2013-08-23 05:58:42 +02:00
|
|
|
#include <stdio.h>
|
|
|
|
|
2017-06-29 16:52:43 +02:00
|
|
|
#include <vector>
|
2017-11-29 21:00:10 +01:00
|
|
|
#include <set>
|
2017-06-29 16:52:43 +02:00
|
|
|
|
2020-07-26 19:11:30 +02:00
|
|
|
#include "LLVMWrapper.h"
|
2013-05-28 01:15:31 +02:00
|
|
|
|
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`.
2015-05-14 21:10:43 +02:00
|
|
|
#include "llvm/Analysis/TargetLibraryInfo.h"
|
|
|
|
#include "llvm/Analysis/TargetTransformInfo.h"
|
2018-12-09 11:20:20 +01:00
|
|
|
#include "llvm/CodeGen/TargetSubtargetInfo.h"
|
2019-12-31 14:02:22 +01:00
|
|
|
#include "llvm/InitializePasses.h"
|
2016-08-03 21:37:57 +02:00
|
|
|
#include "llvm/IR/AutoUpgrade.h"
|
2017-06-29 16:52:43 +02:00
|
|
|
#include "llvm/IR/AssemblyAnnotationWriter.h"
|
2018-12-09 11:20:20 +01:00
|
|
|
#include "llvm/IR/IntrinsicInst.h"
|
2020-01-05 19:16:58 +01:00
|
|
|
#include "llvm/IR/Verifier.h"
|
2020-04-23 20:45:55 +02:00
|
|
|
#include "llvm/Object/ObjectFile.h"
|
|
|
|
#include "llvm/Object/IRObjectFile.h"
|
2020-01-05 19:16:58 +01:00
|
|
|
#include "llvm/Passes/PassBuilder.h"
|
|
|
|
#include "llvm/Passes/StandardInstrumentations.h"
|
2016-12-30 12:22:11 +01:00
|
|
|
#include "llvm/Support/CBindingWrapping.h"
|
|
|
|
#include "llvm/Support/FileSystem.h"
|
|
|
|
#include "llvm/Support/Host.h"
|
2015-07-17 00:48:16 +02:00
|
|
|
#include "llvm/Target/TargetMachine.h"
|
2013-08-23 05:58:42 +02:00
|
|
|
#include "llvm/Transforms/IPO/PassManagerBuilder.h"
|
2016-09-24 18:37:04 +02:00
|
|
|
#include "llvm/Transforms/IPO/AlwaysInliner.h"
|
rustc: Implement ThinLTO
This commit is an implementation of LLVM's ThinLTO for consumption in rustc
itself. Currently today LTO works by merging all relevant LLVM modules into one
and then running optimization passes. "Thin" LTO operates differently by having
more sharded work and allowing parallelism opportunities between optimizing
codegen units. Further down the road Thin LTO also allows *incremental* LTO
which should enable even faster release builds without compromising on the
performance we have today.
This commit uses a `-Z thinlto` flag to gate whether ThinLTO is enabled. It then
also implements two forms of ThinLTO:
* In one mode we'll *only* perform ThinLTO over the codegen units produced in a
single compilation. That is, we won't load upstream rlibs, but we'll instead
just perform ThinLTO amongst all codegen units produced by the compiler for
the local crate. This is intended to emulate a desired end point where we have
codegen units turned on by default for all crates and ThinLTO allows us to do
this without performance loss.
* In anther mode, like full LTO today, we'll optimize all upstream dependencies
in "thin" mode. Unlike today, however, this LTO step is fully parallelized so
should finish much more quickly.
There's a good bit of comments about what the implementation is doing and where
it came from, but the tl;dr; is that currently most of the support here is
copied from upstream LLVM. This code duplication is done for a number of
reasons:
* Controlling parallelism means we can use the existing jobserver support to
avoid overloading machines.
* We will likely want a slightly different form of incremental caching which
integrates with our own incremental strategy, but this is yet to be
determined.
* This buys us some flexibility about when/where we run ThinLTO, as well as
having it tailored to fit our needs for the time being.
* Finally this allows us to reuse some artifacts such as our `TargetMachine`
creation, where all our options we used today aren't necessarily supported by
upstream LLVM yet.
My hope is that we can get some experience with this copy/paste in tree and then
eventually upstream some work to LLVM itself to avoid the duplication while
still ensuring our needs are met. Otherwise I fear that maintaining these
bindings may be quite costly over the years with LLVM updates!
2017-07-23 17:14:38 +02:00
|
|
|
#include "llvm/Transforms/IPO/FunctionImport.h"
|
|
|
|
#include "llvm/Transforms/Utils/FunctionImportUtils.h"
|
|
|
|
#include "llvm/LTO/LTO.h"
|
2013-08-23 05:58:42 +02:00
|
|
|
#include "llvm-c/Transforms/PassManagerBuilder.h"
|
2013-05-28 01:15:31 +02:00
|
|
|
|
2019-11-18 01:00:00 +01:00
|
|
|
#include "llvm/Transforms/Instrumentation.h"
|
|
|
|
#include "llvm/Transforms/Instrumentation/AddressSanitizer.h"
|
2020-02-01 00:58:28 +01:00
|
|
|
#include "llvm/Support/TimeProfiler.h"
|
2019-11-18 01:00:00 +01:00
|
|
|
#include "llvm/Transforms/Instrumentation/ThreadSanitizer.h"
|
|
|
|
#include "llvm/Transforms/Instrumentation/MemorySanitizer.h"
|
2020-01-05 19:16:58 +01:00
|
|
|
#include "llvm/Transforms/Utils/CanonicalizeAliases.h"
|
|
|
|
#include "llvm/Transforms/Utils/NameAnonGlobals.h"
|
2019-11-18 01:00:00 +01:00
|
|
|
|
2013-08-23 05:58:42 +02:00
|
|
|
using namespace llvm;
|
2013-05-28 01:15:31 +02:00
|
|
|
|
2013-08-23 05:58:42 +02:00
|
|
|
typedef struct LLVMOpaquePass *LLVMPassRef;
|
|
|
|
typedef struct LLVMOpaqueTargetMachine *LLVMTargetMachineRef;
|
|
|
|
|
|
|
|
DEFINE_STDCXX_CONVERSION_FUNCTIONS(Pass, LLVMPassRef)
|
|
|
|
DEFINE_STDCXX_CONVERSION_FUNCTIONS(TargetMachine, LLVMTargetMachineRef)
|
2020-06-26 03:52:41 +02:00
|
|
|
#if LLVM_VERSION_LT(11, 0)
|
2016-12-30 12:22:11 +01:00
|
|
|
DEFINE_STDCXX_CONVERSION_FUNCTIONS(PassManagerBuilder,
|
|
|
|
LLVMPassManagerBuilderRef)
|
2020-06-26 03:52:41 +02:00
|
|
|
#endif
|
2013-05-28 01:15:31 +02:00
|
|
|
|
2016-12-30 12:22:11 +01:00
|
|
|
extern "C" void LLVMInitializePasses() {
|
2013-05-29 10:08:20 +02:00
|
|
|
PassRegistry &Registry = *PassRegistry::getPassRegistry();
|
|
|
|
initializeCore(Registry);
|
|
|
|
initializeCodeGen(Registry);
|
|
|
|
initializeScalarOpts(Registry);
|
|
|
|
initializeVectorization(Registry);
|
|
|
|
initializeIPO(Registry);
|
|
|
|
initializeAnalysis(Registry);
|
|
|
|
initializeTransformUtils(Registry);
|
|
|
|
initializeInstCombine(Registry);
|
|
|
|
initializeInstrumentation(Registry);
|
|
|
|
initializeTarget(Registry);
|
|
|
|
}
|
2013-05-28 01:15:31 +02:00
|
|
|
|
2020-02-01 00:58:28 +01:00
|
|
|
extern "C" void LLVMTimeTraceProfilerInitialize() {
|
2020-02-04 20:35:50 +01:00
|
|
|
#if LLVM_VERSION_GE(10, 0)
|
|
|
|
timeTraceProfilerInitialize(
|
|
|
|
/* TimeTraceGranularity */ 0,
|
|
|
|
/* ProcName */ "rustc");
|
2020-11-07 23:25:45 +01:00
|
|
|
#else
|
2020-02-01 00:58:28 +01:00
|
|
|
timeTraceProfilerInitialize();
|
|
|
|
#endif
|
|
|
|
}
|
|
|
|
|
|
|
|
extern "C" void LLVMTimeTraceProfilerFinish(const char* FileName) {
|
|
|
|
StringRef FN(FileName);
|
|
|
|
std::error_code EC;
|
|
|
|
raw_fd_ostream OS(FN, EC, sys::fs::CD_CreateAlways);
|
|
|
|
|
|
|
|
timeTraceProfilerWrite(OS);
|
|
|
|
timeTraceProfilerCleanup();
|
|
|
|
}
|
|
|
|
|
2016-08-01 23:16:16 +02:00
|
|
|
enum class LLVMRustPassKind {
|
|
|
|
Other,
|
2016-01-25 02:22:24 +01:00
|
|
|
Function,
|
|
|
|
Module,
|
|
|
|
};
|
|
|
|
|
2016-12-31 18:01:23 +01:00
|
|
|
static LLVMRustPassKind toRust(PassKind Kind) {
|
|
|
|
switch (Kind) {
|
2016-08-01 23:16:16 +02:00
|
|
|
case PT_Function:
|
2016-12-30 12:22:11 +01:00
|
|
|
return LLVMRustPassKind::Function;
|
2016-08-01 23:16:16 +02:00
|
|
|
case PT_Module:
|
2016-12-30 12:22:11 +01:00
|
|
|
return LLVMRustPassKind::Module;
|
2016-08-01 23:16:16 +02:00
|
|
|
default:
|
2016-12-30 12:22:11 +01:00
|
|
|
return LLVMRustPassKind::Other;
|
2016-08-01 23:16:16 +02:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2016-12-30 12:22:11 +01:00
|
|
|
extern "C" LLVMPassRef LLVMRustFindAndCreatePass(const char *PassName) {
|
|
|
|
StringRef SR(PassName);
|
|
|
|
PassRegistry *PR = PassRegistry::getPassRegistry();
|
2013-05-28 01:15:31 +02:00
|
|
|
|
2016-12-30 12:22:11 +01:00
|
|
|
const PassInfo *PI = PR->getPassInfo(SR);
|
|
|
|
if (PI) {
|
|
|
|
return wrap(PI->createPass());
|
|
|
|
}
|
2016-12-30 13:21:21 +01:00
|
|
|
return nullptr;
|
2016-01-25 02:22:24 +01:00
|
|
|
}
|
|
|
|
|
2019-11-18 01:00:00 +01:00
|
|
|
extern "C" LLVMPassRef LLVMRustCreateAddressSanitizerFunctionPass(bool Recover) {
|
|
|
|
const bool CompileKernel = false;
|
2020-01-27 01:00:00 +01:00
|
|
|
const bool UseAfterScope = true;
|
2019-11-18 01:00:00 +01:00
|
|
|
|
2020-01-27 01:00:00 +01:00
|
|
|
return wrap(createAddressSanitizerFunctionPass(CompileKernel, Recover, UseAfterScope));
|
2019-11-18 01:00:00 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
extern "C" LLVMPassRef LLVMRustCreateModuleAddressSanitizerPass(bool Recover) {
|
|
|
|
const bool CompileKernel = false;
|
|
|
|
|
|
|
|
return wrap(createModuleAddressSanitizerLegacyPassPass(CompileKernel, Recover));
|
|
|
|
}
|
|
|
|
|
|
|
|
extern "C" LLVMPassRef LLVMRustCreateMemorySanitizerPass(int TrackOrigins, bool Recover) {
|
|
|
|
const bool CompileKernel = false;
|
|
|
|
|
|
|
|
return wrap(createMemorySanitizerLegacyPassPass(
|
|
|
|
MemorySanitizerOptions{TrackOrigins, Recover, CompileKernel}));
|
|
|
|
}
|
|
|
|
|
|
|
|
extern "C" LLVMPassRef LLVMRustCreateThreadSanitizerPass() {
|
|
|
|
return wrap(createThreadSanitizerLegacyPassPass());
|
|
|
|
}
|
|
|
|
|
2016-12-31 18:01:23 +01:00
|
|
|
extern "C" LLVMRustPassKind LLVMRustPassKind(LLVMPassRef RustPass) {
|
|
|
|
assert(RustPass);
|
|
|
|
Pass *Pass = unwrap(RustPass);
|
|
|
|
return toRust(Pass->getPassKind());
|
2016-01-25 02:22:24 +01:00
|
|
|
}
|
|
|
|
|
2016-12-31 18:01:23 +01:00
|
|
|
extern "C" void LLVMRustAddPass(LLVMPassManagerRef PMR, LLVMPassRef RustPass) {
|
|
|
|
assert(RustPass);
|
|
|
|
Pass *Pass = unwrap(RustPass);
|
|
|
|
PassManagerBase *PMB = unwrap(PMR);
|
|
|
|
PMB->add(Pass);
|
2013-08-23 05:58:42 +02:00
|
|
|
}
|
|
|
|
|
rustc: Implement ThinLTO
This commit is an implementation of LLVM's ThinLTO for consumption in rustc
itself. Currently today LTO works by merging all relevant LLVM modules into one
and then running optimization passes. "Thin" LTO operates differently by having
more sharded work and allowing parallelism opportunities between optimizing
codegen units. Further down the road Thin LTO also allows *incremental* LTO
which should enable even faster release builds without compromising on the
performance we have today.
