GCC was complaining about this.) This allows various presentation of this data using an external tool. This was first recommended here[1]. As an example, consider this module: 1 int foo(); 2 int bar(); 3 4 int baz() { 5 return foo() + bar(); 6 } The inliner generates these missed-optimization remarks today (the hotness information is pulled from PGO): remark: /tmp/s.c:5:10: foo will not be inlined into baz (hotness: 30) remark: /tmp/s.c:5:18: bar will not be inlined into baz (hotness: 30) Now with -pass-remarks-output=<yaml-file>, we generate this YAML file: --- !Missed Pass: inline Name: NotInlined DebugLoc: { File: /tmp/s.c, Line: 5, Column: 10 } Function: baz Hotness: 30 Args: - Callee: foo - String: will not be inlined into - Caller: baz ... --- !Missed Pass: inline Name: NotInlined DebugLoc: { File: /tmp/s.c, Line: 5, Column: 18 } Function: baz Hotness: 30 Args: - Callee: bar - String: will not be inlined into - Caller: baz ... This is a summary of the high-level decisions: * There is a new streaming interface to emit optimization remarks. E.g. for the inliner remark above: ORE.emit(DiagnosticInfoOptimizationRemarkMissed( DEBUG_TYPE, "NotInlined", &I) << NV("Callee", Callee) << " will not be inlined into " << NV("Caller", CS.getCaller()) << setIsVerbose()); NV stands for named value and allows the YAML client to process a remark using its name (NotInlined) and the named arguments (Callee and Caller) without parsing the text of the message. Subsequent patches will update ORE users to use the new streaming API. * I am using YAML I/O for writing the YAML file. YAML I/O requires you to specify reading and writing at once but reading is highly non-trivial for some of the more complex LLVM types. Since it's not clear that we (ever) want to use LLVM to parse this YAML file, the code supports and asserts that we're writing only. On the other hand, I did experiment that the class hierarchy starting at DiagnosticInfoOptimizationBase can be mapped back from YAML generated here (see D24479). * The YAML stream is stored in the LLVM context. * In the example, we can probably further specify the IR value used, i.e. print "Function" rather than "Value". * As before hotness is computed in the analysis pass instead of DiganosticInfo. This avoids the layering problem since BFI is in Analysis while DiagnosticInfo is in IR. [1] https://reviews.llvm.org/D19678#419445 Differential Revision: https://reviews.llvm.org/D24587 llvm-svn: 282539
2016-09-27 22:55:07 +02:00
#include "llvm/Support/YAMLTraits.h"
#include <algorithm>
#include <cstdint>
#include <functional>
#include <iterator>
#include <string>
namespace llvm {
// Forward declarations.
class DiagnosticPrinter;
class Function;
class Instruction;
class LLVMContext;
class Module;
class SMDiagnostic;
/// \brief Defines the different supported severity of a diagnostic.
enum DiagnosticSeverity : char {
DS_Error,
DS_Warning,
DS_Remark,
// A note attaches additional information to one of the previous diagnostic
// types.
DS_Note
};
/// \brief Defines the different supported kind of a diagnostic.
/// This enum should be extended with a new ID for each added concrete subclass.
enum DiagnosticKind {
DK_InlineAsm,
DK_ResourceLimit,
DK_StackSize,
DK_Linker,
DK_DebugMetadataVersion,
DK_DebugMetadataInvalid,
DK_ISelFallback,
DK_SampleProfile,
DK_OptimizationRemark,
DK_OptimizationRemarkMissed,
DK_OptimizationRemarkAnalysis,
DK_OptimizationRemarkAnalysisFPCommute,
DK_OptimizationRemarkAnalysisAliasing,
DK_OptimizationFailure,
DK_FirstRemark = DK_OptimizationRemark,
DK_LastRemark = DK_OptimizationFailure,
DK_MachineOptimizationRemark,
DK_MachineOptimizationRemarkMissed,
DK_MachineOptimizationRemarkAnalysis,
DK_FirstMachineRemark = DK_MachineOptimizationRemark,
DK_LastMachineRemark = DK_MachineOptimizationRemarkAnalysis,
DK_MIRParser,
DK_PGOProfile,
DK_Unsupported,
DK_FirstPluginKind
};
/// \brief Get the next available kind ID for a plugin diagnostic.
/// Each time this function is called, it returns a different number.
/// Therefore, a plugin that wants to "identify" its own classes
/// with a dynamic identifier, just have to use this method to get a new ID
/// and assign it to each of its classes.
/// The returned ID will be greater than or equal to DK_FirstPluginKind.
/// Thus, the plugin identifiers will not conflict with the
/// DiagnosticKind values.
int getNextAvailablePluginDiagnosticKind();
/// \brief This is the base abstract class for diagnostic reporting in
/// the backend.
/// The print method must be overloaded by the subclasses to print a
/// user-friendly message in the client of the backend (let us call it a
/// frontend).
class DiagnosticInfo {
private:
/// Kind defines the kind of report this is about.
const /* DiagnosticKind */ int Kind;
/// Severity gives the severity of the diagnostic.
const DiagnosticSeverity Severity;
public:
DiagnosticInfo(/* DiagnosticKind */ int Kind, DiagnosticSeverity Severity)
: Kind(Kind), Severity(Severity) {}
virtual ~DiagnosticInfo() = default;
/* DiagnosticKind */ int getKind() const { return Kind; }
DiagnosticSeverity getSeverity() const { return Severity; }
/// Print using the given \p DP a user-friendly message.
/// This is the default message that will be printed to the user.
/// It is used when the frontend does not directly take advantage
/// of the information contained in fields of the subclasses.
/// The printed message must not end with '.' nor start with a severity
/// keyword.
virtual void print(DiagnosticPrinter &DP) const = 0;
};
using DiagnosticHandlerFunction = std::function<void(const DiagnosticInfo &)>;
Use the DiagnosticHandler to print diagnostics when reading bitcode. The bitcode reading interface used std::error_code to report an error to the callers and it is the callers job to print diagnostics. This is not ideal for error handling or diagnostic reporting: * For error handling, all that the callers care about is 3 possibilities: * It worked * The bitcode file is corrupted/invalid. * The file is not bitcode at all. * For diagnostic, it is user friendly to include far more information about the invalid case so the user can find out what is wrong with the bitcode file. This comes up, for example, when a developer introduces a bug while extending the format. The compromise we had was to have a lot of error codes. With this patch we use the DiagnosticHandler to communicate with the human and std::error_code to communicate with the caller. This allows us to have far fewer error codes and adds the infrastructure to print better diagnostics. This is so because the diagnostics are printed when he issue is found. The code that detected the problem in alive in the stack and can pass down as much context as needed. As an example the patch updates test/Bitcode/invalid.ll. Using a DiagnosticHandler also moves the fatal/non-fatal error decision to the caller. A simple one like llvm-dis can just use fatal errors. The gold plugin needs a bit more complex treatment because of being passed non-bitcode files. An hypothetical interactive tool would make all bitcode errors non-fatal. llvm-svn: 225562
2015-01-10 01:07:30 +01:00
/// Diagnostic information for inline asm reporting.
/// This is basically a message and an optional location.
class DiagnosticInfoInlineAsm : public DiagnosticInfo {
private:
/// Optional line information. 0 if not set.
unsigned LocCookie = 0;
/// Message to be reported.
const Twine &MsgStr;
/// Optional origin of the problem.
const Instruction *Instr = nullptr;
public:
/// \p MsgStr is the message to be reported to the frontend.
/// This class does not copy \p MsgStr, therefore the reference must be valid
/// for the whole life time of the Diagnostic.
DiagnosticInfoInlineAsm(const Twine &MsgStr,
DiagnosticSeverity Severity = DS_Error)
: DiagnosticInfo(DK_InlineAsm, Severity), MsgStr(MsgStr) {}
/// \p LocCookie if non-zero gives the line number for this report.
/// \p MsgStr gives the message.
/// This class does not copy \p MsgStr, therefore the reference must be valid
/// for the whole life time of the Diagnostic.
DiagnosticInfoInlineAsm(unsigned LocCookie, const Twine &MsgStr,
DiagnosticSeverity Severity = DS_Error)
: DiagnosticInfo(DK_InlineAsm, Severity), LocCookie(LocCookie),
MsgStr(MsgStr) {}
/// \p Instr gives the original instruction that triggered the diagnostic.
/// \p MsgStr gives the message.
/// This class does not copy \p MsgStr, therefore the reference must be valid
/// for the whole life time of the Diagnostic.
/// Same for \p I.
DiagnosticInfoInlineAsm(const Instruction &I, const Twine &MsgStr,
DiagnosticSeverity Severity = DS_Error);
unsigned getLocCookie() const { return LocCookie; }
const Twine &getMsgStr() const { return MsgStr; }
const Instruction *getInstruction() const { return Instr; }
/// \see DiagnosticInfo::print.
void print(DiagnosticPrinter &DP) const override;
static bool classof(const DiagnosticInfo *DI) {
return DI->getKind() == DK_InlineAsm;
}
};
/// Diagnostic information for stack size etc. reporting.
/// This is basically a function and a size.
class DiagnosticInfoResourceLimit : public DiagnosticInfo {
private:
/// The function that is concerned by this resource limit diagnostic.
const Function &Fn;
/// Description of the resource type (e.g. stack size)
const char *ResourceName;
/// The computed size usage
uint64_t ResourceSize;
// Threshould passed
uint64_t ResourceLimit;
public:
/// \p The function that is concerned by this stack size diagnostic.
/// \p The computed stack size.
