xur created this revision.
xur added a reviewer: tejohnson.
Herald added subscribers: dexonsmith, steven_wu, hiraditya, mehdi_amini.
For IR input files, we currently use LLVM diagnostic handler even the
compilation is from clang. As a result, we are not able to use -Rpass
to get the transformation reports. Some warnings are not handled
properly either: We found many mysterious warnings in our ThinLTO backend
compilations in SamplePGO and CSPGO. An example of the warning:
warning: net/proto2/public/metadata_lite.h:51:21: 0.02% (1 / 4999)
This turns out to be a warning by Wmisexpect, which is supposed to be
filtered out by default. But since the filter is in clang's
diagnostic hander, we emit these incomplete warnings from LLVM's
diagnostic handler.
This patch uses clang diagnostic handler for IR input files. We create
a fake backendconsumer just to install the diagnostic handler.
The location information is tightly coupled with the sourcemanager which
is not available in IR input files. I just prepend the location from DebugLoc
to the warning/remark messages. I think this is much simpler way than
to pass DebugLoc to DiagnosticEngine (I tried this approach, but it seemed to
require much more extensive changes).
With this change, we will have proper handling of all the warnings and we can
use -Rpass* options in IR input files compilation.
Also note that with is patch, LLVM's diagnostic options, like
"-mllvm -pass-remarks=*", are no longer be able to get optimization remarks.
https://reviews.llvm.org/D72523
Files:
clang/lib/CodeGen/CodeGenAction.cpp
clang/test/CodeGen/Inputs/thinlto_expect1.proftext
clang/test/CodeGen/Inputs/thinlto_expect2.proftext
clang/test/CodeGen/thinlto-clang-diagnostic-handler-in-be.c
clang/test/CodeGen/thinlto-diagnostic-handler-remarks-with-hotness.ll
Index: clang/test/CodeGen/thinlto-diagnostic-handler-remarks-with-hotness.ll
===================================================================
--- clang/test/CodeGen/thinlto-diagnostic-handler-remarks-with-hotness.ll
+++ clang/test/CodeGen/thinlto-diagnostic-handler-remarks-with-hotness.ll
@@ -28,7 +28,7 @@
; YAML-NEXT: ...
; Next try with pass remarks to stderr
-; RUN: %clang -target x86_64-scei-ps4 -O2 -x ir %t.o -fthinlto-index=%t.thinlto.bc -mllvm -pass-remarks=inline -fdiagnostics-show-hotness -o %t2.o -c 2>&1 | FileCheck %s
+; RUN: %clang -target x86_64-scei-ps4 -O2 -x ir %t.o -fthinlto-index=%t.thinlto.bc -Rpass=inline -fdiagnostics-show-hotness -o %t2.o -c 2>&1 | FileCheck %s
; CHECK: tinkywinky inlined into main with (cost=0, threshold=337) (hotness: 300)
Index: clang/test/CodeGen/thinlto-clang-diagnostic-handler-in-be.c
===================================================================
--- /dev/null
+++ clang/test/CodeGen/thinlto-clang-diagnostic-handler-in-be.c
@@ -0,0 +1,22 @@
+// Test clang diagnositic handler works in IR file compilaiton.
+
+// REQUIRES: x86-registered-target
+
+// RUN: llvm-profdata merge -o %t1.profdata %S/Inputs/thinlto_expect1.proftext
+// RUN: %clang -O2 -fexperimental-new-pass-manager -flto=thin -g -fprofile-use=%t1.profdata -c -o %t1.bo %s
+// RUN: llvm-lto -thinlto -o %t %t1.bo
+// RUN: %clang -cc1 -O2 -fexperimental-new-pass-manager -x ir %t1.bo -fthinlto-index=%t.thinlto.bc -emit-obj -Rpass-analysis=info 2>&1 | FileCheck %s -check-prefix=CHECK-REMARK
+// RUN: llvm-profdata merge -o %t2.profdata %S/Inputs/thinlto_expect2.proftext
+// RUN: %clang -cc1 -O2 -fexperimental-new-pass-manager -x ir %t1.bo -fthinlto-index=%t.thinlto.bc -fprofile-instrument-use-path=%t2.profdata -emit-obj -Wmisexpect 2>&1 | FileCheck %s -check-prefix=CHECK-WARNING
+
+int sum;
+__attribute__((noinline)) void bar() {
+ sum = 1234;
+}
+
+__attribute__((noinline)) void foo(int m) {
+ if (__builtin_expect(m > 9, 1))
+ bar();
+}
+// CHECK-REMARK: remark: In {{.*}}.c:
+// CHECK-WARNING: warning: Potential performance regression from use of __builtin_expect(): Annotation was correct on {{.*}}.c:{{[0-9]*}}:26: 50.00% (12 / 24) of profiled executions.