This commit uses a `-Z thinlto` flag to gate whether ThinLTO is enabled. It then
also implements two forms of ThinLTO:
* In one mode we'll *only* perform ThinLTO over the codegen units produced in a
single compilation. That is, we won't load upstream rlibs, but we'll instead
just perform ThinLTO amongst all codegen units produced by the compiler for
the local crate. This is intended to emulate a desired end point where we have
codegen units turned on by default for all crates and ThinLTO allows us to do
this without performance loss.
* In anther mode, like full LTO today, we'll optimize all upstream dependencies
in "thin" mode. Unlike today, however, this LTO step is fully parallelized so
should finish much more quickly.
There's a good bit of comments about what the implementation is doing and where
it came from, but the tl;dr; is that currently most of the support here is
copied from upstream LLVM. This code duplication is done for a number of
reasons:
* Controlling parallelism means we can use the existing jobserver support to
avoid overloading machines.
* We will likely want a slightly different form of incremental caching which
integrates with our own incremental strategy, but this is yet to be
determined.
* This buys us some flexibility about when/where we run ThinLTO, as well as
having it tailored to fit our needs for the time being.
* Finally this allows us to reuse some artifacts such as our `TargetMachine`
creation, where all our options we used today aren't necessarily supported by
upstream LLVM yet.
My hope is that we can get some experience with this copy/paste in tree and then
eventually upstream some work to LLVM itself to avoid the duplication while
still ensuring our needs are met. Otherwise I fear that maintaining these
bindings may be quite costly over the years with LLVM updates!
2017-07-23 17:14:38 +02:00
|
|
|
extern "C"
|
2018-11-05 14:52:08 +01:00
|
|
|
void LLVMRustPassManagerBuilderPopulateThinLTOPassManager(
|
rustc: Implement ThinLTO
This commit is an implementation of LLVM's ThinLTO for consumption in rustc
itself. Currently today LTO works by merging all relevant LLVM modules into one
and then running optimization passes. "Thin" LTO operates differently by having
more sharded work and allowing parallelism opportunities between optimizing
codegen units. Further down the road Thin LTO also allows *incremental* LTO
which should enable even faster release builds without compromising on the
performance we have today.
This commit uses a `-Z thinlto` flag to gate whether ThinLTO is enabled. It then
also implements two forms of ThinLTO:
* In one mode we'll *only* perform ThinLTO over the codegen units produced in a
single compilation. That is, we won't load upstream rlibs, but we'll instead
just perform ThinLTO amongst all codegen units produced by the compiler for
the local crate. This is intended to emulate a desired end point where we have
codegen units turned on by default for all crates and ThinLTO allows us to do
this without performance loss.
* In anther mode, like full LTO today, we'll optimize all upstream dependencies
in "thin" mode. Unlike today, however, this LTO step is fully parallelized so
should finish much more quickly.
There's a good bit of comments about what the implementation is doing and where
it came from, but the tl;dr; is that currently most of the support here is
copied from upstream LLVM. This code duplication is done for a number of
reasons:
* Controlling parallelism means we can use the existing jobserver support to
avoid overloading machines.
* We will likely want a slightly different form of incremental caching which
integrates with our own incremental strategy, but this is yet to be
determined.
* This buys us some flexibility about when/where we run ThinLTO, as well as
having it tailored to fit our needs for the time being.
* Finally this allows us to reuse some artifacts such as our `TargetMachine`
creation, where all our options we used today aren't necessarily supported by
upstream LLVM yet.
My hope is that we can get some experience with this copy/paste in tree and then
eventually upstream some work to LLVM itself to avoid the duplication while
still ensuring our needs are met. Otherwise I fear that maintaining these
bindings may be quite costly over the years with LLVM updates!
2017-07-23 17:14:38 +02:00
|
|
|
LLVMPassManagerBuilderRef PMBR,
|
|
|
|
LLVMPassManagerRef PMR
|
|
|
|
) {
|
|
|
|
unwrap(PMBR)->populateThinLTOPassManager(*unwrap(PMR));
|
|
|
|
}
|
|
|
|
|
2019-07-13 20:17:16 +02:00
|
|
|
extern "C"
|
|
|
|
void LLVMRustAddLastExtensionPasses(
|
|
|
|
LLVMPassManagerBuilderRef PMBR, LLVMPassRef *Passes, size_t NumPasses) {
|
|
|
|
auto AddExtensionPasses = [Passes, NumPasses](
|
|
|
|
const PassManagerBuilder &Builder, PassManagerBase &PM) {
|
|
|
|
for (size_t I = 0; I < NumPasses; I++) {
|
|
|
|
PM.add(unwrap(Passes[I]));
|
|
|
|
}
|
|
|
|
};
|
|
|
|
// Add the passes to both of the pre-finalization extension points,
|
|
|
|
// so they are run for optimized and non-optimized builds.
|
|
|
|
unwrap(PMBR)->addExtension(PassManagerBuilder::EP_OptimizerLast,
|
|
|
|
AddExtensionPasses);
|
|
|
|
unwrap(PMBR)->addExtension(PassManagerBuilder::EP_EnabledOnOptLevel0,
|
|
|
|
AddExtensionPasses);
|
|
|
|
}
|
|
|
|
|
2016-02-16 17:07:30 +01:00
|
|
|
#ifdef LLVM_COMPONENT_X86
|
|
|
|
#define SUBTARGET_X86 SUBTARGET(X86)
|
|
|
|
#else
|
|
|
|
#define SUBTARGET_X86
|
|
|
|
#endif
|
|
|
|
|
|
|
|
#ifdef LLVM_COMPONENT_ARM
|
|
|
|
#define SUBTARGET_ARM SUBTARGET(ARM)
|
|
|
|
#else
|
|
|
|
#define SUBTARGET_ARM
|
|
|
|
#endif
|
|
|
|
|
|
|
|
#ifdef LLVM_COMPONENT_AARCH64
|
|
|
|
#define SUBTARGET_AARCH64 SUBTARGET(AArch64)
|
|
|
|
#else
|
|
|
|
#define SUBTARGET_AARCH64
|
|
|
|
#endif
|
|
|
|
|
2016-05-06 15:32:10 +02:00
|
|
|
#ifdef LLVM_COMPONENT_AVR
|
|
|
|
#define SUBTARGET_AVR SUBTARGET(AVR)
|
|
|
|
#else
|
|
|
|
#define SUBTARGET_AVR
|
|
|
|
#endif
|
|
|
|
|
2016-02-16 17:07:30 +01:00
|
|
|
#ifdef LLVM_COMPONENT_MIPS
|
|
|
|
#define SUBTARGET_MIPS SUBTARGET(Mips)
|
|
|
|
#else
|
|
|
|
#define SUBTARGET_MIPS
|
|
|
|
#endif
|
|
|
|
|
|
|
|
#ifdef LLVM_COMPONENT_POWERPC
|
|
|
|
#define SUBTARGET_PPC SUBTARGET(PPC)
|
|
|
|
#else
|
|
|
|
#define SUBTARGET_PPC
|
|
|
|
#endif
|
|
|
|
|
2016-08-28 20:18:28 +02:00
|
|
|
#ifdef LLVM_COMPONENT_SYSTEMZ
|
|
|
|
#define SUBTARGET_SYSTEMZ SUBTARGET(SystemZ)
|
|
|
|
#else
|
|
|
|
#define SUBTARGET_SYSTEMZ
|
|
|
|
#endif
|
|
|
|
|
2016-11-09 22:56:10 +01:00
|
|
|
#ifdef LLVM_COMPONENT_MSP430
|
|
|
|
#define SUBTARGET_MSP430 SUBTARGET(MSP430)
|
|
|
|
#else
|
|
|
|
#define SUBTARGET_MSP430
|
|
|
|
#endif
|
|
|
|
|
2018-07-24 12:03:28 +02:00
|
|
|
#ifdef LLVM_COMPONENT_RISCV
|
|
|
|
#define SUBTARGET_RISCV SUBTARGET(RISCV)
|
|
|
|
#else
|
|
|
|
#define SUBTARGET_RISCV
|
|
|
|
#endif
|
|
|
|
|
2016-12-03 17:53:31 +01:00
|
|
|
#ifdef LLVM_COMPONENT_SPARC
|
|
|
|
#define SUBTARGET_SPARC SUBTARGET(Sparc)
|
|
|
|
#else
|
|
|
|
#define SUBTARGET_SPARC
|
|
|
|
#endif
|
|
|
|
|
2017-04-09 08:03:31 +02:00
|
|
|
#ifdef LLVM_COMPONENT_HEXAGON
|
|
|
|
#define SUBTARGET_HEXAGON SUBTARGET(Hexagon)
|
|
|
|
#else
|
|
|
|
#define SUBTARGET_HEXAGON
|
|
|
|
#endif
|
|
|
|
|
2016-12-30 12:22:11 +01:00
|
|
|
#define GEN_SUBTARGETS \
|
|
|
|
SUBTARGET_X86 \
|
|
|
|
SUBTARGET_ARM \
|
|
|
|
SUBTARGET_AARCH64 \
|
2016-05-06 15:32:10 +02:00
|
|
|
SUBTARGET_AVR \
|
2016-12-30 12:22:11 +01:00
|
|
|
SUBTARGET_MIPS \
|
|
|
|
SUBTARGET_PPC \
|
|
|
|
SUBTARGET_SYSTEMZ \
|
2017-01-01 04:40:10 +01:00
|
|
|
SUBTARGET_MSP430 \
|
2017-04-09 08:03:31 +02:00
|
|
|
SUBTARGET_SPARC \
|
2018-07-24 12:03:28 +02:00
|
|
|
SUBTARGET_HEXAGON \
|
|
|
|
SUBTARGET_RISCV \
|
2016-12-30 12:22:11 +01:00
|
|
|
|
|
|
|
#define SUBTARGET(x) \
|
|
|
|
namespace llvm { \
|
|
|
|
extern const SubtargetFeatureKV x##FeatureKV[]; \
|
|
|
|
extern const SubtargetFeatureKV x##SubTypeKV[]; \
|
2016-02-16 17:07:30 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
GEN_SUBTARGETS
|
|
|
|
#undef SUBTARGET
|
|
|
|
|
2016-12-30 12:22:11 +01:00
|
|
|
extern "C" bool LLVMRustHasFeature(LLVMTargetMachineRef TM,
|
2016-12-31 18:01:23 +01:00
|
|
|
const char *Feature) {
|
2016-12-30 12:22:11 +01:00
|
|
|
TargetMachine *Target = unwrap(TM);
|
|
|
|
const MCSubtargetInfo *MCInfo = Target->getMCSubtargetInfo();
|
2018-03-27 21:27:45 +02:00
|
|
|
return MCInfo->checkFeatures(std::string("+") + Feature);
|
2016-02-16 17:07:30 +01:00
|
|
|
}
|
|
|
|
|
2016-08-01 23:16:16 +02:00
|
|
|
enum class LLVMRustCodeModel {
|
2020-05-07 02:34:27 +02:00
|
|
|
Tiny,
|
2016-12-30 12:22:11 +01:00
|
|
|
Small,
|
|
|
|
Kernel,
|
|
|
|
Medium,
|
|
|
|
Large,
|
2018-01-23 02:01:36 +01:00
|
|
|
None,
|
2016-08-01 23:16:16 +02:00
|
|
|
};
|
|
|
|
|
2020-05-07 02:34:27 +02:00
|
|
|
static Optional<CodeModel::Model> fromRust(LLVMRustCodeModel Model) {
|
2016-12-31 18:01:23 +01:00
|
|
|
switch (Model) {
|
2020-05-07 02:34:27 +02:00
|
|
|
case LLVMRustCodeModel::Tiny:
|
|
|
|
return CodeModel::Tiny;
|
2016-12-30 12:22:11 +01:00
|
|
|
case LLVMRustCodeModel::Small:
|
|
|
|
return CodeModel::Small;
|
|
|
|
case LLVMRustCodeModel::Kernel:
|
|
|
|
return CodeModel::Kernel;
|
|
|
|
case LLVMRustCodeModel::Medium:
|
|
|
|
return CodeModel::Medium;
|
|
|
|
case LLVMRustCodeModel::Large:
|
|
|
|
return CodeModel::Large;
|
2020-05-07 02:34:27 +02:00
|
|
|
case LLVMRustCodeModel::None:
|
|
|
|
return None;
|
2016-12-30 12:22:11 +01:00
|
|
|
default:
|
2017-11-20 17:47:29 +01:00
|
|
|
report_fatal_error("Bad CodeModel.");
|
2016-08-01 23:16:16 +02:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
enum class LLVMRustCodeGenOptLevel {