DiagnosticInfoResourceLimit(const Function &Fn,
const char *ResourceName,
uint64_t ResourceSize,
DiagnosticSeverity Severity = DS_Warning,
DiagnosticKind Kind = DK_ResourceLimit,
uint64_t ResourceLimit = 0)
: DiagnosticInfo(Kind, Severity),
Fn(Fn),
ResourceName(ResourceName),
ResourceSize(ResourceSize),
ResourceLimit(ResourceLimit) {}
const Function &getFunction() const { return Fn; }
const char *getResourceName() const { return ResourceName; }
uint64_t getResourceSize() const { return ResourceSize; }
uint64_t getResourceLimit() const { return ResourceLimit; }
/// \see DiagnosticInfo::print.
void print(DiagnosticPrinter &DP) const override;
static bool classof(const DiagnosticInfo *DI) {
return DI->getKind() == DK_ResourceLimit ||
DI->getKind() == DK_StackSize;
}
};
class DiagnosticInfoStackSize : public DiagnosticInfoResourceLimit {
public:
DiagnosticInfoStackSize(const Function &Fn,
uint64_t StackSize,
DiagnosticSeverity Severity = DS_Warning,
uint64_t StackLimit = 0)
: DiagnosticInfoResourceLimit(Fn, "stack size", StackSize,
Severity, DK_StackSize, StackLimit) {}
uint64_t getStackSize() const { return getResourceSize(); }
uint64_t getStackLimit() const { return getResourceLimit(); }
static bool classof(const DiagnosticInfo *DI) {
return DI->getKind() == DK_StackSize;
}
};
/// Diagnostic information for debug metadata version reporting.
/// This is basically a module and a version.
class DiagnosticInfoDebugMetadataVersion : public DiagnosticInfo {
private:
/// The module that is concerned by this debug metadata version diagnostic.
const Module &M;
/// The actual metadata version.
unsigned MetadataVersion;
public:
/// \p The module that is concerned by this debug metadata version diagnostic.
/// \p The actual metadata version.
DiagnosticInfoDebugMetadataVersion(const Module &M, unsigned MetadataVersion,
DiagnosticSeverity Severity = DS_Warning)
: DiagnosticInfo(DK_DebugMetadataVersion, Severity), M(M),
MetadataVersion(MetadataVersion) {}
const Module &getModule() const { return M; }
unsigned getMetadataVersion() const { return MetadataVersion; }
/// \see DiagnosticInfo::print.
void print(DiagnosticPrinter &DP) const override;
static bool classof(const DiagnosticInfo *DI) {
return DI->getKind() == DK_DebugMetadataVersion;
}
};
/// Diagnostic information for stripping invalid debug metadata.
class DiagnosticInfoIgnoringInvalidDebugMetadata : public DiagnosticInfo {
private:
/// The module that is concerned by this debug metadata version diagnostic.
const Module &M;
public:
/// \p The module that is concerned by this debug metadata version diagnostic.
DiagnosticInfoIgnoringInvalidDebugMetadata(
const Module &M, DiagnosticSeverity Severity = DS_Warning)
: DiagnosticInfo(DK_DebugMetadataVersion, Severity), M(M) {}
const Module &getModule() const { return M; }
/// \see DiagnosticInfo::print.
void print(DiagnosticPrinter &DP) const override;
static bool classof(const DiagnosticInfo *DI) {
return DI->getKind() == DK_DebugMetadataInvalid;
}
};
/// Diagnostic information for the sample profiler.
class DiagnosticInfoSampleProfile : public DiagnosticInfo {
public:
DiagnosticInfoSampleProfile(StringRef FileName, unsigned LineNum,
const Twine &Msg,
DiagnosticSeverity Severity = DS_Error)
: DiagnosticInfo(DK_SampleProfile, Severity), FileName(FileName),
LineNum(LineNum), Msg(Msg) {}
DiagnosticInfoSampleProfile(StringRef FileName, const Twine &Msg,
DiagnosticSeverity Severity = DS_Error)
: DiagnosticInfo(DK_SampleProfile, Severity), FileName(FileName),
Msg(Msg) {}
DiagnosticInfoSampleProfile(const Twine &Msg,
DiagnosticSeverity Severity = DS_Error)
: DiagnosticInfo(DK_SampleProfile, Severity), Msg(Msg) {}
/// \see DiagnosticInfo::print.
void print(DiagnosticPrinter &DP) const override;
static bool classof(const DiagnosticInfo *DI) {
return DI->getKind() == DK_SampleProfile;
}
StringRef getFileName() const { return FileName; }
unsigned getLineNum() const { return LineNum; }
const Twine &getMsg() const { return Msg; }
private:
/// Name of the input file associated with this diagnostic.
StringRef FileName;
/// Line number where the diagnostic occurred. If 0, no line number will
/// be emitted in the message.
unsigned LineNum = 0;
/// Message to report.
const Twine &Msg;
};
/// Diagnostic information for the PGO profiler.
class DiagnosticInfoPGOProfile : public DiagnosticInfo {
public:
DiagnosticInfoPGOProfile(const char *FileName, const Twine &Msg,
DiagnosticSeverity Severity = DS_Error)
: DiagnosticInfo(DK_PGOProfile, Severity), FileName(FileName), Msg(Msg) {}
/// \see DiagnosticInfo::print.
void print(DiagnosticPrinter &DP) const override;
static bool classof(const DiagnosticInfo *DI) {
return DI->getKind() == DK_PGOProfile;
}
const char *getFileName() const { return FileName; }
const Twine &getMsg() const { return Msg; }
private:
/// Name of the input file associated with this diagnostic.
const char *FileName;
/// Message to report.
const Twine &Msg;
};
class DiagnosticLocation {
StringRef Filename;
unsigned Line = 0;
unsigned Column = 0;
public:
DiagnosticLocation() = default;
DiagnosticLocation(const DebugLoc &DL);
DiagnosticLocation(const DISubprogram *SP);
bool isValid() const { return !Filename.empty(); }
StringRef getFilename() const { return Filename; }
unsigned getLine() const { return Line; }
unsigned getColumn() const { return Column; }
};
/// Common features for diagnostics with an associated location.
class DiagnosticInfoWithLocationBase : public DiagnosticInfo {
public:
/// \p Fn is the function where the diagnostic is being emitted. \p Loc is
/// the location information to use in the diagnostic.
DiagnosticInfoWithLocationBase(enum DiagnosticKind Kind,
enum DiagnosticSeverity Severity,
const Function &Fn,
const DiagnosticLocation &Loc)
: DiagnosticInfo(Kind, Severity), Fn(Fn), Loc(Loc) {}
/// Return true if location information is available for this diagnostic.
bool isLocationAvailable() const { return Loc.isValid(); }
/// Return a string with the location information for this diagnostic
/// in the format "file:line:col". If location information is not available,
/// it returns "<unknown>:0:0".
const std::string getLocationStr() const;
/// Return location information for this diagnostic in three parts:
/// the source file name, line number and column.
void getLocation(StringRef *Filename, unsigned *Line, unsigned *Column) const;
const Function &getFunction() const { return Fn; }
DiagnosticLocation getLocation() const { return Loc; }
private:
/// Function where this diagnostic is triggered.
const Function &Fn;
/// Debug location where this diagnostic is triggered.
DiagnosticLocation Loc;
};
/// \brief Common features for diagnostics dealing with optimization remarks
/// that are used by both IR and MIR passes.
class DiagnosticInfoOptimizationBase : public DiagnosticInfoWithLocationBase {
public:
Output optimization remarks in YAML (Re-committed after moving the template specialization under the yaml namespace. GCC was complaining about this.) This allows various presentation of this data using an external tool. This was first recommended here[1]. As an example, consider this module: 1 int foo(); 2 int bar(); 3 4 int baz() { 5 return foo() + bar(); 6 } The inliner generates these missed-optimization remarks today (the hotness information is pulled from PGO): remark: /tmp/s.c:5:10: foo will not be inlined into baz (hotness: 30) remark: /tmp/s.c:5:18: bar will not be inlined into baz (hotness: 30) Now with -pass-remarks-output=<yaml-file>, we generate this YAML file: --- !Missed Pass: inline Name: NotInlined DebugLoc: { File: /tmp/s.c, Line: 5, Column: 10 } Function: baz Hotness: 30 Args: - Callee: foo - String: will not be inlined into - Caller: baz ... --- !Missed Pass: inline Name: NotInlined DebugLoc: { File: /tmp/s.c, Line: 5, Column: 18 } Function: baz Hotness: 30 Args: - Callee: bar - String: will not be inlined into - Caller: baz ... This is a summary of the high-level decisions: * There is a new streaming interface to emit optimization remarks. E.g. for the inliner remark above: ORE.emit(DiagnosticInfoOptimizationRemarkMissed( DEBUG_TYPE, "NotInlined", &I) << NV("Callee", Callee) << " will not be inlined into " << NV("Caller", CS.getCaller()) << setIsVerbose()); NV stands for named value and allows the YAML client to process a remark using its name (NotInlined) and the named arguments (Callee and Caller) without parsing the text of the message. Subsequent patches will update ORE users to use the new streaming API. * I am using YAML I/O for writing the YAML file. YAML I/O requires you to specify reading and writing at once but reading is highly non-trivial for some of the more complex LLVM types. Since it's not clear that we (ever) want to use LLVM to parse this YAML file, the code supports and asserts that we're writing only. On the other hand, I did experiment that the class hierarchy starting at DiagnosticInfoOptimizationBase can be mapped back from YAML generated here (see D24479). * The YAML stream is stored in the LLVM context. * In the example, we can probably further specify the IR value used, i.e. print "Function" rather than "Value". * As before hotness is computed in the analysis pass instead of DiganosticInfo. This avoids the layering problem since BFI is in Analysis while DiagnosticInfo is in IR. [1] https://reviews.llvm.org/D19678#419445 Differential Revision: https://reviews.llvm.org/D24587 llvm-svn: 282539
2016-09-27 22:55:07 +02:00
/// \brief Used to set IsVerbose via the stream interface.
struct setIsVerbose {};
/// \brief When an instance of this is inserted into the stream, the arguments
/// following will not appear in the remark printed in the compiler output
/// (-Rpass) but only in the optimization record file
/// (-fsave-optimization-record).
