rustc: Update LLVM

This commit updates the LLVM submodule in use to the current HEAD of the LLVM
repository. This is primarily being done to start picking up unwinding support
for MSVC, which is currently unimplemented in the revision of LLVM we are using.
Along the way a few changes had to be made:

* As usual, lots of C++ debuginfo bindings in LLVM changed, so there were some
  significant changes to our RustWrapper.cpp
* As usual, some pass management changed in LLVM, so clang was re-scrutinized to
  ensure that we're doing the same thing as clang.
* Some optimization options are now passed directly into the
  `PassManagerBuilder` instead of through CLI switches to LLVM.
* The `NoFramePointerElim` option was removed from LLVM, favoring instead the
  `no-frame-pointer-elim` function attribute instead.

Additionally, LLVM has picked up some new optimizations which required fixing an
existing soundness hole in the IR we generate. It appears that the current LLVM
we use does not expose this hole. When an enum is moved, the previous slot in
memory is overwritten with a bit pattern corresponding to "dropped". When the
drop glue for this slot is run, however, the switch on the discriminant can
often start executing the `unreachable` block of the switch due to the
discriminant now being outside the normal range. This was patched over locally
for now by having the `unreachable` block just change to a `ret void`.
This commit is contained in:
Alex Crichton
2015-05-14 12:10:43 -07:00
parent 71a8d313c8
commit f9d4149c29
17 changed files with 302 additions and 109 deletions

View File

@@ -15,12 +15,19 @@
#include "llvm/Support/CBindingWrapping.h"
#include "llvm/Support/FileSystem.h"
#include "llvm/Support/Host.h"
#if LLVM_VERSION_MINOR >= 7
#include "llvm/Analysis/TargetLibraryInfo.h"
#include "llvm/Analysis/TargetTransformInfo.h"
#else
#include "llvm/Target/TargetLibraryInfo.h"
#endif
#include "llvm/Transforms/IPO/PassManagerBuilder.h"
#include "llvm-c/Transforms/PassManagerBuilder.h"
using namespace llvm;
using namespace llvm::legacy;
extern cl::opt<bool> EnableARMEHABI;
@@ -71,7 +78,6 @@ LLVMRustCreateTargetMachine(const char *triple,
CodeGenOpt::Level OptLevel,
bool EnableSegmentedStacks,
bool UseSoftFloat,
bool NoFramePointerElim,
bool PositionIndependentExecutable,
bool FunctionSections,
bool DataSections) {
@@ -91,12 +97,12 @@ LLVMRustCreateTargetMachine(const char *triple,
TargetOptions Options;
Options.PositionIndependentExecutable = PositionIndependentExecutable;
Options.NoFramePointerElim = NoFramePointerElim;
Options.FloatABIType = FloatABI::Default;
Options.UseSoftFloat = UseSoftFloat;
if (UseSoftFloat) {
Options.FloatABIType = FloatABI::Soft;
}
Options.DataSections = DataSections;
Options.FunctionSections = FunctionSections;
TargetMachine *TM = TheTarget->createTargetMachine(Trip.getTriple(),
real_cpu,
@@ -105,8 +111,6 @@ LLVMRustCreateTargetMachine(const char *triple,
RM,
CM,
OptLevel);
TM->setDataSections(DataSections);
TM->setFunctionSections(FunctionSections);
return wrap(TM);
}
@@ -123,12 +127,32 @@ LLVMRustAddAnalysisPasses(LLVMTargetMachineRef TM,
LLVMPassManagerRef PMR,
LLVMModuleRef M) {
PassManagerBase *PM = unwrap(PMR);
#if LLVM_VERSION_MINOR >= 6
#if LLVM_VERSION_MINOR >= 7
PM->add(createTargetTransformInfoWrapperPass(
unwrap(TM)->getTargetIRAnalysis()));
#else
#if LLVM_VERSION_MINOR == 6
PM->add(new DataLayoutPass());
#else
PM->add(new DataLayoutPass(unwrap(M)));
#endif
unwrap(TM)->addAnalysisPasses(*PM);
#endif
}
extern "C" void
LLVMRustConfigurePassManagerBuilder(LLVMPassManagerBuilderRef PMB,
CodeGenOpt::Level OptLevel,
bool MergeFunctions,
bool SLPVectorize,
bool LoopVectorize) {
#if LLVM_VERSION_MINOR >= 6
// Ignore mergefunc for now as enabling it causes crashes.
//unwrap(PMB)->MergeFunctions = MergeFunctions;
#endif
unwrap(PMB)->SLPVectorize = SLPVectorize;
unwrap(PMB)->OptLevel = OptLevel;
unwrap(PMB)->LoopVectorize = LoopVectorize;
}
// Unfortunately, the LLVM C API doesn't provide a way to set the `LibraryInfo`
@@ -138,7 +162,11 @@ LLVMRustAddBuilderLibraryInfo(LLVMPassManagerBuilderRef PMB,
LLVMModuleRef M,
bool DisableSimplifyLibCalls) {
Triple TargetTriple(unwrap(M)->getTargetTriple());
#if LLVM_VERSION_MINOR >= 7
TargetLibraryInfoImpl *TLI = new TargetLibraryInfoImpl(TargetTriple);
#else
TargetLibraryInfo *TLI = new TargetLibraryInfo(TargetTriple);
#endif
if (DisableSimplifyLibCalls)
TLI->disableAllFunctions();
unwrap(PMB)->LibraryInfo = TLI;
@@ -151,10 +179,17 @@ LLVMRustAddLibraryInfo(LLVMPassManagerRef PMB,
LLVMModuleRef M,
bool DisableSimplifyLibCalls) {
Triple TargetTriple(unwrap(M)->getTargetTriple());
#if LLVM_VERSION_MINOR >= 7
TargetLibraryInfoImpl TLII(TargetTriple);
if (DisableSimplifyLibCalls)
TLII.disableAllFunctions();
unwrap(PMB)->add(new TargetLibraryInfoWrapperPass(TLII));
#else
TargetLibraryInfo *TLI = new TargetLibraryInfo(TargetTriple);
if (DisableSimplifyLibCalls)
TLI->disableAllFunctions();
unwrap(PMB)->add(TLI);
#endif
}
// Unfortunately, the LLVM C API doesn't provide an easy way of iterating over
@@ -204,10 +239,19 @@ LLVMRustWriteOutputFile(LLVMTargetMachineRef Target,
LLVMRustSetLastError(ErrorInfo.c_str());
return false;
}
formatted_raw_ostream FOS(OS);
#if LLVM_VERSION_MINOR >= 7
unwrap(Target)->addPassesToEmitFile(*PM, OS, FileType, false);
#else
formatted_raw_ostream FOS(OS);
unwrap(Target)->addPassesToEmitFile(*PM, FOS, FileType, false);
#endif
PM->run(*unwrap(M));
// Apparently `addPassesToEmitFile` adds an pointer to our on-the-stack output
// stream (OS), so the only real safe place to delete this is here? Don't we
// wish this was written in Rust?
delete PM;
return true;
}