|
2016-12-30 12:22:11 +01:00
|
|
|
None,
|
|
|
|
Less,
|
|
|
|
Default,
|
|
|
|
Aggressive,
|
2016-08-01 23:16:16 +02:00
|
|
|
};
|
|
|
|
|
2016-12-31 18:01:23 +01:00
|
|
|
static CodeGenOpt::Level fromRust(LLVMRustCodeGenOptLevel Level) {
|
|
|
|
switch (Level) {
|
2016-12-30 12:22:11 +01:00
|
|
|
case LLVMRustCodeGenOptLevel::None:
|
|
|
|
return CodeGenOpt::None;
|
|
|
|
case LLVMRustCodeGenOptLevel::Less:
|
|
|
|
return CodeGenOpt::Less;
|
|
|
|
case LLVMRustCodeGenOptLevel::Default:
|
|
|
|
return CodeGenOpt::Default;
|
|
|
|
case LLVMRustCodeGenOptLevel::Aggressive:
|
|
|
|
return CodeGenOpt::Aggressive;
|
|
|
|
default:
|
2017-11-20 17:47:29 +01:00
|
|
|
report_fatal_error("Bad CodeGenOptLevel.");
|
2016-08-01 23:16:16 +02:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2020-01-05 19:16:58 +01:00
|
|
|
enum class LLVMRustPassBuilderOptLevel {
|
|
|
|
O0,
|
|
|
|
O1,
|
|
|
|
O2,
|
|
|
|
O3,
|
|
|
|
Os,
|
|
|
|
Oz,
|
|
|
|
};
|
|
|
|
|
|
|
|
static PassBuilder::OptimizationLevel fromRust(LLVMRustPassBuilderOptLevel Level) {
|
|
|
|
switch (Level) {
|
|
|
|
case LLVMRustPassBuilderOptLevel::O0:
|
2020-06-26 03:52:41 +02:00
|
|
|
return PassBuilder::OptimizationLevel::O0;
|
2020-01-05 19:16:58 +01:00
|
|
|
case LLVMRustPassBuilderOptLevel::O1:
|
2020-06-26 03:52:41 +02:00
|
|
|
return PassBuilder::OptimizationLevel::O1;
|
2020-01-05 19:16:58 +01:00
|
|
|
case LLVMRustPassBuilderOptLevel::O2:
|
2020-06-26 03:52:41 +02:00
|
|
|
return PassBuilder::OptimizationLevel::O2;
|
2020-01-05 19:16:58 +01:00
|
|
|
case LLVMRustPassBuilderOptLevel::O3:
|
2020-06-26 03:52:41 +02:00
|
|
|
return PassBuilder::OptimizationLevel::O3;
|
2020-01-05 19:16:58 +01:00
|
|
|
case LLVMRustPassBuilderOptLevel::Os:
|
2020-06-26 03:52:41 +02:00
|
|
|
return PassBuilder::OptimizationLevel::Os;
|
2020-01-05 19:16:58 +01:00
|
|
|
case LLVMRustPassBuilderOptLevel::Oz:
|
2020-06-26 03:52:41 +02:00
|
|
|
return PassBuilder::OptimizationLevel::Oz;
|
2020-01-05 19:16:58 +01:00
|
|
|
default:
|
|
|
|
report_fatal_error("Bad PassBuilderOptLevel.");
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2020-04-23 19:49:00 +02:00
|
|
|
enum class LLVMRustRelocModel {
|
2017-04-29 00:21:59 +02:00
|
|
|
Static,
|
|
|
|
PIC,
|
|
|
|
DynamicNoPic,
|
|
|
|
ROPI,
|
|
|
|
RWPI,
|
|
|
|
ROPIRWPI,
|
|
|
|
};
|
|
|
|
|
2020-04-23 19:49:00 +02:00
|
|
|
static Reloc::Model fromRust(LLVMRustRelocModel RustReloc) {
|
2017-04-29 00:21:59 +02:00
|
|
|
switch (RustReloc) {
|
2020-04-23 19:49:00 +02:00
|
|
|
case LLVMRustRelocModel::Static:
|
2017-04-29 00:21:59 +02:00
|
|
|
return Reloc::Static;
|
2020-04-23 19:49:00 +02:00
|
|
|
case LLVMRustRelocModel::PIC:
|
2017-04-29 00:21:59 +02:00
|
|
|
return Reloc::PIC_;
|
2020-04-23 19:49:00 +02:00
|
|
|
case LLVMRustRelocModel::DynamicNoPic:
|
2017-04-29 00:21:59 +02:00
|
|
|
return Reloc::DynamicNoPIC;
|
2020-04-23 19:49:00 +02:00
|
|
|
case LLVMRustRelocModel::ROPI:
|
2017-04-29 00:21:59 +02:00
|
|
|
return Reloc::ROPI;
|
2020-04-23 19:49:00 +02:00
|
|
|
case LLVMRustRelocModel::RWPI:
|
2017-04-29 00:21:59 +02:00
|
|
|
return Reloc::RWPI;
|
2020-04-23 19:49:00 +02:00
|
|
|
case LLVMRustRelocModel::ROPIRWPI:
|
2017-04-29 00:21:59 +02:00
|
|
|
return Reloc::ROPI_RWPI;
|
|
|
|
}
|
2017-11-20 17:47:29 +01:00
|
|
|
report_fatal_error("Bad RelocModel.");
|
2017-04-29 00:21:59 +02:00
|
|
|
}
|
|
|
|
|
2018-11-29 23:05:23 +01:00
|
|
|
#ifdef LLVM_RUSTLLVM
|
2016-07-10 16:22:13 +02:00
|
|
|
/// getLongestEntryLength - Return the length of the longest entry in the table.
|
2019-07-06 18:52:26 +02:00
|
|
|
template<typename KV>
|
|
|
|
static size_t getLongestEntryLength(ArrayRef<KV> Table) {
|
2016-07-10 16:22:13 +02:00
|
|
|
size_t MaxLen = 0;
|
|
|
|
for (auto &I : Table)
|
|
|
|
MaxLen = std::max(MaxLen, std::strlen(I.Key));
|
|
|
|
return MaxLen;
|
|
|
|
}
|
|
|
|
|
2016-12-30 12:22:11 +01:00
|
|
|
extern "C" void LLVMRustPrintTargetCPUs(LLVMTargetMachineRef TM) {
|
|
|
|
const TargetMachine *Target = unwrap(TM);
|
|
|
|
const MCSubtargetInfo *MCInfo = Target->getMCSubtargetInfo();
|
2017-09-08 03:07:16 +02:00
|
|
|
const Triple::ArchType HostArch = Triple(sys::getProcessTriple()).getArch();
|
|
|
|
const Triple::ArchType TargetArch = Target->getTargetTriple().getArch();
|
2019-07-06 18:52:26 +02:00
|
|
|
const ArrayRef<SubtargetSubTypeKV> CPUTable = MCInfo->getCPUTable();
|
2016-12-30 12:22:11 +01:00
|
|
|
unsigned MaxCPULen = getLongestEntryLength(CPUTable);
|
|
|
|
|
|
|
|
printf("Available CPUs for this target:\n");
|
2017-09-08 03:07:16 +02:00
|
|
|
if (HostArch == TargetArch) {
|
|
|
|
const StringRef HostCPU = sys::getHostCPUName();
|
|
|
|
printf(" %-*s - Select the CPU of the current host (currently %.*s).\n",
|
|
|
|
MaxCPULen, "native", (int)HostCPU.size(), HostCPU.data());
|
|
|
|
}
|
2016-12-30 12:22:11 +01:00
|
|
|
for (auto &CPU : CPUTable)
|
2019-07-06 18:52:26 +02:00
|
|
|
printf(" %-*s\n", MaxCPULen, CPU.Key);
|
2016-12-30 12:22:11 +01:00
|
|
|
printf("\n");
|
2016-07-10 16:22:13 +02:00
|
|
|
}
|
|
|
|
|
2016-12-30 12:22:11 +01:00
|
|
|
extern "C" void LLVMRustPrintTargetFeatures(LLVMTargetMachineRef TM) {
|
|
|
|
const TargetMachine *Target = unwrap(TM);
|
|
|
|
const MCSubtargetInfo *MCInfo = Target->getMCSubtargetInfo();
|
|
|
|
const ArrayRef<SubtargetFeatureKV> FeatTable = MCInfo->getFeatureTable();
|
|
|
|
unsigned MaxFeatLen = getLongestEntryLength(FeatTable);
|
|
|
|
|
|
|
|
printf("Available features for this target:\n");
|
|
|
|
for (auto &Feature : FeatTable)
|
|
|
|
printf(" %-*s - %s.\n", MaxFeatLen, Feature.Key, Feature.Desc);
|
2020-06-17 07:01:46 +02:00
|
|
|
printf("\nRust-specific features:\n");
|
2020-06-16 20:53:30 +02:00
|
|
|
printf(" %-*s - %s.\n",
|
|
|
|
MaxFeatLen,
|
|
|
|
"crt-static",
|
2020-06-16 20:14:32 +02:00
|
|
|
"Enables libraries with C Run-time Libraries(CRT) to be statically linked"
|
|
|
|
);
|
2020-06-14 18:14:11 +02:00
|
|
|
printf("\n");
|
2016-12-30 12:22:11 +01:00
|
|
|
|
|
|
|
printf("Use +feature to enable a feature, or -feature to disable it.\n"
|
|
|
|
"For example, rustc -C -target-cpu=mycpu -C "
|
|
|
|
"target-feature=+feature1,-feature2\n\n");
|
2016-07-10 16:22:13 +02:00
|
|
|
}
|
|
|
|
|
2016-07-24 11:49:10 +02:00
|
|
|
#else
|
|
|
|
|
2016-12-30 12:22:11 +01:00
|
|
|
extern "C" void LLVMRustPrintTargetCPUs(LLVMTargetMachineRef) {
|
|
|
|
printf("Target CPU help is not supported by this LLVM version.\n\n");
|
2016-07-24 11:49:10 +02:00
|
|
|
}
|
|
|
|
|
2016-12-30 12:22:11 +01:00
|
|
|
extern "C" void LLVMRustPrintTargetFeatures(LLVMTargetMachineRef) {
|
|
|
|
printf("Target features help is not supported by this LLVM version.\n\n");
|
2016-07-24 11:49:10 +02:00
|
|
|
}
|
|
|
|
#endif
|
|
|
|
|
2018-08-23 20:03:22 +02:00
|
|
|
extern "C" const char* LLVMRustGetHostCPUName(size_t *len) {
|
|
|
|
StringRef Name = sys::getHostCPUName();
|
|
|
|
*len = Name.size();
|
|
|
|
return Name.data();
|
|
|
|
}
|
|
|
|
|
2016-12-30 12:22:11 +01:00
|
|
|
extern "C" LLVMTargetMachineRef LLVMRustCreateTargetMachine(
|
2016-12-31 18:01:23 +01:00
|
|
|
const char *TripleStr, const char *CPU, const char *Feature,
|
2020-04-23 19:49:00 +02:00
|
|
|
const char *ABIStr, LLVMRustCodeModel RustCM, LLVMRustRelocModel RustReloc,
|
2016-12-31 18:01:23 +01:00
|
|
|
LLVMRustCodeGenOptLevel RustOptLevel, bool UseSoftFloat,
|
2020-05-21 19:53:41 +02:00
|
|
|
bool FunctionSections,
|
2017-10-23 05:01:00 +02:00
|
|
|
bool DataSections,
|
|
|
|
bool TrapUnreachable,
|
2018-08-12 19:59:18 +02:00
|
|
|
bool Singlethread,
|
2018-09-13 19:43:15 +02:00
|
|
|
bool AsmComments,
|
2019-12-02 13:22:45 +01:00
|
|
|
bool EmitStackSizeSection,
|
2020-04-17 04:40:11 +02:00
|
|
|
bool RelaxELFRelocations,
|
2020-09-23 17:25:20 +02:00
|
|
|
bool UseInitArray,
|
|
|
|
const char *SplitDwarfFile) {
|
2016-07-24 22:31:16 +02:00
|
|
|
|
2016-12-31 18:01:23 +01:00
|
|
|
auto OptLevel = fromRust(RustOptLevel);
|
2017-04-29 00:21:59 +02:00
|
|
|
auto RM = fromRust(RustReloc);
|
2020-05-07 02:34:27 +02:00
|
|
|
auto CM = fromRust(RustCM);
|
2016-07-24 22:31:16 +02:00
|
|
|
|
2016-12-30 12:22:11 +01:00
|
|
|
std::string Error;
|
2016-12-31 18:01:23 +01:00
|
|
|
Triple Trip(Triple::normalize(TripleStr));
|
2016-12-30 12:22:11 +01:00
|
|
|
const llvm::Target *TheTarget =
|
|
|
|
TargetRegistry::lookupTarget(Trip.getTriple(), Error);
|
2016-12-30 13:21:21 +01:00
|
|
|
if (TheTarget == nullptr) {
|
2016-12-30 12:22:11 +01:00
|
|
|
LLVMRustSetLastError(Error.c_str());
|
2016-12-30 13:21:21 +01:00
|
|
|
return nullptr;
|
2016-12-30 12:22:11 +01:00
|
|
|
}
|
2013-08-23 05:58:42 +02:00
|
|
|
|
2016-12-30 12:22:11 +01:00
|
|
|
TargetOptions Options;
|
2016-07-13 00:41:40 +02:00
|
|
|
|
2016-12-30 12:22:11 +01:00
|
|
|
Options.FloatABIType = FloatABI::Default;
|
|
|
|
if (UseSoftFloat) {
|
|
|
|
Options.FloatABIType = FloatABI::Soft;
|
|
|
|
}
|
|
|
|
Options.DataSections = DataSections;
|
|
|
|
Options.FunctionSections = FunctionSections;
|
2018-08-12 19:59:18 +02:00
|
|
|
Options.MCOptions.AsmVerbose = AsmComments;
|
|
|
|
Options.MCOptions.PreserveAsmComments = AsmComments;
|
2019-10-30 05:12:05 +01:00
|
|
|
Options.MCOptions.ABIName = ABIStr;
|
2020-09-23 17:25:20 +02:00
|
|
|
if (SplitDwarfFile) {
|
|
|
|
Options.MCOptions.SplitDwarfFile = SplitDwarfFile;
|
|
|
|
}
|
2019-12-02 13:22:45 +01:00
|
|
|
Options.RelaxELFRelocations = RelaxELFRelocations;
|
2020-04-17 04:40:11 +02:00
|
|
|
Options.UseInitArray = UseInitArray;
|
2016-12-30 12:22:11 +01:00
|
|
|
|
2017-11-11 16:08:00 +01:00
|
|
|
if (TrapUnreachable) {
|
2018-05-08 15:10:16 +02:00
|
|
|
// Tell LLVM to codegen `unreachable` into an explicit trap instruction.