struct setExtraArgs {};
Output optimization remarks in YAML (Re-committed after moving the template specialization under the yaml namespace. GCC was complaining about this.) This allows various presentation of this data using an external tool. This was first recommended here[1]. As an example, consider this module: 1 int foo(); 2 int bar(); 3 4 int baz() { 5 return foo() + bar(); 6 } The inliner generates these missed-optimization remarks today (the hotness information is pulled from PGO): remark: /tmp/s.c:5:10: foo will not be inlined into baz (hotness: 30) remark: /tmp/s.c:5:18: bar will not be inlined into baz (hotness: 30) Now with -pass-remarks-output=<yaml-file>, we generate this YAML file: --- !Missed Pass: inline Name: NotInlined DebugLoc: { File: /tmp/s.c, Line: 5, Column: 10 } Function: baz Hotness: 30 Args: - Callee: foo - String: will not be inlined into - Caller: baz ... --- !Missed Pass: inline Name: NotInlined DebugLoc: { File: /tmp/s.c, Line: 5, Column: 18 } Function: baz Hotness: 30 Args: - Callee: bar - String: will not be inlined into - Caller: baz ... This is a summary of the high-level decisions: * There is a new streaming interface to emit optimization remarks. E.g. for the inliner remark above: ORE.emit(DiagnosticInfoOptimizationRemarkMissed( DEBUG_TYPE, "NotInlined", &I) << NV("Callee", Callee) << " will not be inlined into " << NV("Caller", CS.getCaller()) << setIsVerbose()); NV stands for named value and allows the YAML client to process a remark using its name (NotInlined) and the named arguments (Callee and Caller) without parsing the text of the message. Subsequent patches will update ORE users to use the new streaming API. * I am using YAML I/O for writing the YAML file. YAML I/O requires you to specify reading and writing at once but reading is highly non-trivial for some of the more complex LLVM types. Since it's not clear that we (ever) want to use LLVM to parse this YAML file, the code supports and asserts that we're writing only. On the other hand, I did experiment that the class hierarchy starting at DiagnosticInfoOptimizationBase can be mapped back from YAML generated here (see D24479). * The YAML stream is stored in the LLVM context. * In the example, we can probably further specify the IR value used, i.e. print "Function" rather than "Value". * As before hotness is computed in the analysis pass instead of DiganosticInfo. This avoids the layering problem since BFI is in Analysis while DiagnosticInfo is in IR. [1] https://reviews.llvm.org/D19678#419445 Differential Revision: https://reviews.llvm.org/D24587 llvm-svn: 282539
2016-09-27 22:55:07 +02:00
/// \brief Used in the streaming interface as the general argument type. It
/// internally converts everything into a key-value pair.
struct Argument {
StringRef Key;
std::string Val;
// If set, the debug location corresponding to the value.
DiagnosticLocation Loc;
Output optimization remarks in YAML (Re-committed after moving the template specialization under the yaml namespace. GCC was complaining about this.) This allows various presentation of this data using an external tool. This was first recommended here[1]. As an example, consider this module: 1 int foo(); 2 int bar(); 3 4 int baz() { 5 return foo() + bar(); 6 } The inliner generates these missed-optimization remarks today (the hotness information is pulled from PGO): remark: /tmp/s.c:5:10: foo will not be inlined into baz (hotness: 30) remark: /tmp/s.c:5:18: bar will not be inlined into baz (hotness: 30) Now with -pass-remarks-output=<yaml-file>, we generate this YAML file: --- !Missed Pass: inline Name: NotInlined DebugLoc: { File: /tmp/s.c, Line: 5, Column: 10 } Function: baz Hotness: 30 Args: - Callee: foo - String: will not be inlined into - Caller: baz ... --- !Missed Pass: inline Name: NotInlined DebugLoc: { File: /tmp/s.c, Line: 5, Column: 18 } Function: baz Hotness: 30 Args: - Callee: bar - String: will not be inlined into - Caller: baz ... This is a summary of the high-level decisions: * There is a new streaming interface to emit optimization remarks. E.g. for the inliner remark above: ORE.emit(DiagnosticInfoOptimizationRemarkMissed( DEBUG_TYPE, "NotInlined", &I) << NV("Callee", Callee) << " will not be inlined into " << NV("Caller", CS.getCaller()) << setIsVerbose()); NV stands for named value and allows the YAML client to process a remark using its name (NotInlined) and the named arguments (Callee and Caller) without parsing the text of the message. Subsequent patches will update ORE users to use the new streaming API. * I am using YAML I/O for writing the YAML file. YAML I/O requires you to specify reading and writing at once but reading is highly non-trivial for some of the more complex LLVM types. Since it's not clear that we (ever) want to use LLVM to parse this YAML file, the code supports and asserts that we're writing only. On the other hand, I did experiment that the class hierarchy starting at DiagnosticInfoOptimizationBase can be mapped back from YAML generated here (see D24479). * The YAML stream is stored in the LLVM context. * In the example, we can probably further specify the IR value used, i.e. print "Function" rather than "Value". * As before hotness is computed in the analysis pass instead of DiganosticInfo. This avoids the layering problem since BFI is in Analysis while DiagnosticInfo is in IR. [1] https://reviews.llvm.org/D19678#419445 Differential Revision: https://reviews.llvm.org/D24587 llvm-svn: 282539
2016-09-27 22:55:07 +02:00
explicit Argument(StringRef Str = "") : Key("String"), Val(Str) {}
Argument(StringRef Key, const Value *V);
Argument(StringRef Key, const Type *T);
Argument(StringRef Key, StringRef S);
Argument(StringRef Key, int N);
Argument(StringRef Key, int64_t N);
Argument(StringRef Key, unsigned N);
Argument(StringRef Key, uint64_t N);
Argument(StringRef Key, bool B) : Key(Key), Val(B ? "true" : "false") {}
Output optimization remarks in YAML (Re-committed after moving the template specialization under the yaml namespace. GCC was complaining about this.) This allows various presentation of this data using an external tool. This was first recommended here[1]. As an example, consider this module: 1 int foo(); 2 int bar(); 3 4 int baz() { 5 return foo() + bar(); 6 } The inliner generates these missed-optimization remarks today (the hotness information is pulled from PGO): remark: /tmp/s.c:5:10: foo will not be inlined into baz (hotness: 30) remark: /tmp/s.c:5:18: bar will not be inlined into baz (hotness: 30) Now with -pass-remarks-output=<yaml-file>, we generate this YAML file: --- !Missed Pass: inline Name: NotInlined DebugLoc: { File: /tmp/s.c, Line: 5, Column: 10 } Function: baz Hotness: 30 Args: - Callee: foo - String: will not be inlined into - Caller: baz ... --- !Missed Pass: inline Name: NotInlined DebugLoc: { File: /tmp/s.c, Line: 5, Column: 18 } Function: baz Hotness: 30 Args: - Callee: bar - String: will not be inlined into - Caller: baz ... This is a summary of the high-level decisions: * There is a new streaming interface to emit optimization remarks. E.g. for the inliner remark above: ORE.emit(DiagnosticInfoOptimizationRemarkMissed( DEBUG_TYPE, "NotInlined", &I) << NV("Callee", Callee) << " will not be inlined into " << NV("Caller", CS.getCaller()) << setIsVerbose()); NV stands for named value and allows the YAML client to process a remark using its name (NotInlined) and the named arguments (Callee and Caller) without parsing the text of the message. Subsequent patches will update ORE users to use the new streaming API. * I am using YAML I/O for writing the YAML file. YAML I/O requires you to specify reading and writing at once but reading is highly non-trivial for some of the more complex LLVM types. Since it's not clear that we (ever) want to use LLVM to parse this YAML file, the code supports and asserts that we're writing only. On the other hand, I did experiment that the class hierarchy starting at DiagnosticInfoOptimizationBase can be mapped back from YAML generated here (see D24479). * The YAML stream is stored in the LLVM context. * In the example, we can probably further specify the IR value used, i.e. print "Function" rather than "Value". * As before hotness is computed in the analysis pass instead of DiganosticInfo. This avoids the layering problem since BFI is in Analysis while DiagnosticInfo is in IR. [1] https://reviews.llvm.org/D19678#419445 Differential Revision: https://reviews.llvm.org/D24587 llvm-svn: 282539
2016-09-27 22:55:07 +02:00
};
/// \p PassName is the name of the pass emitting this diagnostic. \p
/// RemarkName is a textual identifier for the remark (single-word,
/// camel-case). \p Fn is the function where the diagnostic is being emitted.
/// \p Loc is the location information to use in the diagnostic. If line table
/// information is available, the diagnostic will include the source code
/// location.