Index: clang/test/CodeGen/Inputs/thinlto_expect2.proftext
===================================================================
--- /dev/null
+++ clang/test/CodeGen/Inputs/thinlto_expect2.proftext
@@ -0,0 +1,20 @@
+# CSIR level Instrumentation Flag
+:csir
+foo
+# Func Hash:
+25571299074
+# Num Counters:
+2
+# Counter Values:
+12
+24
+
+foo
+# Func Hash:
+1152921530178146050
+# Num Counters:
+2
+# Counter Values:
+24
+12
+
Index: clang/test/CodeGen/Inputs/thinlto_expect1.proftext
===================================================================
--- /dev/null
+++ clang/test/CodeGen/Inputs/thinlto_expect1.proftext
@@ -0,0 +1,11 @@
+# IR level Instrumentation Flag
+:ir
+foo
+# Func Hash:
+25571299074
+# Num Counters:
+2
+# Counter Values:
+12
+24
+
Index: clang/lib/CodeGen/CodeGenAction.cpp
===================================================================
--- clang/lib/CodeGen/CodeGenAction.cpp
+++ clang/lib/CodeGen/CodeGenAction.cpp
@@ -151,6 +151,25 @@
FrontendTimesIsEnabled = TimePasses;
llvm::TimePassesIsEnabled = TimePasses;
}
+ BackendConsumer(BackendAction Action, DiagnosticsEngine &Diags,
+ const HeaderSearchOptions &HeaderSearchOpts,
+ const PreprocessorOptions &PPOpts,
+ const CodeGenOptions &CodeGenOpts,
+ const TargetOptions &TargetOpts,
+ const LangOptions &LangOpts, bool TimePasses,
+ SmallVector<LinkModule, 4> LinkModules, LLVMContext &C,
+ CoverageSourceInfo *CoverageInfo = nullptr)
+ : Diags(Diags), Action(Action), HeaderSearchOpts(HeaderSearchOpts),
+ CodeGenOpts(CodeGenOpts), TargetOpts(TargetOpts), LangOpts(LangOpts),
+ Context(nullptr),
+ LLVMIRGeneration("irgen", "LLVM IR Generation Time"),
+ LLVMIRGenerationRefCount(0),
+ Gen(CreateLLVMCodeGen(Diags, "", HeaderSearchOpts, PPOpts,
+ CodeGenOpts, C, CoverageInfo)),
+ LinkModules(std::move(LinkModules)) {
+ FrontendTimesIsEnabled = TimePasses;
+ llvm::TimePassesIsEnabled = TimePasses;
+ }
llvm::Module *getModule() const { return Gen->GetModule(); }
std::unique_ptr<llvm::Module> takeModule() {
return std::unique_ptr<llvm::Module>(Gen->ReleaseModule());
@@ -570,6 +589,12 @@
const FullSourceLoc BackendConsumer::getBestLocationFromDebugLoc(
const llvm::DiagnosticInfoWithLocationBase &D, bool &BadDebugInfo,
StringRef &Filename, unsigned &Line, unsigned &Column) const {
+ // It's possible to have a null context in which case return an empty
+ // SourceLoc.