|
2017-11-11 16:08:00 +01:00
|
|
|
// This limits the extent of possible undefined behavior in some cases, as
|
|
|
|
// it prevents control flow from "falling through" into whatever code
|
|
|
|
// happens to be laid out next in memory.
|
|
|
|
Options.TrapUnreachable = true;
|
|
|
|
}
|
2017-11-10 20:00:52 +01:00
|
|
|
|
2017-10-23 05:01:00 +02:00
|
|
|
if (Singlethread) {
|
|
|
|
Options.ThreadModel = ThreadModel::Single;
|
|
|
|
}
|
|
|
|
|
2018-09-13 19:43:15 +02:00
|
|
|
Options.EmitStackSizeSection = EmitStackSizeSection;
|
|
|
|
|
2016-12-30 12:22:11 +01:00
|
|
|
TargetMachine *TM = TheTarget->createTargetMachine(
|
2018-08-23 20:03:22 +02:00
|
|
|
Trip.getTriple(), CPU, Feature, Options, RM, CM, OptLevel);
|
2016-12-30 12:22:11 +01:00
|
|
|
return wrap(TM);
|
2013-08-23 05:58:42 +02:00
|
|
|
}
|
|
|
|
|
2016-12-30 12:22:11 +01:00
|
|
|
extern "C" void LLVMRustDisposeTargetMachine(LLVMTargetMachineRef TM) {
|
|
|
|
delete unwrap(TM);
|
2013-08-23 05:58:42 +02:00
|
|
|
}
|
|
|
|
|
2016-12-30 12:22:11 +01:00
|
|
|
extern "C" void LLVMRustConfigurePassManagerBuilder(
|
2016-12-31 18:01:23 +01:00
|
|
|
LLVMPassManagerBuilderRef PMBR, LLVMRustCodeGenOptLevel OptLevel,
|
2018-05-12 14:07:20 +02:00
|
|
|
bool MergeFunctions, bool SLPVectorize, bool LoopVectorize, bool PrepareForThinLTO,
|
2018-02-19 01:57:12 +01:00
|
|
|
const char* PGOGenPath, const char* PGOUsePath) {
|
2018-03-29 09:41:41 +02:00
|
|
|
unwrap(PMBR)->MergeFunctions = MergeFunctions;
|
2016-12-31 18:01:23 +01:00
|
|
|
unwrap(PMBR)->SLPVectorize = SLPVectorize;
|
|
|
|
unwrap(PMBR)->OptLevel = fromRust(OptLevel);
|
|
|
|
unwrap(PMBR)->LoopVectorize = LoopVectorize;
|
2018-05-12 14:07:20 +02:00
|
|
|
unwrap(PMBR)->PrepareForThinLTO = PrepareForThinLTO;
|
2018-03-15 16:56:45 +01:00
|
|
|
|
2018-02-19 01:57:12 +01:00
|
|
|
if (PGOGenPath) {
|
|
|
|
assert(!PGOUsePath);
|
|
|
|
unwrap(PMBR)->EnablePGOInstrGen = true;
|
|
|
|
unwrap(PMBR)->PGOInstrGen = PGOGenPath;
|
|
|
|
}
|
|
|
|
if (PGOUsePath) {
|
|
|
|
assert(!PGOGenPath);
|
|
|
|
unwrap(PMBR)->PGOInstrUse = PGOUsePath;
|
|
|
|
}
|
2013-08-23 05:58:42 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
// Unfortunately, the LLVM C API doesn't provide a way to set the `LibraryInfo`
|
|
|
|
// field of a PassManagerBuilder, we expose our own method of doing so.
|
2016-12-31 18:01:23 +01:00
|
|
|
extern "C" void LLVMRustAddBuilderLibraryInfo(LLVMPassManagerBuilderRef PMBR,
|
2016-12-30 12:22:11 +01:00
|
|
|
LLVMModuleRef M,
|
|
|
|
bool DisableSimplifyLibCalls) {
|
|
|
|
Triple TargetTriple(unwrap(M)->getTargetTriple());
|
|
|
|
TargetLibraryInfoImpl *TLI = new TargetLibraryInfoImpl(TargetTriple);
|
|
|
|
if (DisableSimplifyLibCalls)
|
|
|
|
TLI->disableAllFunctions();
|
2016-12-31 18:01:23 +01:00
|
|
|
unwrap(PMBR)->LibraryInfo = TLI;
|
2013-08-23 05:58:42 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
// Unfortunately, the LLVM C API doesn't provide a way to create the
|
|
|
|
// TargetLibraryInfo pass, so we use this method to do so.
|
2016-12-31 18:01:23 +01:00
|
|
|
extern "C" void LLVMRustAddLibraryInfo(LLVMPassManagerRef PMR, LLVMModuleRef M,
|
2016-12-30 12:22:11 +01:00
|
|
|
bool DisableSimplifyLibCalls) {
|
|
|
|
Triple TargetTriple(unwrap(M)->getTargetTriple());
|
|
|
|
TargetLibraryInfoImpl TLII(TargetTriple);
|
|
|
|
if (DisableSimplifyLibCalls)
|
|
|
|
TLII.disableAllFunctions();
|
2016-12-31 18:01:23 +01:00
|
|
|
unwrap(PMR)->add(new TargetLibraryInfoWrapperPass(TLII));
|
2013-08-23 05:58:42 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
// Unfortunately, the LLVM C API doesn't provide an easy way of iterating over
|
|
|
|
// all the functions in a module, so we do that manually here. You'll find
|
|
|
|
// similar code in clang's BackendUtil.cpp file.
|
2016-12-31 18:01:23 +01:00
|
|
|
extern "C" void LLVMRustRunFunctionPassManager(LLVMPassManagerRef PMR,
|
2016-12-30 12:22:11 +01:00
|
|
|
LLVMModuleRef M) {
|
|
|
|
llvm::legacy::FunctionPassManager *P =
|
2016-12-31 18:01:23 +01:00
|
|
|
unwrap<llvm::legacy::FunctionPassManager>(PMR);
|
2016-12-30 12:22:11 +01:00
|
|
|
P->doInitialization();
|
|
|
|
|
|
|
|
// Upgrade all calls to old intrinsics first.
|
|
|
|
for (Module::iterator I = unwrap(M)->begin(), E = unwrap(M)->end(); I != E;)
|
|
|
|
UpgradeCallsToIntrinsic(&*I++); // must be post-increment, as we remove
|
|
|
|
|
|
|
|
for (Module::iterator I = unwrap(M)->begin(), E = unwrap(M)->end(); I != E;
|
|
|
|
++I)
|
|
|
|
if (!I->isDeclaration())
|
|
|
|
P->run(*I);
|
|
|
|
|
|
|
|
P->doFinalization();
|
2013-08-23 05:58:42 +02:00
|
|
|
}
|
|
|
|
|
2016-12-30 12:22:11 +01:00
|
|
|
extern "C" void LLVMRustSetLLVMOptions(int Argc, char **Argv) {
|
|
|
|
// Initializing the command-line options more than once is not allowed. So,
|
|
|
|
// check if they've already been initialized. (This could happen if we're
|
|
|
|
// being called from rustpkg, for example). If the arguments change, then
|
|
|
|
// that's just kinda unfortunate.
|
2016-12-31 18:01:23 +01:00
|
|
|
static bool Initialized = false;
|
|
|
|
if (Initialized)
|
2016-12-30 12:22:11 +01:00
|
|
|
return;
|
2016-12-31 18:01:23 +01:00
|
|
|
Initialized = true;
|
2016-12-30 12:22:11 +01:00
|
|
|
cl::ParseCommandLineOptions(Argc, Argv);
|
2013-08-23 05:58:42 +02:00
|
|
|
}
|
|
|
|
|
2016-08-01 23:16:16 +02:00
|
|
|
enum class LLVMRustFileType {
|
2016-12-30 12:22:11 +01:00
|
|
|
AssemblyFile,
|
|
|
|
ObjectFile,
|
2016-08-01 23:16:16 +02:00
|
|
|
};
|
|
|
|
|
2019-12-31 14:05:34 +01:00
|
|
|
#if LLVM_VERSION_GE(10, 0)
|
|
|
|
static CodeGenFileType fromRust(LLVMRustFileType Type) {
|
|
|
|
switch (Type) {
|
|
|
|
case LLVMRustFileType::AssemblyFile:
|
|
|
|
return CGFT_AssemblyFile;
|
|
|
|
case LLVMRustFileType::ObjectFile:
|
|
|
|
return CGFT_ObjectFile;
|
|
|
|
default:
|
|
|
|
report_fatal_error("Bad FileType.");
|
|
|
|
}
|
|
|
|
}
|
|
|
|
#else
|
2016-12-31 18:01:23 +01:00
|
|
|
static TargetMachine::CodeGenFileType fromRust(LLVMRustFileType Type) {
|
|
|
|
switch (Type) {
|
2016-12-30 12:22:11 +01:00
|
|
|
case LLVMRustFileType::AssemblyFile:
|
|
|
|
return TargetMachine::CGFT_AssemblyFile;
|
|
|
|
case LLVMRustFileType::ObjectFile:
|
|
|
|
return TargetMachine::CGFT_ObjectFile;
|
|
|
|
default:
|
2017-11-20 17:47:29 +01:00
|
|
|
report_fatal_error("Bad FileType.");
|
2016-08-01 23:16:16 +02:00
|
|
|
}
|
|
|
|
}
|
2019-12-31 14:05:34 +01:00
|
|
|
#endif
|
2016-08-01 23:16:16 +02:00
|
|
|
|
|
|
|
extern "C" LLVMRustResult
|
2016-12-30 12:22:11 +01:00
|
|
|
LLVMRustWriteOutputFile(LLVMTargetMachineRef Target, LLVMPassManagerRef PMR,
|
2020-09-23 17:25:20 +02:00
|
|
|
LLVMModuleRef M, const char *Path, const char *DwoPath,
|
2016-12-31 18:01:23 +01:00
|
|
|
LLVMRustFileType RustFileType) {
|
2016-07-13 00:42:20 +02:00
|
|
|
llvm::legacy::PassManager *PM = unwrap<llvm::legacy::PassManager>(PMR);
|
2016-12-31 18:01:23 +01:00
|
|
|
auto FileType = fromRust(RustFileType);
|
2013-08-23 05:58:42 +02:00
|
|
|
|
|
|
|
std::string ErrorInfo;
|
2014-09-30 23:20:22 +02:00
|
|
|
std::error_code EC;
|
2016-12-31 18:01:23 +01:00
|
|
|
raw_fd_ostream OS(Path, EC, sys::fs::F_None);
|
2014-09-30 23:20:22 +02:00
|
|
|
if (EC)
|
|
|
|
ErrorInfo = EC.message();
|
2013-08-23 05:58:42 +02:00
|
|
|
if (ErrorInfo != "") {
|
2014-04-15 16:25:22 +02:00
|
|
|
LLVMRustSetLastError(ErrorInfo.c_str());
|
2016-08-01 23:16:16 +02:00
|
|
|
return LLVMRustResult::Failure;
|
2013-08-23 05:58:42 +02:00
|
|
|
}
|
|
|
|
|
2018-08-20 08:44:25 +02:00
|
|
|
buffer_ostream BOS(OS);
|
2020-09-23 17:25:20 +02:00
|
|
|
if (DwoPath) {
|
|
|
|
raw_fd_ostream DOS(DwoPath, EC, sys::fs::F_None);
|
|
|
|
EC.clear();
|
|
|
|
if (EC)
|
|
|
|
ErrorInfo = EC.message();
|
|
|
|
if (ErrorInfo != "") {
|
|
|
|
LLVMRustSetLastError(ErrorInfo.c_str());
|
|
|
|
return LLVMRustResult::Failure;
|
|
|
|
}
|
|
|
|
buffer_ostream DBOS(DOS);
|
|
|
|
unwrap(Target)->addPassesToEmitFile(*PM, BOS, &DBOS, FileType, false);
|
|
|
|
PM->run(*unwrap(M));
|
|
|
|
} else {
|
|
|
|
unwrap(Target)->addPassesToEmitFile(*PM, BOS, nullptr, FileType, false);
|
|
|
|
PM->run(*unwrap(M));
|
|
|
|
}
|
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`.