Output optimization remarks in YAML (Re-committed after moving the template specialization under the yaml namespace. GCC was complaining about this.) This allows various presentation of this data using an external tool. This was first recommended here[1]. As an example, consider this module: 1 int foo(); 2 int bar(); 3 4 int baz() { 5 return foo() + bar(); 6 } The inliner generates these missed-optimization remarks today (the hotness information is pulled from PGO): remark: /tmp/s.c:5:10: foo will not be inlined into baz (hotness: 30) remark: /tmp/s.c:5:18: bar will not be inlined into baz (hotness: 30) Now with -pass-remarks-output=<yaml-file>, we generate this YAML file: --- !Missed Pass: inline Name: NotInlined DebugLoc: { File: /tmp/s.c, Line: 5, Column: 10 } Function: baz Hotness: 30 Args: - Callee: foo - String: will not be inlined into - Caller: baz ... --- !Missed Pass: inline Name: NotInlined DebugLoc: { File: /tmp/s.c, Line: 5, Column: 18 } Function: baz Hotness: 30 Args: - Callee: bar - String: will not be inlined into - Caller: baz ... This is a summary of the high-level decisions: * There is a new streaming interface to emit optimization remarks. E.g. for the inliner remark above: ORE.emit(DiagnosticInfoOptimizationRemarkMissed( DEBUG_TYPE, "NotInlined", &I) << NV("Callee", Callee) << " will not be inlined into " << NV("Caller", CS.getCaller()) << setIsVerbose()); NV stands for named value and allows the YAML client to process a remark using its name (NotInlined) and the named arguments (Callee and Caller) without parsing the text of the message. Subsequent patches will update ORE users to use the new streaming API. * I am using YAML I/O for writing the YAML file. YAML I/O requires you to specify reading and writing at once but reading is highly non-trivial for some of the more complex LLVM types. Since it's not clear that we (ever) want to use LLVM to parse this YAML file, the code supports and asserts that we're writing only. On the other hand, I did experiment that the class hierarchy starting at DiagnosticInfoOptimizationBase can be mapped back from YAML generated here (see D24479). * The YAML stream is stored in the LLVM context. * In the example, we can probably further specify the IR value used, i.e. print "Function" rather than "Value". * As before hotness is computed in the analysis pass instead of DiganosticInfo. This avoids the layering problem since BFI is in Analysis while DiagnosticInfo is in IR. [1] https://reviews.llvm.org/D19678#419445 Differential Revision: https://reviews.llvm.org/D24587 llvm-svn: 282539
2016-09-27 22:55:07 +02:00
DiagnosticInfoOptimizationBase(enum DiagnosticKind Kind,
enum DiagnosticSeverity Severity,
const char *PassName, StringRef RemarkName,
const Function &Fn,
const DiagnosticLocation &Loc)
: DiagnosticInfoWithLocationBase(Kind, Severity, Fn, Loc),
PassName(PassName), RemarkName(RemarkName) {}
Output optimization remarks in YAML (Re-committed after moving the template specialization under the yaml namespace. GCC was complaining about this.) This allows various presentation of this data using an external tool. This was first recommended here[1]. As an example, consider this module: 1 int foo(); 2 int bar(); 3 4 int baz() { 5 return foo() + bar(); 6 } The inliner generates these missed-optimization remarks today (the hotness information is pulled from PGO): remark: /tmp/s.c:5:10: foo will not be inlined into baz (hotness: 30) remark: /tmp/s.c:5:18: bar will not be inlined into baz (hotness: 30) Now with -pass-remarks-output=<yaml-file>, we generate this YAML file: --- !Missed Pass: inline Name: NotInlined DebugLoc: { File: /tmp/s.c, Line: 5, Column: 10 } Function: baz Hotness: 30 Args: - Callee: foo - String: will not be inlined into - Caller: baz ... --- !Missed Pass: inline Name: NotInlined DebugLoc: { File: /tmp/s.c, Line: 5, Column: 18 } Function: baz Hotness: 30 Args: - Callee: bar - String: will not be inlined into - Caller: baz ... This is a summary of the high-level decisions: * There is a new streaming interface to emit optimization remarks. E.g. for the inliner remark above: ORE.emit(DiagnosticInfoOptimizationRemarkMissed( DEBUG_TYPE, "NotInlined", &I) << NV("Callee", Callee) << " will not be inlined into " << NV("Caller", CS.getCaller()) << setIsVerbose()); NV stands for named value and allows the YAML client to process a remark using its name (NotInlined) and the named arguments (Callee and Caller) without parsing the text of the message. Subsequent patches will update ORE users to use the new streaming API. * I am using YAML I/O for writing the YAML file. YAML I/O requires you to specify reading and writing at once but reading is highly non-trivial for some of the more complex LLVM types. Since it's not clear that we (ever) want to use LLVM to parse this YAML file, the code supports and asserts that we're writing only. On the other hand, I did experiment that the class hierarchy starting at DiagnosticInfoOptimizationBase can be mapped back from YAML generated here (see D24479). * The YAML stream is stored in the LLVM context. * In the example, we can probably further specify the IR value used, i.e. print "Function" rather than "Value". * As before hotness is computed in the analysis pass instead of DiganosticInfo. This avoids the layering problem since BFI is in Analysis while DiagnosticInfo is in IR. [1] https://reviews.llvm.org/D19678#419445 Differential Revision: https://reviews.llvm.org/D24587 llvm-svn: 282539
2016-09-27 22:55:07 +02:00
DiagnosticInfoOptimizationBase &operator<<(StringRef S);
DiagnosticInfoOptimizationBase &operator<<(Argument A);
DiagnosticInfoOptimizationBase &operator<<(setIsVerbose V);
DiagnosticInfoOptimizationBase &operator<<(setExtraArgs EA);
/// \see DiagnosticInfo::print.
void print(DiagnosticPrinter &DP) const override;
/// Return true if this optimization remark is enabled by one of
/// of the LLVM command line flags (-pass-remarks, -pass-remarks-missed,
/// or -pass-remarks-analysis). Note that this only handles the LLVM
/// flags. We cannot access Clang flags from here (they are handled
/// in BackendConsumer::OptimizationRemarkHandler).
virtual bool isEnabled() const = 0;
StringRef getPassName() const { return PassName; }
Output optimization remarks in YAML (Re-committed after moving the template specialization under the yaml namespace. GCC was complaining about this.) This allows various presentation of this data using an external tool. This was first recommended here[1]. As an example, consider this module: 1 int foo(); 2 int bar(); 3 4 int baz() { 5 return foo() + bar(); 6 } The inliner generates these missed-optimization remarks today (the hotness information is pulled from PGO): remark: /tmp/s.c:5:10: foo will not be inlined into baz (hotness: 30) remark: /tmp/s.c:5:18: bar will not be inlined into baz (hotness: 30) Now with -pass-remarks-output=<yaml-file>, we generate this YAML file: --- !Missed Pass: inline Name: NotInlined DebugLoc: { File: /tmp/s.c, Line: 5, Column: 10 } Function: baz Hotness: 30 Args: - Callee: foo - String: will not be inlined into - Caller: baz ... --- !Missed Pass: inline Name: NotInlined DebugLoc: { File: /tmp/s.c, Line: 5, Column: 18 } Function: baz Hotness: 30 Args: - Callee: bar - String: will not be inlined into - Caller: baz ... This is a summary of the high-level decisions: * There is a new streaming interface to emit optimization remarks. E.g. for the inliner remark above: ORE.emit(DiagnosticInfoOptimizationRemarkMissed( DEBUG_TYPE, "NotInlined", &I) << NV("Callee", Callee) << " will not be inlined into " << NV("Caller", CS.getCaller()) << setIsVerbose()); NV stands for named value and allows the YAML client to process a remark using its name (NotInlined) and the named arguments (Callee and Caller) without parsing the text of the message. Subsequent patches will update ORE users to use the new streaming API. * I am using YAML I/O for writing the YAML file. YAML I/O requires you to specify reading and writing at once but reading is highly non-trivial for some of the more complex LLVM types. Since it's not clear that we (ever) want to use LLVM to parse this YAML file, the code supports and asserts that we're writing only. On the other hand, I did experiment that the class hierarchy starting at DiagnosticInfoOptimizationBase can be mapped back from YAML generated here (see D24479). * The YAML stream is stored in the LLVM context. * In the example, we can probably further specify the IR value used, i.e. print "Function" rather than "Value". * As before hotness is computed in the analysis pass instead of DiganosticInfo. This avoids the layering problem since BFI is in Analysis while DiagnosticInfo is in IR. [1] https://reviews.llvm.org/D19678#419445 Differential Revision: https://reviews.llvm.org/D24587 llvm-svn: 282539
2016-09-27 22:55:07 +02:00
std::string getMsg() const;
Optional<uint64_t> getHotness() const { return Hotness; }
Output optimization remarks in YAML (Re-committed after moving the template specialization under the yaml namespace. GCC was complaining about this.) This allows various presentation of this data using an external tool. This was first recommended here[1]. As an example, consider this module: 1 int foo(); 2 int bar(); 3 4 int baz() { 5 return foo() + bar(); 6 } The inliner generates these missed-optimization remarks today (the hotness information is pulled from PGO): remark: /tmp/s.c:5:10: foo will not be inlined into baz (hotness: 30) remark: /tmp/s.c:5:18: bar will not be inlined into baz (hotness: 30) Now with -pass-remarks-output=<yaml-file>, we generate this YAML file: --- !Missed Pass: inline Name: NotInlined DebugLoc: { File: /tmp/s.c, Line: 5, Column: 10 } Function: baz Hotness: 30 Args: - Callee: foo - String: will not be inlined into - Caller: baz ... --- !Missed Pass: inline Name: NotInlined DebugLoc: { File: /tmp/s.c, Line: 5, Column: 18 } Function: baz Hotness: 30 Args: - Callee: bar - String: will not be inlined into - Caller: baz ... This is a summary of the high-level decisions: * There is a new streaming interface to emit optimization remarks. E.g. for the inliner remark above: ORE.emit(DiagnosticInfoOptimizationRemarkMissed( DEBUG_TYPE, "NotInlined", &I) << NV("Callee", Callee) << " will not be inlined into " << NV("Caller", CS.getCaller()) << setIsVerbose()); NV stands for named value and allows the YAML client to process a remark using its name (NotInlined) and the named arguments (Callee and Caller) without parsing the text of the message. Subsequent patches will update ORE users to use the new streaming API. * I am using YAML I/O for writing the YAML file. YAML I/O requires you to specify reading and writing at once but reading is highly non-trivial for some of the more complex LLVM types. Since it's not clear that we (ever) want to use LLVM to parse this YAML file, the code supports and asserts that we're writing only. On the other hand, I did experiment that the class hierarchy starting at DiagnosticInfoOptimizationBase can be mapped back from YAML generated here (see D24479). * The YAML stream is stored in the LLVM context. * In the example, we can probably further specify the IR value used, i.e. print "Function" rather than "Value". * As before hotness is computed in the analysis pass instead of DiganosticInfo. This avoids the layering problem since BFI is in Analysis while DiagnosticInfo is in IR. [1] https://reviews.llvm.org/D19678#419445 Differential Revision: https://reviews.llvm.org/D24587 llvm-svn: 282539
2016-09-27 22:55:07 +02:00
void setHotness(Optional<uint64_t> H) { Hotness = H; }
bool isVerbose() const { return IsVerbose; }
static bool classof(const DiagnosticInfo *DI) {
return (DI->getKind() >= DK_FirstRemark &&
DI->getKind() <= DK_LastRemark) ||
(DI->getKind() >= DK_FirstMachineRemark &&
DI->getKind() <= DK_LastMachineRemark);
}
bool isPassed() const {
return (getKind() == DK_OptimizationRemark ||
getKind() == DK_MachineOptimizationRemark);
}
bool isMissed() const {
return (getKind() == DK_OptimizationRemarkMissed ||
getKind() == DK_MachineOptimizationRemarkMissed);
}
bool isAnalysis() const {
return (getKind() == DK_OptimizationRemarkAnalysis ||
getKind() == DK_MachineOptimizationRemarkAnalysis);
}
protected:
/// Name of the pass that triggers this report. If this matches the
/// regular expression given in -Rpass=regexp, then the remark will
/// be emitted.
const char *PassName;
/// Textual identifier for the remark (single-word, camel-case). Can be used
/// by external tools reading the YAML output file for optimization remarks to
/// identify the remark.