+ if (Context == nullptr) {
+ return FullSourceLoc();
+ }
+
SourceManager &SourceMgr = Context->getSourceManager();
FileManager &FileMgr = SourceMgr.getFileManager();
SourceLocation DILoc;
@@ -619,7 +644,16 @@
FullSourceLoc Loc =
getBestLocationFromDebugLoc(D, BadDebugInfo, Filename, Line, Column);
- Diags.Report(Loc, diag::err_fe_backend_unsupported) << D.getMessage().str();
+ std::string Msg;
+ raw_string_ostream MsgStream(Msg);
+ if (!Loc.isValid()) {
+ MsgStream << D.getLocationStr() << ": in function "
+ << D.getFunction().getName() << ' '
+ << *(D.getFunction().getFunctionType()) << ": ";
+ }
+ MsgStream << D.getMessage();
+
+ Diags.Report(Loc, diag::err_fe_backend_unsupported) << MsgStream.str();
if (BadDebugInfo)
// If we were not able to translate the file:line:col information
@@ -638,7 +672,12 @@
FullSourceLoc Loc =
getBestLocationFromDebugLoc(D, BadDebugInfo, Filename, Line, Column);
- Diags.Report(Loc, diag::warn_profile_data_misexpect) << D.getMsg().str();
+ std::string Msg;
+ if (!Loc.isValid())
+ Msg = D.getLocationStr() + ": ";
+ raw_string_ostream MsgStream(Msg);
+ MsgStream << D.getMsg();
+ Diags.Report(Loc, diag::warn_profile_data_misexpect) << MsgStream.str();
if (BadDebugInfo)
// If we were not able to translate the file:line:col information
@@ -662,6 +701,8 @@
getBestLocationFromDebugLoc(D, BadDebugInfo, Filename, Line, Column);
std::string Msg;
+ if (!Loc.isValid())
+ Msg = "In " + D.getLocationStr() + ": ";
raw_string_ostream MsgStream(Msg);
MsgStream << D.getMsg();
@@ -1088,12 +1129,20 @@
Ctx.setInlineAsmDiagnosticHandler(BitcodeInlineAsmDiagHandler,
&Diagnostics);
+ // Set clang diagnostic handler. To do this we need to create a fake
+ // BackendConsumer.
+ BackendConsumer Result(BA, CI.getDiagnostics(), CI.getHeaderSearchOpts(),
+ CI.getPreprocessorOpts(), CI.getCodeGenOpts(),
+ CI.getTargetOpts(), CI.getLangOpts(),
+ CI.getFrontendOpts().ShowTimers,
+ std::move(LinkModules), *VMContext, nullptr);
+ Ctx.setDiagnosticHandler(
+ std::make_unique<ClangDiagnosticHandler>(CodeGenOpts, &Result));
+
Expected<std::unique_ptr<llvm::ToolOutputFile>> OptRecordFileOrErr =
setupOptimizationRemarks(
- Ctx, CodeGenOpts.OptRecordFile,
- CodeGenOpts.OptRecordPasses,
- CodeGenOpts.OptRecordFormat,
- CodeGenOpts.DiagnosticsWithHotness,
+ Ctx, CodeGenOpts.OptRecordFile, CodeGenOpts.OptRecordPasses,
+ CodeGenOpts.OptRecordFormat, CodeGenOpts.DiagnosticsWithHotness,
CodeGenOpts.DiagnosticsHotnessThreshold);
if (Error E = OptRecordFileOrErr.takeError()) {
@@ -1101,10 +1150,10 @@
return;
}
std::unique_ptr<llvm::ToolOutputFile> OptRecordFile =
- std::move(*OptRecordFileOrErr);
+ std::move(*OptRecordFileOrErr);
- EmitBackendOutput(Diagnostics, CI.getHeaderSearchOpts(),
- CodeGenOpts, TargetOpts, CI.getLangOpts(),
+ EmitBackendOutput(Diagnostics, CI.getHeaderSearchOpts(), CodeGenOpts,
+ TargetOpts, CI.getLangOpts(),
CI.getTarget().getDataLayout(), TheModule.get(), BA,
std::move(OS));
_______________________________________________
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