2015-05-14 21:10:43 +02:00
|
|
|
|
2015-10-08 00:11:25 +02:00
|
|
|
// Apparently `addPassesToEmitFile` adds a pointer to our on-the-stack output
|
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`.
2015-05-14 21:10:43 +02:00
|
|
|
// stream (OS), so the only real safe place to delete this is here? Don't we
|
|
|
|
// wish this was written in Rust?
|
2019-11-25 21:44:26 +01:00
|
|
|
LLVMDisposePassManager(PMR);
|
2016-08-01 23:16:16 +02:00
|
|
|
return LLVMRustResult::Success;
|
2013-08-23 05:58:42 +02:00
|
|
|
}
|
|
|
|
|
2020-02-11 22:37:16 +01:00
|
|
|
extern "C" typedef void (*LLVMRustSelfProfileBeforePassCallback)(void*, // LlvmSelfProfiler
|
|
|
|
const char*, // pass name
|
|
|
|
const char*); // IR name
|
|
|
|
extern "C" typedef void (*LLVMRustSelfProfileAfterPassCallback)(void*); // LlvmSelfProfiler
|
|
|
|
|
|
|
|
std::string LLVMRustwrappedIrGetName(const llvm::Any &WrappedIr) {
|
|
|
|
if (any_isa<const Module *>(WrappedIr))
|
|
|
|
return any_cast<const Module *>(WrappedIr)->getName().str();
|
|
|
|
if (any_isa<const Function *>(WrappedIr))
|
|
|
|
return any_cast<const Function *>(WrappedIr)->getName().str();
|
|
|
|
if (any_isa<const Loop *>(WrappedIr))
|
|
|
|
return any_cast<const Loop *>(WrappedIr)->getName().str();
|
|
|
|
if (any_isa<const LazyCallGraph::SCC *>(WrappedIr))
|
|
|
|
return any_cast<const LazyCallGraph::SCC *>(WrappedIr)->getName();
|
|
|
|
return "<UNKNOWN>";
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
void LLVMSelfProfileInitializeCallbacks(
|
|
|
|
PassInstrumentationCallbacks& PIC, void* LlvmSelfProfiler,
|
|
|
|
LLVMRustSelfProfileBeforePassCallback BeforePassCallback,
|
|
|
|
LLVMRustSelfProfileAfterPassCallback AfterPassCallback) {
|
|
|
|
PIC.registerBeforePassCallback([LlvmSelfProfiler, BeforePassCallback](
|
|
|
|
StringRef Pass, llvm::Any Ir) {
|
|
|
|
std::string PassName = Pass.str();
|
|
|
|
std::string IrName = LLVMRustwrappedIrGetName(Ir);
|
|
|
|
BeforePassCallback(LlvmSelfProfiler, PassName.c_str(), IrName.c_str());
|
|
|
|
return true;
|
|
|
|
});
|
|
|
|
|
|
|
|
PIC.registerAfterPassCallback(
|
|
|
|
[LlvmSelfProfiler, AfterPassCallback](StringRef Pass, llvm::Any Ir) {
|
|
|
|
AfterPassCallback(LlvmSelfProfiler);
|
|
|
|
});
|
|
|
|
|
|
|
|
PIC.registerAfterPassInvalidatedCallback(
|
|
|
|
[LlvmSelfProfiler, AfterPassCallback](StringRef Pass) {
|
|
|
|
AfterPassCallback(LlvmSelfProfiler);
|
|
|
|
});
|
|
|
|
|
|
|
|
PIC.registerBeforeAnalysisCallback([LlvmSelfProfiler, BeforePassCallback](
|
|
|
|
StringRef Pass, llvm::Any Ir) {
|
|
|
|
std::string PassName = Pass.str();
|
|
|
|
std::string IrName = LLVMRustwrappedIrGetName(Ir);
|
|
|
|
BeforePassCallback(LlvmSelfProfiler, PassName.c_str(), IrName.c_str());
|
|
|
|
});
|
|
|
|
|
|
|
|
PIC.registerAfterAnalysisCallback(
|
|
|
|
[LlvmSelfProfiler, AfterPassCallback](StringRef Pass, llvm::Any Ir) {
|
|
|
|
AfterPassCallback(LlvmSelfProfiler);
|
|
|
|
});
|
|
|
|
}
|
|
|
|
|
2020-01-05 19:16:58 +01:00
|
|
|
enum class LLVMRustOptStage {
|
|
|
|
PreLinkNoLTO,
|
|
|
|
PreLinkThinLTO,
|
|
|
|
PreLinkFatLTO,
|
|
|
|
ThinLTO,
|
|
|
|
FatLTO,
|
|
|
|
};
|
|
|
|
|
|
|
|
struct LLVMRustSanitizerOptions {
|
2020-06-14 02:00:00 +02:00
|
|
|
bool SanitizeAddress;
|
|
|
|
bool SanitizeAddressRecover;
|
2020-01-05 19:16:58 +01:00
|
|
|
bool SanitizeMemory;
|
2020-06-14 02:00:00 +02:00
|
|
|
bool SanitizeMemoryRecover;
|
|
|
|
int SanitizeMemoryTrackOrigins;
|
2020-01-05 19:16:58 +01:00
|
|
|
bool SanitizeThread;
|
|
|
|
};
|
|
|
|
|
|
|
|
extern "C" void
|
|
|
|
LLVMRustOptimizeWithNewPassManager(
|
|
|
|
LLVMModuleRef ModuleRef,
|
|
|
|
LLVMTargetMachineRef TMRef,
|
|
|
|
LLVMRustPassBuilderOptLevel OptLevelRust,
|
|
|
|
LLVMRustOptStage OptStage,
|
|
|
|
bool NoPrepopulatePasses, bool VerifyIR, bool UseThinLTOBuffers,
|
|
|
|
bool MergeFunctions, bool UnrollLoops, bool SLPVectorize, bool LoopVectorize,
|
2020-05-13 02:00:00 +02:00
|
|
|
bool DisableSimplifyLibCalls, bool EmitLifetimeMarkers,
|
2020-01-05 19:16:58 +01:00
|
|
|
LLVMRustSanitizerOptions *SanitizerOptions,
|
2020-02-11 22:37:16 +01:00
|
|
|
const char *PGOGenPath, const char *PGOUsePath,
|
|
|
|
void* LlvmSelfProfiler,
|
|
|
|
LLVMRustSelfProfileBeforePassCallback BeforePassCallback,
|
|
|
|
LLVMRustSelfProfileAfterPassCallback AfterPassCallback) {
|
2020-01-05 19:16:58 +01:00
|
|
|
Module *TheModule = unwrap(ModuleRef);
|
|
|
|
TargetMachine *TM = unwrap(TMRef);
|
|
|
|
PassBuilder::OptimizationLevel OptLevel = fromRust(OptLevelRust);
|
|
|
|
|
|
|
|
// FIXME: MergeFunctions is not supported by NewPM yet.
|
|
|
|
(void) MergeFunctions;
|
|
|
|
|
|
|
|
PipelineTuningOptions PTO;
|
|
|
|
PTO.LoopUnrolling = UnrollLoops;
|
|
|
|
PTO.LoopInterleaving = UnrollLoops;
|
|
|
|
PTO.LoopVectorization = LoopVectorize;
|
|
|
|
PTO.SLPVectorization = SLPVectorize;
|
|
|
|
|
|
|
|
PassInstrumentationCallbacks PIC;
|
|
|
|
StandardInstrumentations SI;
|
|
|
|
SI.registerCallbacks(PIC);
|
|
|
|
|
2020-02-11 22:37:16 +01:00
|
|
|
if (LlvmSelfProfiler){
|
|
|
|
LLVMSelfProfileInitializeCallbacks(PIC,LlvmSelfProfiler,BeforePassCallback,AfterPassCallback);
|
|
|
|
}
|
|
|
|
|
2020-01-05 19:16:58 +01:00
|
|
|
Optional<PGOOptions> PGOOpt;
|
|
|
|
if (PGOGenPath) {
|
|
|
|
assert(!PGOUsePath);
|
|
|
|
PGOOpt = PGOOptions(PGOGenPath, "", "", PGOOptions::IRInstr);
|
|
|
|
} else if (PGOUsePath) {
|
|
|
|
assert(!PGOGenPath);
|
|
|
|
PGOOpt = PGOOptions(PGOUsePath, "", "", PGOOptions::IRUse);
|
|
|
|
}
|
|
|
|
|
|
|
|
PassBuilder PB(TM, PTO, PGOOpt, &PIC);
|
|
|
|
|
|
|
|
// FIXME: We may want to expose this as an option.
|
|
|
|
bool DebugPassManager = false;
|
|
|
|
LoopAnalysisManager LAM(DebugPassManager);
|
|
|
|
FunctionAnalysisManager FAM(DebugPassManager);
|
|
|
|
CGSCCAnalysisManager CGAM(DebugPassManager);
|
|
|
|
ModuleAnalysisManager MAM(DebugPassManager);
|
|
|
|
|
|
|
|
FAM.registerPass([&] { return PB.buildDefaultAAPipeline(); });
|
|
|
|
|
|
|
|
Triple TargetTriple(TheModule->getTargetTriple());
|
|
|
|
std::unique_ptr<TargetLibraryInfoImpl> TLII(new TargetLibraryInfoImpl(TargetTriple));
|
|
|
|
if (DisableSimplifyLibCalls)
|
|
|
|
TLII->disableAllFunctions();
|
|
|
|
FAM.registerPass([&] { return TargetLibraryAnalysis(*TLII); });
|
|
|
|
|
|
|
|
PB.registerModuleAnalyses(MAM);
|
|
|
|
PB.registerCGSCCAnalyses(CGAM);
|
|
|
|
PB.registerFunctionAnalyses(FAM);
|
|
|
|
PB.registerLoopAnalyses(LAM);
|
|
|
|
PB.crossRegisterProxies(LAM, FAM, CGAM, MAM);
|
|
|
|
|
|
|
|
// We manually collect pipeline callbacks so we can apply them at O0, where the
|
|
|
|
// PassBuilder does not create a pipeline.