Output optimization remarks in YAML (Re-committed after moving the template specialization under the yaml namespace. GCC was complaining about this.) This allows various presentation of this data using an external tool. This was first recommended here[1]. As an example, consider this module: 1 int foo(); 2 int bar(); 3 4 int baz() { 5 return foo() + bar(); 6 } The inliner generates these missed-optimization remarks today (the hotness information is pulled from PGO): remark: /tmp/s.c:5:10: foo will not be inlined into baz (hotness: 30) remark: /tmp/s.c:5:18: bar will not be inlined into baz (hotness: 30) Now with -pass-remarks-output=<yaml-file>, we generate this YAML file: --- !Missed Pass: inline Name: NotInlined DebugLoc: { File: /tmp/s.c, Line: 5, Column: 10 } Function: baz Hotness: 30 Args: - Callee: foo - String: will not be inlined into - Caller: baz ... --- !Missed Pass: inline Name: NotInlined DebugLoc: { File: /tmp/s.c, Line: 5, Column: 18 } Function: baz Hotness: 30 Args: - Callee: bar - String: will not be inlined into - Caller: baz ... This is a summary of the high-level decisions: * There is a new streaming interface to emit optimization remarks. E.g. for the inliner remark above: ORE.emit(DiagnosticInfoOptimizationRemarkMissed( DEBUG_TYPE, "NotInlined", &I) << NV("Callee", Callee) << " will not be inlined into " << NV("Caller", CS.getCaller()) << setIsVerbose()); NV stands for named value and allows the YAML client to process a remark using its name (NotInlined) and the named arguments (Callee and Caller) without parsing the text of the message. Subsequent patches will update ORE users to use the new streaming API. * I am using YAML I/O for writing the YAML file. YAML I/O requires you to specify reading and writing at once but reading is highly non-trivial for some of the more complex LLVM types. Since it's not clear that we (ever) want to use LLVM to parse this YAML file, the code supports and asserts that we're writing only. On the other hand, I did experiment that the class hierarchy starting at DiagnosticInfoOptimizationBase can be mapped back from YAML generated here (see D24479). * The YAML stream is stored in the LLVM context. * In the example, we can probably further specify the IR value used, i.e. print "Function" rather than "Value". * As before hotness is computed in the analysis pass instead of DiganosticInfo. This avoids the layering problem since BFI is in Analysis while DiagnosticInfo is in IR. [1] https://reviews.llvm.org/D19678#419445 Differential Revision: https://reviews.llvm.org/D24587 llvm-svn: 282539
2016-09-27 22:55:07 +02:00
StringRef RemarkName;
[OptRemark,LDist] RFC: Add hotness attribute Summary: This is the first set of changes implementing the RFC from http://thread.gmane.org/gmane.comp.compilers.llvm.devel/98334 This is a cross-sectional patch; rather than implementing the hotness attribute for all optimization remarks and all passes in a patch set, it implements it for the 'missed-optimization' remark for Loop Distribution. My goal is to shake out the design issues before scaling it up to other types and passes. Hotness is computed as an integer as the multiplication of the block frequency with the function entry count. It's only printed in opt currently since clang prints the diagnostic fields directly. E.g.: remark: /tmp/t.c:3:3: loop not distributed: use -Rpass-analysis=loop-distribute for more info (hotness: 300) A new API added is similar to emitOptimizationRemarkMissed. The difference is that it additionally takes a code region that the diagnostic corresponds to. From this, hotness is computed using BFI. The new API is exposed via an analysis pass so that it can be made dependent on LazyBFI. (Thanks to Hal for the analysis pass idea.) This feature can all be enabled by setDiagnosticHotnessRequested in the LLVM context. If this is off, LazyBFI is not calculated (D22141) so there should be no overhead. A new command-line option is added to turn this on in opt. My plan is to switch all user of emitOptimizationRemark* to use this module instead. Reviewers: hfinkel Subscribers: rcox2, mzolotukhin, llvm-commits Differential Revision: http://reviews.llvm.org/D21771 llvm-svn: 275583
2016-07-15 19:23:20 +02:00
/// If profile information is available, this is the number of times the
/// corresponding code was executed in a profile instrumentation run.
Optional<uint64_t> Hotness;
Output optimization remarks in YAML (Re-committed after moving the template specialization under the yaml namespace. GCC was complaining about this.) This allows various presentation of this data using an external tool. This was first recommended here[1]. As an example, consider this module: 1 int foo(); 2 int bar(); 3 4 int baz() { 5 return foo() + bar(); 6 } The inliner generates these missed-optimization remarks today (the hotness information is pulled from PGO): remark: /tmp/s.c:5:10: foo will not be inlined into baz (hotness: 30) remark: /tmp/s.c:5:18: bar will not be inlined into baz (hotness: 30) Now with -pass-remarks-output=<yaml-file>, we generate this YAML file: --- !Missed Pass: inline Name: NotInlined DebugLoc: { File: /tmp/s.c, Line: 5, Column: 10 } Function: baz Hotness: 30 Args: - Callee: foo - String: will not be inlined into - Caller: baz ... --- !Missed Pass: inline Name: NotInlined DebugLoc: { File: /tmp/s.c, Line: 5, Column: 18 } Function: baz Hotness: 30 Args: - Callee: bar - String: will not be inlined into - Caller: baz ... This is a summary of the high-level decisions: * There is a new streaming interface to emit optimization remarks. E.g. for the inliner remark above: ORE.emit(DiagnosticInfoOptimizationRemarkMissed( DEBUG_TYPE, "NotInlined", &I) << NV("Callee", Callee) << " will not be inlined into " << NV("Caller", CS.getCaller()) << setIsVerbose()); NV stands for named value and allows the YAML client to process a remark using its name (NotInlined) and the named arguments (Callee and Caller) without parsing the text of the message. Subsequent patches will update ORE users to use the new streaming API. * I am using YAML I/O for writing the YAML file. YAML I/O requires you to specify reading and writing at once but reading is highly non-trivial for some of the more complex LLVM types. Since it's not clear that we (ever) want to use LLVM to parse this YAML file, the code supports and asserts that we're writing only. On the other hand, I did experiment that the class hierarchy starting at DiagnosticInfoOptimizationBase can be mapped back from YAML generated here (see D24479). * The YAML stream is stored in the LLVM context. * In the example, we can probably further specify the IR value used, i.e. print "Function" rather than "Value". * As before hotness is computed in the analysis pass instead of DiganosticInfo. This avoids the layering problem since BFI is in Analysis while DiagnosticInfo is in IR. [1] https://reviews.llvm.org/D19678#419445 Differential Revision: https://reviews.llvm.org/D24587 llvm-svn: 282539
2016-09-27 22:55:07 +02:00
/// Arguments collected via the streaming interface.
SmallVector<Argument, 4> Args;
/// The remark is expected to be noisy.
bool IsVerbose = false;
Output optimization remarks in YAML (Re-committed after moving the template specialization under the yaml namespace. GCC was complaining about this.) This allows various presentation of this data using an external tool. This was first recommended here[1]. As an example, consider this module: 1 int foo(); 2 int bar(); 3 4 int baz() { 5 return foo() + bar(); 6 } The inliner generates these missed-optimization remarks today (the hotness information is pulled from PGO): remark: /tmp/s.c:5:10: foo will not be inlined into baz (hotness: 30) remark: /tmp/s.c:5:18: bar will not be inlined into baz (hotness: 30) Now with -pass-remarks-output=<yaml-file>, we generate this YAML file: --- !Missed Pass: inline Name: NotInlined DebugLoc: { File: /tmp/s.c, Line: 5, Column: 10 } Function: baz Hotness: 30 Args: - Callee: foo - String: will not be inlined into - Caller: baz ... --- !Missed Pass: inline Name: NotInlined DebugLoc: { File: /tmp/s.c, Line: 5, Column: 18 } Function: baz Hotness: 30 Args: - Callee: bar - String: will not be inlined into - Caller: baz ... This is a summary of the high-level decisions: * There is a new streaming interface to emit optimization remarks. E.g. for the inliner remark above: ORE.emit(DiagnosticInfoOptimizationRemarkMissed( DEBUG_TYPE, "NotInlined", &I) << NV("Callee", Callee) << " will not be inlined into " << NV("Caller", CS.getCaller()) << setIsVerbose()); NV stands for named value and allows the YAML client to process a remark using its name (NotInlined) and the named arguments (Callee and Caller) without parsing the text of the message. Subsequent patches will update ORE users to use the new streaming API. * I am using YAML I/O for writing the YAML file. YAML I/O requires you to specify reading and writing at once but reading is highly non-trivial for some of the more complex LLVM types. Since it's not clear that we (ever) want to use LLVM to parse this YAML file, the code supports and asserts that we're writing only. On the other hand, I did experiment that the class hierarchy starting at DiagnosticInfoOptimizationBase can be mapped back from YAML generated here (see D24479). * The YAML stream is stored in the LLVM context. * In the example, we can probably further specify the IR value used, i.e. print "Function" rather than "Value". * As before hotness is computed in the analysis pass instead of DiganosticInfo. This avoids the layering problem since BFI is in Analysis while DiagnosticInfo is in IR. [1] https://reviews.llvm.org/D19678#419445 Differential Revision: https://reviews.llvm.org/D24587 llvm-svn: 282539
2016-09-27 22:55:07 +02:00
/// \brief If positive, the index of the first argument that only appear in
/// the optimization records and not in the remark printed in the compiler
/// output.
int FirstExtraArgIndex = -1;
Output optimization remarks in YAML (Re-committed after moving the template specialization under the yaml namespace. GCC was complaining about this.) This allows various presentation of this data using an external tool. This was first recommended here[1]. As an example, consider this module: 1 int foo(); 2 int bar(); 3 4 int baz() { 5 return foo() + bar(); 6 } The inliner generates these missed-optimization remarks today (the hotness information is pulled from PGO): remark: /tmp/s.c:5:10: foo will not be inlined into baz (hotness: 30) remark: /tmp/s.c:5:18: bar will not be inlined into baz (hotness: 30) Now with -pass-remarks-output=<yaml-file>, we generate this YAML file: --- !Missed Pass: inline Name: NotInlined DebugLoc: { File: /tmp/s.c, Line: 5, Column: 10 } Function: baz Hotness: 30 Args: - Callee: foo - String: will not be inlined into - Caller: baz ... --- !Missed Pass: inline Name: NotInlined DebugLoc: { File: /tmp/s.c, Line: 5, Column: 18 } Function: baz Hotness: 30 Args: - Callee: bar - String: will not be inlined into - Caller: baz ... This is a summary of the high-level decisions: * There is a new streaming interface to emit optimization remarks. E.g. for the inliner remark above: ORE.emit(DiagnosticInfoOptimizationRemarkMissed( DEBUG_TYPE, "NotInlined", &I) << NV("Callee", Callee) << " will not be inlined into " << NV("Caller", CS.getCaller()) << setIsVerbose()); NV stands for named value and allows the YAML client to process a remark using its name (NotInlined) and the named arguments (Callee and Caller) without parsing the text of the message. Subsequent patches will update ORE users to use the new streaming API. * I am using YAML I/O for writing the YAML file. YAML I/O requires you to specify reading and writing at once but reading is highly non-trivial for some of the more complex LLVM types. Since it's not clear that we (ever) want to use LLVM to parse this YAML file, the code supports and asserts that we're writing only. On the other hand, I did experiment that the class hierarchy starting at DiagnosticInfoOptimizationBase can be mapped back from YAML generated here (see D24479). * The YAML stream is stored in the LLVM context. * In the example, we can probably further specify the IR value used, i.e. print "Function" rather than "Value". * As before hotness is computed in the analysis pass instead of DiganosticInfo. This avoids the layering problem since BFI is in Analysis while DiagnosticInfo is in IR. [1] https://reviews.llvm.org/D19678#419445 Differential Revision: https://reviews.llvm.org/D24587 llvm-svn: 282539
2016-09-27 22:55:07 +02:00
friend struct yaml::MappingTraits<DiagnosticInfoOptimizationBase *>;
};
/// \brief Common features for diagnostics dealing with optimization remarks
/// that are used by IR passes.
class DiagnosticInfoIROptimization : public DiagnosticInfoOptimizationBase {
public:
/// \p PassName is the name of the pass emitting this diagnostic. \p
/// RemarkName is a textual identifier for the remark (single-word,
/// camel-case). \p Fn is the function where the diagnostic is being emitted.
/// \p Loc is the location information to use in the diagnostic. If line table
/// information is available, the diagnostic will include the source code
/// location. \p CodeRegion is IR value (currently basic block) that the
/// optimization operates on. This is currently used to provide run-time
/// hotness information with PGO.
DiagnosticInfoIROptimization(enum DiagnosticKind Kind,
enum DiagnosticSeverity Severity,
const char *PassName, StringRef RemarkName,
const Function &Fn,
const DiagnosticLocation &Loc,
const Value *CodeRegion = nullptr)
: DiagnosticInfoOptimizationBase(Kind, Severity, PassName, RemarkName, Fn,
Loc),
CodeRegion(CodeRegion) {}
/// \brief This is ctor variant allows a pass to build an optimization remark
/// from an existing remark.
///
/// This is useful when a transformation pass (e.g LV) wants to emit a remark
/// (\p Orig) generated by one of its analyses (e.g. LAA) as its own analysis
/// remark. The string \p Prepend will be emitted before the original
/// message.
DiagnosticInfoIROptimization(const char *PassName, StringRef Prepend,
const DiagnosticInfoIROptimization &Orig)
: DiagnosticInfoOptimizationBase(
(DiagnosticKind)Orig.getKind(), Orig.getSeverity(), PassName,
Orig.RemarkName, Orig.getFunction(), Orig.getLocation()),
CodeRegion(Orig.getCodeRegion()) {
*this << Prepend;
std::copy(Orig.Args.begin(), Orig.Args.end(), std::back_inserter(Args));
}
/// Legacy interface.
/// \p PassName is the name of the pass emitting this diagnostic.
/// \p Fn is the function where the diagnostic is being emitted. \p Loc is
/// the location information to use in the diagnostic. If line table
/// information is available, the diagnostic will include the source code
/// location. \p Msg is the message to show. Note that this class does not
/// copy this message, so this reference must be valid for the whole life time
/// of the diagnostic.
DiagnosticInfoIROptimization(enum DiagnosticKind Kind,
enum DiagnosticSeverity Severity,
const char *PassName, const Function &Fn,
const DiagnosticLocation &Loc, const Twine &Msg)
: DiagnosticInfoOptimizationBase(Kind, Severity, PassName, "", Fn, Loc) {
*this << Msg.str();
}
const Value *getCodeRegion() const { return CodeRegion; }
static bool classof(const DiagnosticInfo *DI) {
return DI->getKind() >= DK_FirstRemark && DI->getKind() <= DK_LastRemark;
}
private:
/// The IR value (currently basic block) that the optimization operates on.
/// This is currently used to provide run-time hotness information with PGO.
const Value *CodeRegion;
};
/// Diagnostic information for applied optimization remarks.
class OptimizationRemark : public DiagnosticInfoIROptimization {
public:
/// \p PassName is the name of the pass emitting this diagnostic. If this name
/// matches the regular expression given in -Rpass=, then the diagnostic will
/// be emitted. \p RemarkName is a textual identifier for the remark (single-
/// word, camel-case). \p Loc is the debug location and \p CodeRegion is the
/// region that the optimization operates on (currently only block is
/// supported).
OptimizationRemark(const char *PassName, StringRef RemarkName,
const DiagnosticLocation &Loc, const Value *CodeRegion);
/// Same as above, but the debug location and code region are derived from \p
/// Instr.
OptimizationRemark(const char *PassName, StringRef RemarkName,
const Instruction *Inst);
/// Same as above, but the debug location and code region are derived from \p
/// Func.
OptimizationRemark(const char *PassName, StringRef RemarkName,
const Function *Func);
static bool classof(const DiagnosticInfo *DI) {
return DI->getKind() == DK_OptimizationRemark;
}
static bool isEnabled(StringRef PassName);
/// \see DiagnosticInfoOptimizationBase::isEnabled.
bool isEnabled() const override { return isEnabled(getPassName()); }
private:
/// This is deprecated now and only used by the function API below.
/// \p PassName is the name of the pass emitting this diagnostic. If
/// this name matches the regular expression given in -Rpass=, then the
/// diagnostic will be emitted. \p Fn is the function where the diagnostic
/// is being emitted. \p Loc is the location information to use in the
/// diagnostic. If line table information is available, the diagnostic
/// will include the source code location. \p Msg is the message to show.
/// Note that this class does not copy this message, so this reference
/// must be valid for the whole life time of the diagnostic.
OptimizationRemark(const char *PassName, const Function &Fn,
const DiagnosticLocation &Loc, const Twine &Msg)
: DiagnosticInfoIROptimization(DK_OptimizationRemark, DS_Remark, PassName,
Fn, Loc, Msg) {}
friend void emitOptimizationRemark(LLVMContext &Ctx, const char *PassName,
const Function &Fn,
const DiagnosticLocation &Loc,
const Twine &Msg);
};
/// Diagnostic information for missed-optimization remarks.
class OptimizationRemarkMissed : public DiagnosticInfoIROptimization {
public:
Output optimization remarks in YAML (Re-committed after moving the template specialization under the yaml namespace. GCC was complaining about this.) This allows various presentation of this data using an external tool. This was first recommended here[1]. As an example, consider this module: 1 int foo(); 2 int bar(); 3 4 int baz() { 5 return foo() + bar(); 6 } The inliner generates these missed-optimization remarks today (the hotness information is pulled from PGO): remark: /tmp/s.c:5:10: foo will not be inlined into baz (hotness: 30) remark: /tmp/s.c:5:18: bar will not be inlined into baz (hotness: 30) Now with -pass-remarks-output=<yaml-file>, we generate this YAML file: --- !Missed Pass: inline Name: NotInlined DebugLoc: { File: /tmp/s.c, Line: 5, Column: 10 } Function: baz Hotness: 30 Args: - Callee: foo - String: will not be inlined into - Caller: baz ... --- !Missed Pass: inline Name: NotInlined DebugLoc: { File: /tmp/s.c, Line: 5, Column: 18 } Function: baz Hotness: 30 Args: - Callee: bar - String: will not be inlined into - Caller: baz ... This is a summary of the high-level decisions: * There is a new streaming interface to emit optimization remarks. E.g. for the inliner remark above: ORE.emit(DiagnosticInfoOptimizationRemarkMissed( DEBUG_TYPE, "NotInlined", &I) << NV("Callee", Callee) << " will not be inlined into " << NV("Caller", CS.getCaller()) << setIsVerbose()); NV stands for named value and allows the YAML client to process a remark using its name (NotInlined) and the named arguments (Callee and Caller) without parsing the text of the message. Subsequent patches will update ORE users to use the new streaming API. * I am using YAML I/O for writing the YAML file. YAML I/O requires you to specify reading and writing at once but reading is highly non-trivial for some of the more complex LLVM types. Since it's not clear that we (ever) want to use LLVM to parse this YAML file, the code supports and asserts that we're writing only. On the other hand, I did experiment that the class hierarchy starting at DiagnosticInfoOptimizationBase can be mapped back from YAML generated here (see D24479). * The YAML stream is stored in the LLVM context. * In the example, we can probably further specify the IR value used, i.e. print "Function" rather than "Value". * As before hotness is computed in the analysis pass instead of DiganosticInfo. This avoids the layering problem since BFI is in Analysis while DiagnosticInfo is in IR. [1] https://reviews.llvm.org/D19678#419445 Differential Revision: https://reviews.llvm.org/D24587 llvm-svn: 282539
2016-09-27 22:55:07 +02:00
/// \p PassName is the name of the pass emitting this diagnostic. If this name
/// matches the regular expression given in -Rpass-missed=, then the
/// diagnostic will be emitted. \p RemarkName is a textual identifier for the
/// remark (single-word, camel-case). \p Loc is the debug location and \p
/// CodeRegion is the region that the optimization operates on (currently only
/// block is supported).