|
|
|
|
std::vector<std::function<void(ModulePassManager &)>> PipelineStartEPCallbacks;
|
2020-06-26 03:52:41 +02:00
|
|
|
#if LLVM_VERSION_GE(11, 0)
|
|
|
|
std::vector<std::function<void(ModulePassManager &, PassBuilder::OptimizationLevel)>>
|
|
|
|
OptimizerLastEPCallbacks;
|
|
|
|
#else
|
2020-01-05 19:16:58 +01:00
|
|
|
std::vector<std::function<void(FunctionPassManager &, PassBuilder::OptimizationLevel)>>
|
|
|
|
OptimizerLastEPCallbacks;
|
2020-06-26 03:52:41 +02:00
|
|
|
#endif
|
2020-01-05 19:16:58 +01:00
|
|
|
|
|
|
|
if (VerifyIR) {
|
|
|
|
PipelineStartEPCallbacks.push_back([VerifyIR](ModulePassManager &MPM) {
|
|
|
|
MPM.addPass(VerifierPass());
|
|
|
|
});
|
|
|
|
}
|
|
|
|
|
|
|
|
if (SanitizerOptions) {
|
|
|
|
if (SanitizerOptions->SanitizeMemory) {
|
|
|
|
MemorySanitizerOptions Options(
|
|
|
|
SanitizerOptions->SanitizeMemoryTrackOrigins,
|
2020-06-14 02:00:00 +02:00
|
|
|
SanitizerOptions->SanitizeMemoryRecover,
|
2020-01-05 19:16:58 +01:00
|
|
|
/*CompileKernel=*/false);
|
2020-06-26 03:52:41 +02:00
|
|
|
#if LLVM_VERSION_GE(11, 0)
|
|
|
|
OptimizerLastEPCallbacks.push_back(
|
|
|
|
[Options](ModulePassManager &MPM, PassBuilder::OptimizationLevel Level) {
|
|
|
|
MPM.addPass(MemorySanitizerPass(Options));
|
|
|
|
MPM.addPass(createModuleToFunctionPassAdaptor(MemorySanitizerPass(Options)));
|
|
|
|
}
|
|
|
|
);
|
|
|
|
#else
|
2020-01-05 19:16:58 +01:00
|
|
|
#if LLVM_VERSION_GE(10, 0)
|
|
|
|
PipelineStartEPCallbacks.push_back([Options](ModulePassManager &MPM) {
|
|
|
|
MPM.addPass(MemorySanitizerPass(Options));
|
|
|
|
});
|
|
|
|
#endif
|
|
|
|
OptimizerLastEPCallbacks.push_back(
|
|
|
|
[Options](FunctionPassManager &FPM, PassBuilder::OptimizationLevel Level) {
|
|
|
|
FPM.addPass(MemorySanitizerPass(Options));
|
|
|
|
}
|
|
|
|
);
|
2020-06-26 03:52:41 +02:00
|
|
|
#endif
|
2020-01-05 19:16:58 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
if (SanitizerOptions->SanitizeThread) {
|
2020-06-26 03:52:41 +02:00
|
|
|
#if LLVM_VERSION_GE(11, 0)
|
|
|
|
OptimizerLastEPCallbacks.push_back(
|
|
|
|
[](ModulePassManager &MPM, PassBuilder::OptimizationLevel Level) {
|
|
|
|
MPM.addPass(ThreadSanitizerPass());
|
|
|
|
MPM.addPass(createModuleToFunctionPassAdaptor(ThreadSanitizerPass()));
|
|
|
|
}
|
|
|
|
);
|
|
|
|
#else
|
2020-01-05 19:16:58 +01:00
|
|
|
#if LLVM_VERSION_GE(10, 0)
|
|
|
|
PipelineStartEPCallbacks.push_back([](ModulePassManager &MPM) {
|
|
|
|
MPM.addPass(ThreadSanitizerPass());
|
|
|
|
});
|
|
|
|
#endif
|
|
|
|
OptimizerLastEPCallbacks.push_back(
|
|
|
|
[](FunctionPassManager &FPM, PassBuilder::OptimizationLevel Level) {
|
|
|
|
FPM.addPass(ThreadSanitizerPass());
|
|
|
|
}
|
|
|
|
);
|
2020-06-26 03:52:41 +02:00
|
|
|
#endif
|
2020-01-05 19:16:58 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
if (SanitizerOptions->SanitizeAddress) {
|
2020-06-26 03:52:41 +02:00
|
|
|
#if LLVM_VERSION_GE(11, 0)
|
|
|
|
OptimizerLastEPCallbacks.push_back(
|
|
|
|
[SanitizerOptions](ModulePassManager &MPM, PassBuilder::OptimizationLevel Level) {
|
|
|
|
MPM.addPass(RequireAnalysisPass<ASanGlobalsMetadataAnalysis, Module>());
|
|
|
|
MPM.addPass(ModuleAddressSanitizerPass(
|
|
|
|
/*CompileKernel=*/false, SanitizerOptions->SanitizeAddressRecover));
|
|
|
|
MPM.addPass(createModuleToFunctionPassAdaptor(AddressSanitizerPass(
|
|
|
|
/*CompileKernel=*/false, SanitizerOptions->SanitizeAddressRecover,
|
|
|
|
/*UseAfterScope=*/true)));
|
|
|
|
}
|
|
|
|
);
|
|
|
|
#else
|
2020-01-05 19:16:58 +01:00
|
|
|
PipelineStartEPCallbacks.push_back([&](ModulePassManager &MPM) {
|
|
|
|
MPM.addPass(RequireAnalysisPass<ASanGlobalsMetadataAnalysis, Module>());
|
|
|
|
});
|
|
|
|
OptimizerLastEPCallbacks.push_back(
|
|
|
|
[SanitizerOptions](FunctionPassManager &FPM, PassBuilder::OptimizationLevel Level) {
|
|
|
|
FPM.addPass(AddressSanitizerPass(
|
2020-06-14 02:00:00 +02:00
|
|
|
/*CompileKernel=*/false, SanitizerOptions->SanitizeAddressRecover,
|
2020-02-13 01:00:00 +01:00
|
|
|
/*UseAfterScope=*/true));
|
2020-01-05 19:16:58 +01:00
|
|
|
}
|
|
|
|
);
|
|
|
|
PipelineStartEPCallbacks.push_back(
|
|
|
|
[SanitizerOptions](ModulePassManager &MPM) {
|
|
|
|
MPM.addPass(ModuleAddressSanitizerPass(
|
2020-06-14 02:00:00 +02:00
|
|
|
/*CompileKernel=*/false, SanitizerOptions->SanitizeAddressRecover));
|
2020-01-05 19:16:58 +01:00
|
|
|
}
|
|
|
|
);
|
2020-06-26 03:52:41 +02:00
|
|
|
#endif
|
2020-01-05 19:16:58 +01:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
ModulePassManager MPM(DebugPassManager);
|
|
|
|
if (!NoPrepopulatePasses) {
|
2020-06-26 03:52:41 +02:00
|
|
|
if (OptLevel == PassBuilder::OptimizationLevel::O0) {
|
2020-01-05 19:16:58 +01:00
|
|
|
for (const auto &C : PipelineStartEPCallbacks)
|
|
|
|
C(MPM);
|
|
|
|
|
2020-06-26 03:52:41 +02:00
|
|
|
#if LLVM_VERSION_GE(11, 0)
|
|
|
|
for (const auto &C : OptimizerLastEPCallbacks)
|
|
|
|
C(MPM, OptLevel);
|
|
|
|
#else
|
2020-01-05 19:16:58 +01:00
|
|
|
if (!OptimizerLastEPCallbacks.empty()) {
|
|
|
|
FunctionPassManager FPM(DebugPassManager);
|
|
|
|
for (const auto &C : OptimizerLastEPCallbacks)
|
|
|
|
C(FPM, OptLevel);
|
|
|
|
MPM.addPass(createModuleToFunctionPassAdaptor(std::move(FPM)));
|
|
|
|
}
|
2020-06-26 03:52:41 +02:00
|
|
|
#endif
|
2020-01-05 19:16:58 +01:00
|
|
|
|
2020-05-13 02:00:00 +02:00
|
|
|
MPM.addPass(AlwaysInlinerPass(EmitLifetimeMarkers));
|
2020-01-05 19:16:58 +01:00
|
|
|
|
|
|
|
#if LLVM_VERSION_GE(10, 0)
|
|
|
|
if (PGOOpt) {
|
|
|
|
PB.addPGOInstrPassesForO0(
|
|
|
|
MPM, DebugPassManager, PGOOpt->Action == PGOOptions::IRInstr,
|
|
|
|
/*IsCS=*/false, PGOOpt->ProfileFile, PGOOpt->ProfileRemappingFile);
|
|
|
|
}
|
|
|
|
#endif
|
|
|
|
} else {
|
|
|
|
for (const auto &C : PipelineStartEPCallbacks)
|
|
|
|
PB.registerPipelineStartEPCallback(C);
|
2020-03-03 01:00:00 +01:00
|
|
|
if (OptStage != LLVMRustOptStage::PreLinkThinLTO) {
|
|
|
|
for (const auto &C : OptimizerLastEPCallbacks)
|
|
|
|
PB.registerOptimizerLastEPCallback(C);
|
|
|
|
}
|
2020-01-05 19:16:58 +01:00
|
|
|
|
|
|
|
switch (OptStage) {
|
|
|
|
case LLVMRustOptStage::PreLinkNoLTO:
|
|
|
|
MPM = PB.buildPerModuleDefaultPipeline(OptLevel, DebugPassManager);
|
|
|
|
break;
|
|
|
|
case LLVMRustOptStage::PreLinkThinLTO:
|
|
|
|
MPM = PB.buildThinLTOPreLinkDefaultPipeline(OptLevel, DebugPassManager);
|
2020-06-26 03:52:41 +02:00
|
|
|
#if LLVM_VERSION_GE(11, 0)
|
|
|
|
for (const auto &C : OptimizerLastEPCallbacks)
|
|
|
|
C(MPM, OptLevel);
|
|
|
|
#else
|
2020-03-03 01:00:00 +01:00
|
|
|
if (!OptimizerLastEPCallbacks.empty()) {
|
|
|
|
FunctionPassManager FPM(DebugPassManager);
|
|
|
|
for (const auto &C : OptimizerLastEPCallbacks)
|
|
|
|
C(FPM, OptLevel);
|
|
|
|
MPM.addPass(createModuleToFunctionPassAdaptor(std::move(FPM)));
|
|
|
|
}
|
2020-06-26 03:52:41 +02:00
|
|
|
#endif
|
2020-01-05 19:16:58 +01:00
|
|
|
break;
|
|
|
|
case LLVMRustOptStage::PreLinkFatLTO:
|
|
|
|
MPM = PB.buildLTOPreLinkDefaultPipeline(OptLevel, DebugPassManager);
|
|
|
|
break;
|
|
|
|
case LLVMRustOptStage::ThinLTO:
|
|
|
|
// FIXME: Does it make sense to pass the ModuleSummaryIndex?
|
|
|
|
// It only seems to be needed for C++ specific optimizations.
|
|
|
|
MPM = PB.buildThinLTODefaultPipeline(OptLevel, DebugPassManager, nullptr);
|
|
|
|
break;
|
|
|
|
case LLVMRustOptStage::FatLTO:
|
|
|
|
MPM = PB.buildLTODefaultPipeline(OptLevel, DebugPassManager, nullptr);
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
if (UseThinLTOBuffers) {
|
|
|
|
MPM.addPass(CanonicalizeAliasesPass());
|
|
|
|
MPM.addPass(NameAnonGlobalPass());
|
|
|
|
}
|
|
|
|
|
|
|
|
// Upgrade all calls to old intrinsics first.
|
|
|
|
for (Module::iterator I = TheModule->begin(), E = TheModule->end(); I != E;)
|
|
|
|
UpgradeCallsToIntrinsic(&*I++); // must be post-increment, as we remove
|
|
|
|
|
|
|
|
MPM.run(*TheModule, MAM);
|
|
|
|
}
|
2017-06-29 16:52:43 +02:00
|
|
|
|
|
|
|
// Callback to demangle function name
|
|
|
|
// Parameters:
|
|
|
|
// * name to be demangled
|
|
|
|
// * name len
|
|
|
|
// * output buffer
|
|
|
|
// * output buffer len
|
|
|
|
// Returns len of demangled string, or 0 if demangle failed.
|
|
|
|
typedef size_t (*DemangleFn)(const char*, size_t, char*, size_t);
|
|
|
|
|
|
|
|
|
|
|
|
namespace {
|
|
|
|
|
|
|
|
class RustAssemblyAnnotationWriter : public AssemblyAnnotationWriter {
|
|
|
|
DemangleFn Demangle;
|
|
|
|
std::vector<char> Buf;
|
|
|
|
|
|
|
|
public:
|
|
|
|
RustAssemblyAnnotationWriter(DemangleFn Demangle) : Demangle(Demangle) {}
|
|
|
|
|
|
|
|
// Return empty string if demangle failed
|
|
|
|
// or if name does not need to be demangled
|
|
|
|
StringRef CallDemangle(StringRef name) {
|
|
|
|
if (!Demangle) {
|
|
|
|
return StringRef();
|
|
|
|
}
|
|
|
|
|
|
|
|
if (Buf.size() < name.size() * 2) {
|
|
|
|
// Semangled name usually shorter than mangled,
|
|
|
|
// but allocate twice as much memory just in case
|
|
|
|
Buf.resize(name.size() * 2);
|
|
|
|
}
|
|
|
|
|
|
|
|
auto R = Demangle(name.data(), name.size(), Buf.data(), Buf.size());
|
|
|
|
if (!R) {
|
|
|
|
// Demangle failed.
|
|
|
|
return StringRef();
|
|
|
|
}
|
|
|
|
|
|
|
|
auto Demangled = StringRef(Buf.data(), R);
|
|
|
|
if (Demangled == name) {
|
|
|
|
// Do not print anything if demangled name is equal to mangled.
|
|
|
|
return StringRef();
|
|
|
|
}
|
|
|
|
|
|
|
|
return Demangled;
|
|
|
|
}
|
|
|
|
|
|
|
|
void emitFunctionAnnot(const Function *F,
|
|
|
|
formatted_raw_ostream &OS) override {
|
|
|
|
StringRef Demangled = CallDemangle(F->getName());
|
|
|
|
if (Demangled.empty()) {
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
OS << "; " << Demangled << "\n";
|
|
|
|
}
|
|
|
|
|
|
|
|
void emitInstructionAnnot(const Instruction *I,
|
|
|
|
formatted_raw_ostream &OS) override {
|
|
|
|
const char *Name;
|
|
|
|
const Value *Value;
|
|
|
|
if (const CallInst *CI = dyn_cast<CallInst>(I)) {
|
|
|
|
Name = "call";
|
2020-06-26 03:52:41 +02:00
|
|
|
Value = CI->getCalledOperand();
|
2017-06-29 16:52:43 +02:00
|
|
|
} else if (const InvokeInst* II = dyn_cast<InvokeInst>(I)) {
|
|
|
|
Name = "invoke";
|
2020-06-26 03:52:41 +02:00
|
|
|
Value = II->getCalledOperand();
|
2017-06-29 16:52:43 +02:00
|
|
|
} else {
|
|
|
|
// Could demangle more operations, e. g.
|
|
|
|
// `store %place, @function`.
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (!Value->hasName()) {
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
StringRef Demangled = CallDemangle(Value->getName());
|
|
|
|
if (Demangled.empty()) {
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
OS << "; " << Name << " " << Demangled << "\n";
|
|
|
|
}
|
|
|
|
};
|
|
|
|
|
|
|
|
} // namespace
|
|
|
|
|
2019-04-06 02:48:23 +02:00
|
|
|
extern "C" LLVMRustResult
|
2019-12-02 10:53:01 +01:00
|
|
|
LLVMRustPrintModule(LLVMModuleRef M, const char *Path, DemangleFn Demangle) {
|
2013-08-23 05:58:42 +02:00
|
|
|
std::string ErrorInfo;
|
2014-09-30 23:20:22 +02:00
|
|
|
std::error_code EC;
|
2016-12-31 18:01:23 +01:00
|
|
|
raw_fd_ostream OS(Path, EC, sys::fs::F_None);
|
2014-09-30 23:20:22 +02:00
|
|
|
if (EC)
|
|
|
|
ErrorInfo = EC.message();
|
2019-04-06 02:48:23 +02:00
|
|
|
if (ErrorInfo != "") {
|
|
|
|
LLVMRustSetLastError(ErrorInfo.c_str());
|
|
|
|
return LLVMRustResult::Failure;
|
|
|
|
}
|
2014-02-26 23:06:27 +01:00
|
|
|
|
2019-12-02 10:53:01 +01:00
|
|
|
RustAssemblyAnnotationWriter AAW(Demangle);
|
2013-08-23 05:58:42 +02:00
|
|
|
formatted_raw_ostream FOS(OS);
|
2019-12-02 10:53:01 +01:00
|
|
|
unwrap(M)->print(FOS, &AAW);
|
2019-04-06 02:48:23 +02:00
|
|
|
|
|
|
|
return LLVMRustResult::Success;
|
2013-08-23 05:58:42 +02:00
|
|
|
}
|
|
|
|
|
2016-12-30 12:22:11 +01:00
|
|
|
extern "C" void LLVMRustPrintPasses() {
|
|
|
|
LLVMInitializePasses();
|
|
|
|
struct MyListener : PassRegistrationListener {
|
2016-12-31 18:01:23 +01:00
|
|
|
void passEnumerate(const PassInfo *Info) {
|
|
|
|
StringRef PassArg = Info->getPassArgument();
|
|
|
|
StringRef PassName = Info->getPassName();
|
2016-12-30 12:22:11 +01:00
|
|
|
if (!PassArg.empty()) {
|
|
|
|
// These unsigned->signed casts could theoretically overflow, but
|
|
|
|
// realistically never will (and even if, the result is implementation
|
|
|
|
// defined rather plain UB).
|
|
|
|
printf("%15.*s - %.*s\n", (int)PassArg.size(), PassArg.data(),
|
|
|
|
(int)PassName.size(), PassName.data());
|
|
|
|
}
|
|
|
|
}
|
2016-12-31 18:01:23 +01:00
|
|
|
} Listener;
|
2013-08-23 05:58:42 +02:00
|
|
|
|
2016-12-30 12:22:11 +01:00
|
|
|
PassRegistry *PR = PassRegistry::getPassRegistry();
|
2016-12-31 18:01:23 +01:00
|
|
|
PR->enumerateWith(&Listener);
|
2013-05-28 01:15:31 +02:00
|
|
|
}
|
2013-06-20 00:18:25 +02:00
|
|
|
|
2016-12-31 18:01:23 +01:00
|
|
|
extern "C" void LLVMRustAddAlwaysInlinePass(LLVMPassManagerBuilderRef PMBR,
|
2016-12-30 12:22:11 +01:00
|
|
|
bool AddLifetimes) {
|
2016-12-31 18:01:23 +01:00
|
|
|
unwrap(PMBR)->Inliner = llvm::createAlwaysInlinerLegacyPass(AddLifetimes);
|
2013-06-20 00:18:25 +02:00
|
|
|
}
|
Implement LTO
This commit implements LTO for rust leveraging LLVM's passes. What this means
is:
* When compiling an rlib, in addition to insdering foo.o into the archive, also
insert foo.bc (the LLVM bytecode) of the optimized module.
* When the compiler detects the -Z lto option, it will attempt to perform LTO on
a staticlib or binary output. The compiler will emit an error if a dylib or
rlib output is being generated.
* The actual act of performing LTO is as follows:
1. Force all upstream libraries to have an rlib version available.
2. Load the bytecode of each upstream library from the rlib.
3. Link all this bytecode into the current LLVM module (just using llvm
apis)
4. Run an internalization pass which internalizes all symbols except those
found reachable for the local crate of compilation.
5. Run the LLVM LTO pass manager over this entire module
6a. If assembling an archive, then add all upstream rlibs into the output
archive. This ignores all of the object/bitcode/metadata files rust
generated and placed inside the rlibs.
6b. If linking a binary, create copies of all upstream rlibs, remove the
rust-generated object-file, and then link everything as usual.
As I have explained in #10741, this process is excruciatingly slow, so this is
*not* turned on by default, and it is also why I have decided to hide it behind
a -Z flag for now. The good news is that the binary sizes are about as small as
they can be as a result of LTO, so it's definitely working.
Closes #10741
Closes #10740
2013-12-03 08:19:29 +01:00
|
|
|
|
2016-12-31 18:01:23 +01:00
|
|
|
extern "C" void LLVMRustRunRestrictionPass(LLVMModuleRef M, char **Symbols,
|
|
|
|
size_t Len) {
|
2016-12-30 12:22:11 +01:00
|
|
|
llvm::legacy::PassManager passes;
|
2016-07-13 00:42:44 +02:00
|
|
|
|
2016-12-30 12:22:11 +01:00
|
|
|
auto PreserveFunctions = [=](const GlobalValue &GV) {
|
2016-12-31 18:01:23 +01:00
|
|
|
for (size_t I = 0; I < Len; I++) {
|
|
|
|
if (GV.getName() == Symbols[I]) {
|
2016-12-30 12:22:11 +01:00
|
|
|
return true;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
return false;
|
|
|
|
};
|
2016-07-13 00:42:44 +02:00
|
|
|
|
2016-12-30 12:22:11 +01:00
|
|
|
passes.add(llvm::createInternalizePass(PreserveFunctions));
|
2016-07-13 00:42:44 +02:00
|
|
|
|
2016-12-30 12:22:11 +01:00
|
|
|
passes.run(*unwrap(M));
|
Implement LTO
This commit implements LTO for rust leveraging LLVM's passes. What this means
is:
* When compiling an rlib, in addition to insdering foo.o into the archive, also
insert foo.bc (the LLVM bytecode) of the optimized module.
* When the compiler detects the -Z lto option, it will attempt to perform LTO on
a staticlib or binary output. The compiler will emit an error if a dylib or
rlib output is being generated.
* The actual act of performing LTO is as follows:
1. Force all upstream libraries to have an rlib version available.
2. Load the bytecode of each upstream library from the rlib.
3. Link all this bytecode into the current LLVM module (just using llvm
apis)
4. Run an internalization pass which internalizes all symbols except those
found reachable for the local crate of compilation.
5. Run the LLVM LTO pass manager over this entire module
6a. If assembling an archive, then add all upstream rlibs into the output
archive. This ignores all of the object/bitcode/metadata files rust
generated and placed inside the rlibs.
6b. If linking a binary, create copies of all upstream rlibs, remove the
rust-generated object-file, and then link everything as usual.
As I have explained in #10741, this process is excruciatingly slow, so this is
*not* turned on by default, and it is also why I have decided to hide it behind
a -Z flag for now. The good news is that the binary sizes are about as small as
they can be as a result of LTO, so it's definitely working.
Closes #10741
Closes #10740
2013-12-03 08:19:29 +01:00
|
|
|
}
|
2013-12-11 08:27:15 +01:00
|
|
|
|
2016-12-30 12:22:11 +01:00
|
|
|
extern "C" void LLVMRustMarkAllFunctionsNounwind(LLVMModuleRef M) {
|
|
|
|
for (Module::iterator GV = unwrap(M)->begin(), E = unwrap(M)->end(); GV != E;
|
|
|
|
++GV) {
|
|
|
|
GV->setDoesNotThrow();
|
|
|
|
Function *F = dyn_cast<Function>(GV);
|
2016-12-30 13:21:21 +01:00
|
|
|
if (F == nullptr)
|
2016-12-30 12:22:11 +01:00
|
|
|
continue;
|
|
|
|
|
|
|
|
for (Function::iterator B = F->begin(), BE = F->end(); B != BE; ++B) {
|
|
|
|
for (BasicBlock::iterator I = B->begin(), IE = B->end(); I != IE; ++I) {
|
|
|
|
if (isa<InvokeInst>(I)) {
|
|
|
|
InvokeInst *CI = cast<InvokeInst>(I);
|
|
|
|
CI->setDoesNotThrow();
|
2013-12-11 08:27:15 +01:00
|
|
|
}
|
2016-12-30 12:22:11 +01:00
|
|
|
}
|
2013-12-11 08:27:15 +01:00
|
|
|
}
|
2016-12-30 12:22:11 +01:00
|
|
|
}
|
2013-12-11 08:27:15 +01:00
|
|
|
}
|
2015-07-17 00:48:16 +02:00
|
|
|
|
|
|
|
extern "C" void
|
|
|
|
LLVMRustSetDataLayoutFromTargetMachine(LLVMModuleRef Module,
|
|
|
|
LLVMTargetMachineRef TMR) {
|
2016-12-30 12:22:11 +01:00
|
|
|
TargetMachine *Target = unwrap(TMR);
|
|
|
|
unwrap(Module)->setDataLayout(Target->createDataLayout());
|
2015-07-17 00:48:16 +02:00
|
|
|
}
|
|
|
|
|
2019-11-03 17:52:00 +01:00
|
|
|
extern "C" void LLVMRustSetModulePICLevel(LLVMModuleRef M) {
|
|
|
|
unwrap(M)->setPICLevel(PICLevel::Level::BigPIC);
|
|
|
|
}
|
|
|
|
|
2016-12-30 12:22:11 +01:00
|
|
|
extern "C" void LLVMRustSetModulePIELevel(LLVMModuleRef M) {
|
|
|
|
unwrap(M)->setPIELevel(PIELevel::Level::Large);
|
2016-07-13 00:41:40 +02:00
|
|
|
}
|
rustc: Implement ThinLTO
This commit is an implementation of LLVM's ThinLTO for consumption in rustc
itself. Currently today LTO works by merging all relevant LLVM modules into one
and then running optimization passes. "Thin" LTO operates differently by having
more sharded work and allowing parallelism opportunities between optimizing
codegen units. Further down the road Thin LTO also allows *incremental* LTO
which should enable even faster release builds without compromising on the
performance we have today.
This commit uses a `-Z thinlto` flag to gate whether ThinLTO is enabled. It then
also implements two forms of ThinLTO:
* In one mode we'll *only* perform ThinLTO over the codegen units produced in a
single compilation. That is, we won't load upstream rlibs, but we'll instead
just perform ThinLTO amongst all codegen units produced by the compiler for
the local crate. This is intended to emulate a desired end point where we have
codegen units turned on by default for all crates and ThinLTO allows us to do
this without performance loss.
* In anther mode, like full LTO today, we'll optimize all upstream dependencies
in "thin" mode. Unlike today, however, this LTO step is fully parallelized so
should finish much more quickly.
There's a good bit of comments about what the implementation is doing and where
it came from, but the tl;dr; is that currently most of the support here is
copied from upstream LLVM. This code duplication is done for a number of
reasons:
* Controlling parallelism means we can use the existing jobserver support to
avoid overloading machines.
* We will likely want a slightly different form of incremental caching which
integrates with our own incremental strategy, but this is yet to be
determined.
* This buys us some flexibility about when/where we run ThinLTO, as well as
having it tailored to fit our needs for the time being.
* Finally this allows us to reuse some artifacts such as our `TargetMachine`
creation, where all our options we used today aren't necessarily supported by
upstream LLVM yet.
My hope is that we can get some experience with this copy/paste in tree and then
eventually upstream some work to LLVM itself to avoid the duplication while
still ensuring our needs are met. Otherwise I fear that maintaining these
bindings may be quite costly over the years with LLVM updates!