OptimizationRemarkMissed(const char *PassName, StringRef RemarkName,
const DiagnosticLocation &Loc,
const Value *CodeRegion);
/// \brief Same as above but \p Inst is used to derive code region and debug
/// location.
OptimizationRemarkMissed(const char *PassName, StringRef RemarkName,
const Instruction *Inst);
Output optimization remarks in YAML (Re-committed after moving the template specialization under the yaml namespace. GCC was complaining about this.) This allows various presentation of this data using an external tool. This was first recommended here[1]. As an example, consider this module: 1 int foo(); 2 int bar(); 3 4 int baz() { 5 return foo() + bar(); 6 } The inliner generates these missed-optimization remarks today (the hotness information is pulled from PGO): remark: /tmp/s.c:5:10: foo will not be inlined into baz (hotness: 30) remark: /tmp/s.c:5:18: bar will not be inlined into baz (hotness: 30) Now with -pass-remarks-output=<yaml-file>, we generate this YAML file: --- !Missed Pass: inline Name: NotInlined DebugLoc: { File: /tmp/s.c, Line: 5, Column: 10 } Function: baz Hotness: 30 Args: - Callee: foo - String: will not be inlined into - Caller: baz ... --- !Missed Pass: inline Name: NotInlined DebugLoc: { File: /tmp/s.c, Line: 5, Column: 18 } Function: baz Hotness: 30 Args: - Callee: bar - String: will not be inlined into - Caller: baz ... This is a summary of the high-level decisions: * There is a new streaming interface to emit optimization remarks. E.g. for the inliner remark above: ORE.emit(DiagnosticInfoOptimizationRemarkMissed( DEBUG_TYPE, "NotInlined", &I) << NV("Callee", Callee) << " will not be inlined into " << NV("Caller", CS.getCaller()) << setIsVerbose()); NV stands for named value and allows the YAML client to process a remark using its name (NotInlined) and the named arguments (Callee and Caller) without parsing the text of the message. Subsequent patches will update ORE users to use the new streaming API. * I am using YAML I/O for writing the YAML file. YAML I/O requires you to specify reading and writing at once but reading is highly non-trivial for some of the more complex LLVM types. Since it's not clear that we (ever) want to use LLVM to parse this YAML file, the code supports and asserts that we're writing only. On the other hand, I did experiment that the class hierarchy starting at DiagnosticInfoOptimizationBase can be mapped back from YAML generated here (see D24479). * The YAML stream is stored in the LLVM context. * In the example, we can probably further specify the IR value used, i.e. print "Function" rather than "Value". * As before hotness is computed in the analysis pass instead of DiganosticInfo. This avoids the layering problem since BFI is in Analysis while DiagnosticInfo is in IR. [1] https://reviews.llvm.org/D19678#419445 Differential Revision: https://reviews.llvm.org/D24587 llvm-svn: 282539
2016-09-27 22:55:07 +02:00
static bool classof(const DiagnosticInfo *DI) {
return DI->getKind() == DK_OptimizationRemarkMissed;
}
static bool isEnabled(StringRef PassName);
/// \see DiagnosticInfoOptimizationBase::isEnabled.
bool isEnabled() const override { return isEnabled(getPassName()); }
private:
/// This is deprecated now and only used by the function API below.
/// \p PassName is the name of the pass emitting this diagnostic. If
/// this name matches the regular expression given in -Rpass-missed=, then the
/// diagnostic will be emitted. \p Fn is the function where the diagnostic
/// is being emitted. \p Loc is the location information to use in the
/// diagnostic. If line table information is available, the diagnostic
/// will include the source code location. \p Msg is the message to show.
/// Note that this class does not copy this message, so this reference
/// must be valid for the whole life time of the diagnostic.
OptimizationRemarkMissed(const char *PassName, const Function &Fn,
const DiagnosticLocation &Loc, const Twine &Msg)
: DiagnosticInfoIROptimization(DK_OptimizationRemarkMissed, DS_Remark,
PassName, Fn, Loc, Msg) {}
friend void emitOptimizationRemarkMissed(LLVMContext &Ctx,
const char *PassName,
const Function &Fn,
const DiagnosticLocation &Loc,
const Twine &Msg);
};
/// Diagnostic information for optimization analysis remarks.
class OptimizationRemarkAnalysis : public DiagnosticInfoIROptimization {
public:
/// \p PassName is the name of the pass emitting this diagnostic. If this name
/// matches the regular expression given in -Rpass-analysis=, then the
/// diagnostic will be emitted. \p RemarkName is a textual identifier for the
/// remark (single-word, camel-case). \p Loc is the debug location and \p
/// CodeRegion is the region that the optimization operates on (currently only
/// block is supported).
OptimizationRemarkAnalysis(const char *PassName, StringRef RemarkName,
const DiagnosticLocation &Loc,
const Value *CodeRegion);
/// \brief This is ctor variant allows a pass to build an optimization remark
/// from an existing remark.
///
/// This is useful when a transformation pass (e.g LV) wants to emit a remark
/// (\p Orig) generated by one of its analyses (e.g. LAA) as its own analysis
/// remark. The string \p Prepend will be emitted before the original
/// message.
OptimizationRemarkAnalysis(const char *PassName, StringRef Prepend,
const OptimizationRemarkAnalysis &Orig)
: DiagnosticInfoIROptimization(PassName, Prepend, Orig) {}
/// \brief Same as above but \p Inst is used to derive code region and debug
/// location.
OptimizationRemarkAnalysis(const char *PassName, StringRef RemarkName,
const Instruction *Inst);
static bool classof(const DiagnosticInfo *DI) {
return DI->getKind() == DK_OptimizationRemarkAnalysis;
}
static bool isEnabled(StringRef PassName);
/// \see DiagnosticInfoOptimizationBase::isEnabled.
bool isEnabled() const override {
return shouldAlwaysPrint() || isEnabled(getPassName());
}
static const char *AlwaysPrint;
bool shouldAlwaysPrint() const { return getPassName() == AlwaysPrint; }
protected:
OptimizationRemarkAnalysis(enum DiagnosticKind Kind, const char *PassName,
const Function &Fn, const DiagnosticLocation &Loc,
const Twine &Msg)
: DiagnosticInfoIROptimization(Kind, DS_Remark, PassName, Fn, Loc, Msg) {}
OptimizationRemarkAnalysis(enum DiagnosticKind Kind, const char *PassName,
StringRef RemarkName,
const DiagnosticLocation &Loc,
const Value *CodeRegion);
private:
/// This is deprecated now and only used by the function API below.
/// \p PassName is the name of the pass emitting this diagnostic. If
/// this name matches the regular expression given in -Rpass-analysis=, then
/// the diagnostic will be emitted. \p Fn is the function where the diagnostic
/// is being emitted. \p Loc is the location information to use in the
/// diagnostic. If line table information is available, the diagnostic will
/// include the source code location. \p Msg is the message to show. Note that
/// this class does not copy this message, so this reference must be valid for
/// the whole life time of the diagnostic.
OptimizationRemarkAnalysis(const char *PassName, const Function &Fn,
const DiagnosticLocation &Loc, const Twine &Msg)
: DiagnosticInfoIROptimization(DK_OptimizationRemarkAnalysis, DS_Remark,
PassName, Fn, Loc, Msg) {}
friend void emitOptimizationRemarkAnalysis(LLVMContext &Ctx,
const char *PassName,
const Function &Fn,
const DiagnosticLocation &Loc,
const Twine &Msg);
};
/// Diagnostic information for optimization analysis remarks related to
/// floating-point non-commutativity.
class OptimizationRemarkAnalysisFPCommute : public OptimizationRemarkAnalysis {
public:
/// \p PassName is the name of the pass emitting this diagnostic. If this name
/// matches the regular expression given in -Rpass-analysis=, then the
/// diagnostic will be emitted. \p RemarkName is a textual identifier for the
/// remark (single-word, camel-case). \p Loc is the debug location and \p
/// CodeRegion is the region that the optimization operates on (currently only
/// block is supported). The front-end will append its own message related to
/// options that address floating-point non-commutativity.
OptimizationRemarkAnalysisFPCommute(const char *PassName,
StringRef RemarkName,
const DiagnosticLocation &Loc,
const Value *CodeRegion)
: OptimizationRemarkAnalysis(DK_OptimizationRemarkAnalysisFPCommute,
PassName, RemarkName, Loc, CodeRegion) {}
static bool classof(const DiagnosticInfo *DI) {
return DI->getKind() == DK_OptimizationRemarkAnalysisFPCommute;
}
private:
/// This is deprecated now and only used by the function API below.
/// \p PassName is the name of the pass emitting this diagnostic. If
/// this name matches the regular expression given in -Rpass-analysis=, then
/// the diagnostic will be emitted. \p Fn is the function where the diagnostic
/// is being emitted. \p Loc is the location information to use in the
/// diagnostic. If line table information is available, the diagnostic will
/// include the source code location. \p Msg is the message to show. The
/// front-end will append its own message related to options that address
/// floating-point non-commutativity. Note that this class does not copy this
/// message, so this reference must be valid for the whole life time of the
/// diagnostic.
OptimizationRemarkAnalysisFPCommute(const char *PassName, const Function &Fn,
const DiagnosticLocation &Loc,
const Twine &Msg)
: OptimizationRemarkAnalysis(DK_OptimizationRemarkAnalysisFPCommute,
PassName, Fn, Loc, Msg) {}
friend void emitOptimizationRemarkAnalysisFPCommute(
LLVMContext &Ctx, const char *PassName, const Function &Fn,
const DiagnosticLocation &Loc, const Twine &Msg);
};
/// Diagnostic information for optimization analysis remarks related to
/// pointer aliasing.
class OptimizationRemarkAnalysisAliasing : public OptimizationRemarkAnalysis {
public:
/// \p PassName is the name of the pass emitting this diagnostic. If this name
/// matches the regular expression given in -Rpass-analysis=, then the
/// diagnostic will be emitted. \p RemarkName is a textual identifier for the
/// remark (single-word, camel-case). \p Loc is the debug location and \p
/// CodeRegion is the region that the optimization operates on (currently only
/// block is supported). The front-end will append its own message related to
/// options that address pointer aliasing legality.
OptimizationRemarkAnalysisAliasing(const char *PassName, StringRef RemarkName,
const DiagnosticLocation &Loc,
const Value *CodeRegion)
: OptimizationRemarkAnalysis(DK_OptimizationRemarkAnalysisAliasing,
PassName, RemarkName, Loc, CodeRegion) {}
static bool classof(const DiagnosticInfo *DI) {
return DI->getKind() == DK_OptimizationRemarkAnalysisAliasing;
}
private:
/// This is deprecated now and only used by the function API below.
/// \p PassName is the name of the pass emitting this diagnostic. If
/// this name matches the regular expression given in -Rpass-analysis=, then
/// the diagnostic will be emitted. \p Fn is the function where the diagnostic
/// is being emitted. \p Loc is the location information to use in the
/// diagnostic. If line table information is available, the diagnostic will
/// include the source code location. \p Msg is the message to show. The
/// front-end will append its own message related to options that address
/// pointer aliasing legality. Note that this class does not copy this
/// message, so this reference must be valid for the whole life time of the
/// diagnostic.
OptimizationRemarkAnalysisAliasing(const char *PassName, const Function &Fn,
const DiagnosticLocation &Loc,
const Twine &Msg)
: OptimizationRemarkAnalysis(DK_OptimizationRemarkAnalysisAliasing,
PassName, Fn, Loc, Msg) {}
friend void emitOptimizationRemarkAnalysisAliasing(
LLVMContext &Ctx, const char *PassName, const Function &Fn,
const DiagnosticLocation &Loc, const Twine &Msg);
};
/// Diagnostic information for machine IR parser.
class DiagnosticInfoMIRParser : public DiagnosticInfo {
const SMDiagnostic &Diagnostic;
public:
DiagnosticInfoMIRParser(DiagnosticSeverity Severity,
const SMDiagnostic &Diagnostic)
: DiagnosticInfo(DK_MIRParser, Severity), Diagnostic(Diagnostic) {}
const SMDiagnostic &getDiagnostic() const { return Diagnostic; }
void print(DiagnosticPrinter &DP) const override;
static bool classof(const DiagnosticInfo *DI) {
return DI->getKind() == DK_MIRParser;
}
};
/// Diagnostic information for ISel fallback path.
class DiagnosticInfoISelFallback : public DiagnosticInfo {
/// The function that is concerned by this diagnostic.
const Function &Fn;
public:
DiagnosticInfoISelFallback(const Function &Fn,
DiagnosticSeverity Severity = DS_Warning)
: DiagnosticInfo(DK_ISelFallback, Severity), Fn(Fn) {}
const Function &getFunction() const { return Fn; }
void print(DiagnosticPrinter &DP) const override;
static bool classof(const DiagnosticInfo *DI) {
return DI->getKind() == DK_ISelFallback;
}
};
// Create wrappers for C Binding types (see CBindingWrapping.h).
DEFINE_SIMPLE_CONVERSION_FUNCTIONS(DiagnosticInfo, LLVMDiagnosticInfoRef)
/// \brief Legacy interface to emit an optimization-applied message. Use
/// (Machine)OptimizationRemarkEmitter instead.
///
/// \p PassName is the name of the pass emitting the message. If -Rpass= is
/// given and \p PassName matches the regular expression in -Rpass, then the
/// remark will be emitted. \p Fn is the function triggering the remark, \p Loc
/// is the debug location where the diagnostic is generated. \p Msg is the
/// message string to use.
void emitOptimizationRemark(LLVMContext &Ctx, const char *PassName,
const Function &Fn, const DiagnosticLocation &Loc,
const Twine &Msg);
/// \brief Legacy interface to emit an optimization-missed message. Use
/// (Machine)OptimizationRemarkEmitter instead.
///
/// \p PassName is the name of the pass emitting the message. If -Rpass-missed=
/// is given and \p PassName matches the regular expression in -Rpass, then the
/// remark will be emitted. \p Fn is the function triggering the remark, \p Loc
/// is the debug location where the diagnostic is generated. \p Msg is the
/// message string to use.
void emitOptimizationRemarkMissed(LLVMContext &Ctx, const char *PassName,
const Function &Fn,
const DiagnosticLocation &Loc,
const Twine &Msg);
/// \brief Legacy interface to emit an optimization analysis remark message.
/// Use (Machine)OptimizationRemarkEmitter instead.
///
/// \p PassName is the name of the pass emitting the message. If
/// -Rpass-analysis= is given and \p PassName matches the regular expression in
/// -Rpass, then the remark will be emitted. \p Fn is the function triggering
/// the remark, \p Loc is the debug location where the diagnostic is
/// generated. \p Msg is the message string to use.
void emitOptimizationRemarkAnalysis(LLVMContext &Ctx, const char *PassName,
const Function &Fn,
const DiagnosticLocation &Loc,
const Twine &Msg);
/// \brief Legacy interface to emit an optimization analysis remark related to
/// messages about floating-point non-commutativity. Use
/// (Machine)OptimizationRemarkEmitter instead.
///
/// \p PassName is the name of the pass emitting the message. If
/// -Rpass-analysis= is given and \p PassName matches the regular expression in
/// -Rpass, then the remark will be emitted. \p Fn is the function triggering
/// the remark, \p Loc is the debug location where the diagnostic is
/// generated. \p Msg is the message string to use.
void emitOptimizationRemarkAnalysisFPCommute(LLVMContext &Ctx,
const char *PassName,
const Function &Fn,
const DiagnosticLocation &Loc,
const Twine &Msg);
/// \brief Legacy interface to emit an optimization analysis remark related to
/// messages about pointer aliasing. Use (Machine)OptimizationRemarkEmitter
/// instead.
///
/// \p PassName is the name of the pass emitting the message.
/// If -Rpass-analysis= is given and \p PassName matches the regular expression
/// in -Rpass, then the remark will be emitted. \p Fn is the function triggering
/// the remark, \p Loc is the debug location where the diagnostic is generated.
/// \p Msg is the message string to use.
void emitOptimizationRemarkAnalysisAliasing(LLVMContext &Ctx,
const char *PassName,
const Function &Fn,
const DiagnosticLocation &Loc,
const Twine &Msg);
/// Diagnostic information for optimization failures.
class DiagnosticInfoOptimizationFailure : public DiagnosticInfoIROptimization {
public:
/// \p Fn is the function where the diagnostic is being emitted. \p Loc is
/// the location information to use in the diagnostic. If line table
/// information is available, the diagnostic will include the source code
/// location. \p Msg is the message to show. Note that this class does not
/// copy this message, so this reference must be valid for the whole life time
/// of the diagnostic.
DiagnosticInfoOptimizationFailure(const Function &Fn,
const DiagnosticLocation &Loc,
const Twine &Msg)
: DiagnosticInfoIROptimization(DK_OptimizationFailure, DS_Warning,
nullptr, Fn, Loc, Msg) {}
/// \p PassName is the name of the pass emitting this diagnostic. \p
/// RemarkName is a textual identifier for the remark (single-word,
/// camel-case). \p Loc is the debug location and \p CodeRegion is the
/// region that the optimization operates on (currently basic block is
/// supported).
DiagnosticInfoOptimizationFailure(const char *PassName, StringRef RemarkName,
const DiagnosticLocation &Loc,
const Value *CodeRegion);
static bool classof(const DiagnosticInfo *DI) {
return DI->getKind() == DK_OptimizationFailure;
}
/// \see DiagnosticInfoOptimizationBase::isEnabled.
bool isEnabled() const override;
};
/// Diagnostic information for unsupported feature in backend.
class DiagnosticInfoUnsupported : public DiagnosticInfoWithLocationBase {
private:
Twine Msg;
public:
/// \p Fn is the function where the diagnostic is being emitted. \p Loc is
/// the location information to use in the diagnostic. If line table
/// information is available, the diagnostic will include the source code
/// location. \p Msg is the message to show. Note that this class does not
/// copy this message, so this reference must be valid for the whole life time
/// of the diagnostic.
DiagnosticInfoUnsupported(
const Function &Fn, const Twine &Msg,
const DiagnosticLocation &Loc = DiagnosticLocation(),
DiagnosticSeverity Severity = DS_Error)
: DiagnosticInfoWithLocationBase(DK_Unsupported, Severity, Fn, Loc),
Msg(Msg) {}
static bool classof(const DiagnosticInfo *DI) {
return DI->getKind() == DK_Unsupported;
}
const Twine &getMessage() const { return Msg; }
void print(DiagnosticPrinter &DP) const override;
};
} // end namespace llvm
#endif // LLVM_IR_DIAGNOSTICINFO_H