2017-07-23 17:14:38 +02:00
|
|
|
|
|
|
|
// Here you'll find an implementation of ThinLTO as used by the Rust compiler
|
|
|
|
// right now. This ThinLTO support is only enabled on "recent ish" versions of
|
|
|
|
// LLVM, and otherwise it's just blanket rejected from other compilers.
|
|
|
|
//
|
|
|
|
// Most of this implementation is straight copied from LLVM. At the time of
|
|
|
|
// this writing it wasn't *quite* suitable to reuse more code from upstream
|
|
|
|
// for our purposes, but we should strive to upstream this support once it's
|
|
|
|
// ready to go! I figure we may want a bit of testing locally first before
|
|
|
|
// sending this upstream to LLVM. I hear though they're quite eager to receive
|
|
|
|
// feedback like this!
|
|
|
|
//
|
|
|
|
// If you're reading this code and wondering "what in the world" or you're
|
|
|
|
// working "good lord by LLVM upgrade is *still* failing due to these bindings"
|
|
|
|
// then fear not! (ok maybe fear a little). All code here is mostly based
|
|
|
|
// on `lib/LTO/ThinLTOCodeGenerator.cpp` in LLVM.
|
|
|
|
//
|
|
|
|
// You'll find that the general layout here roughly corresponds to the `run`
|
|
|
|
// method in that file as well as `ProcessThinLTOModule`. Functions are
|
|
|
|
// specifically commented below as well, but if you're updating this code
|
|
|
|
// or otherwise trying to understand it, the LLVM source will be useful in
|
|
|
|
// interpreting the mysteries within.
|
|
|
|
//
|
|
|
|
// Otherwise I'll apologize in advance, it probably requires a relatively
|
|
|
|
// significant investment on your part to "truly understand" what's going on
|
|
|
|
// here. Not saying I do myself, but it took me awhile staring at LLVM's source
|
|
|
|
// and various online resources about ThinLTO to make heads or tails of all
|
|
|
|
// this.
|
|
|
|
|
|
|
|
// This is a shared data structure which *must* be threadsafe to share
|
|
|
|
// read-only amongst threads. This also corresponds basically to the arguments
|
|
|
|
// of the `ProcessThinLTOModule` function in the LLVM source.
|
|
|
|
struct LLVMRustThinLTOData {
|
|
|
|
// The combined index that is the global analysis over all modules we're
|
|
|
|
// performing ThinLTO for. This is mostly managed by LLVM.
|
|
|
|
ModuleSummaryIndex Index;
|
|
|
|
|
|
|
|
// All modules we may look at, stored as in-memory serialized versions. This
|
|
|
|
// is later used when inlining to ensure we can extract any module to inline
|
|
|
|
// from.
|
|
|
|
StringMap<MemoryBufferRef> ModuleMap;
|
|
|
|
|
|
|
|
// A set that we manage of everything we *don't* want internalized. Note that
|
|
|
|
// this includes all transitive references right now as well, but it may not
|
|
|
|
// always!
|
|
|
|
DenseSet<GlobalValue::GUID> GUIDPreservedSymbols;
|
|
|
|
|
|
|
|
// Not 100% sure what these are, but they impact what's internalized and
|
|
|
|
// what's inlined across modules, I believe.
|
|
|
|
StringMap<FunctionImporter::ImportMapTy> ImportLists;
|
|
|
|
StringMap<FunctionImporter::ExportSetTy> ExportLists;
|
|
|
|
StringMap<GVSummaryMapTy> ModuleToDefinedGVSummaries;
|
Use llvm::computeLTOCacheKey to determine post-ThinLTO CGU reuse
During incremental ThinLTO compilation, we attempt to re-use the
optimized (post-ThinLTO) bitcode file for a module if it is 'safe' to do
so.
Up until now, 'safe' has meant that the set of modules that our current
modules imports from/exports to is unchanged from the previous
compilation session. See PR #67020 and PR #71131 for more details.
However, this turns out be insufficient to guarantee that it's safe
to reuse the post-LTO module (i.e. that optimizing the pre-LTO module
would produce the same result). When LLVM optimizes a module during
ThinLTO, it may look at other information from the 'module index', such
as whether a (non-imported!) global variable is used. If this
information changes between compilation runs, we may end up re-using an
optimized module that (for example) had dead-code elimination run on a
function that is now used by another module.
Fortunately, LLVM implements its own ThinLTO module cache, which is used
when ThinLTO is performed by a linker plugin (e.g. when clang is used to
compile a C proect). Using this cache directly would require extensive
refactoring of our code - but fortunately for us, LLVM provides a
function that does exactly what we need.
The function `llvm::computeLTOCacheKey` is used to compute a SHA-1 hash
from all data that might influence the result of ThinLTO on a module.
In addition to the module imports/exports that we manually track, it
also hashes information about global variables (e.g. their liveness)
which might be used during optimization. By using this function, we
shouldn't have to worry about new LLVM passes breaking our module re-use
behavior.
In LLVM, the output of this function forms part of the filename used to
store the post-ThinLTO module. To keep our current filename structure
intact, this PR just writes out the mapping 'CGU name -> Hash' to a
file. To determine if a post-LTO module should be reused, we compare
hashes from the previous session.
This should unblock PR #75199 - by sheer chance, it seems to have hit
this issue due to the particular CGU partitioning and optimization
decisions that end up getting made.
2020-09-17 23:36:13 +02:00
|
|
|
StringMap<std::map<GlobalValue::GUID, GlobalValue::LinkageTypes>> ResolvedODR;
|
2018-01-30 20:53:18 +01:00
|
|
|
|
2019-01-16 18:59:03 +01:00
|
|
|
LLVMRustThinLTOData() : Index(/* HaveGVs = */ false) {}
|
rustc: Implement ThinLTO
This commit is an implementation of LLVM's ThinLTO for consumption in rustc
itself. Currently today LTO works by merging all relevant LLVM modules into one
and then running optimization passes. "Thin" LTO operates differently by having
more sharded work and allowing parallelism opportunities between optimizing
codegen units. Further down the road Thin LTO also allows *incremental* LTO
which should enable even faster release builds without compromising on the
performance we have today.
This commit uses a `-Z thinlto` flag to gate whether ThinLTO is enabled. It then
also implements two forms of ThinLTO:
* In one mode we'll *only* perform ThinLTO over the codegen units produced in a
single compilation. That is, we won't load upstream rlibs, but we'll instead
just perform ThinLTO amongst all codegen units produced by the compiler for
the local crate. This is intended to emulate a desired end point where we have
codegen units turned on by default for all crates and ThinLTO allows us to do
this without performance loss.
* In anther mode, like full LTO today, we'll optimize all upstream dependencies
in "thin" mode. Unlike today, however, this LTO step is fully parallelized so
should finish much more quickly.
There's a good bit of comments about what the implementation is doing and where
it came from, but the tl;dr; is that currently most of the support here is
copied from upstream LLVM. This code duplication is done for a number of
reasons:
* Controlling parallelism means we can use the existing jobserver support to
avoid overloading machines.
* We will likely want a slightly different form of incremental caching which
integrates with our own incremental strategy, but this is yet to be
determined.
* This buys us some flexibility about when/where we run ThinLTO, as well as
having it tailored to fit our needs for the time being.
* Finally this allows us to reuse some artifacts such as our `TargetMachine`
creation, where all our options we used today aren't necessarily supported by
upstream LLVM yet.
My hope is that we can get some experience with this copy/paste in tree and then
eventually upstream some work to LLVM itself to avoid the duplication while
still ensuring our needs are met. Otherwise I fear that maintaining these
bindings may be quite costly over the years with LLVM updates!
2017-07-23 17:14:38 +02:00
|
|
|
};
|
|
|
|
|
|
|
|
// Just an argument to the `LLVMRustCreateThinLTOData` function below.
|
|
|
|
struct LLVMRustThinLTOModule {
|
|
|
|
const char *identifier;
|
|
|
|
const char *data;
|
|
|
|
size_t len;
|
|
|
|
};
|
|
|
|
|
|
|
|
// This is copied from `lib/LTO/ThinLTOCodeGenerator.cpp`, not sure what it
|
|
|
|
// does.
|
|
|
|
static const GlobalValueSummary *
|
|
|
|
getFirstDefinitionForLinker(const GlobalValueSummaryList &GVSummaryList) {
|
|
|
|
auto StrongDefForLinker = llvm::find_if(
|
|
|
|
GVSummaryList, [](const std::unique_ptr<GlobalValueSummary> &Summary) {
|
|
|
|
auto Linkage = Summary->linkage();
|
|
|
|
return !GlobalValue::isAvailableExternallyLinkage(Linkage) &&
|
|
|
|
!GlobalValue::isWeakForLinker(Linkage);
|
|
|
|
});
|
|
|
|
if (StrongDefForLinker != GVSummaryList.end())
|
|
|
|
return StrongDefForLinker->get();
|
|
|
|
|
|
|
|
auto FirstDefForLinker = llvm::find_if(
|
|
|
|
GVSummaryList, [](const std::unique_ptr<GlobalValueSummary> &Summary) {
|
|
|
|
auto Linkage = Summary->linkage();
|
|
|
|
return !GlobalValue::isAvailableExternallyLinkage(Linkage);
|
|
|
|
});
|
|
|
|
if (FirstDefForLinker == GVSummaryList.end())
|
|
|
|
return nullptr;
|
|
|
|
return FirstDefForLinker->get();
|
|
|
|
}
|
|
|
|
|
|
|
|
// The main entry point for creating the global ThinLTO analysis. The structure
|
|
|
|
// here is basically the same as before threads are spawned in the `run`
|
|
|
|
// function of `lib/LTO/ThinLTOCodeGenerator.cpp`.
|
|
|
|
extern "C" LLVMRustThinLTOData*
|
|
|
|
LLVMRustCreateThinLTOData(LLVMRustThinLTOModule *modules,
|
|
|
|
int num_modules,
|
|
|
|
const char **preserved_symbols,
|
|
|
|
int num_symbols) {
|
2019-12-31 14:08:25 +01:00
|
|
|
#if LLVM_VERSION_GE(10, 0)
|
|
|
|
auto Ret = std::make_unique<LLVMRustThinLTOData>();
|
|
|
|
#else
|
rustc: Implement ThinLTO
This commit is an implementation of LLVM's ThinLTO for consumption in rustc
itself. Currently today LTO works by merging all relevant LLVM modules into one
and then running optimization passes. "Thin" LTO operates differently by having
more sharded work and allowing parallelism opportunities between optimizing
codegen units. Further down the road Thin LTO also allows *incremental* LTO
which should enable even faster release builds without compromising on the
performance we have today.
This commit uses a `-Z thinlto` flag to gate whether ThinLTO is enabled. It then
also implements two forms of ThinLTO:
* In one mode we'll *only* perform ThinLTO over the codegen units produced in a
single compilation. That is, we won't load upstream rlibs, but we'll instead
just perform ThinLTO amongst all codegen units produced by the compiler for
the local crate. This is intended to emulate a desired end point where we have
codegen units turned on by default for all crates and ThinLTO allows us to do
this without performance loss.
* In anther mode, like full LTO today, we'll optimize all upstream dependencies
in "thin" mode. Unlike today, however, this LTO step is fully parallelized so
should finish much more quickly.
There's a good bit of comments about what the implementation is doing and where
it came from, but the tl;dr; is that currently most of the support here is
copied from upstream LLVM. This code duplication is done for a number of
reasons:
* Controlling parallelism means we can use the existing jobserver support to
avoid overloading machines.
* We will likely want a slightly different form of incremental caching which
integrates with our own incremental strategy, but this is yet to be
determined.
* This buys us some flexibility about when/where we run ThinLTO, as well as
having it tailored to fit our needs for the time being.
* Finally this allows us to reuse some artifacts such as our `TargetMachine`
creation, where all our options we used today aren't necessarily supported by
upstream LLVM yet.
My hope is that we can get some experience with this copy/paste in tree and then
eventually upstream some work to LLVM itself to avoid the duplication while
still ensuring our needs are met. Otherwise I fear that maintaining these
bindings may be quite costly over the years with LLVM updates!
2017-07-23 17:14:38 +02:00
|
|
|
auto Ret = llvm::make_unique<LLVMRustThinLTOData>();
|
2019-12-31 14:08:25 +01:00
|
|
|
#endif
|
rustc: Implement ThinLTO
This commit is an implementation of LLVM's ThinLTO for consumption in rustc
itself. Currently today LTO works by merging all relevant LLVM modules into one
and then running optimization passes. "Thin" LTO operates differently by having
more sharded work and allowing parallelism opportunities between optimizing
codegen units. Further down the road Thin LTO also allows *incremental* LTO
which should enable even faster release builds without compromising on the
performance we have today.
This commit uses a `-Z thinlto` flag to gate whether ThinLTO is enabled. It then
also implements two forms of ThinLTO:
* In one mode we'll *only* perform ThinLTO over the codegen units produced in a
single compilation. That is, we won't load upstream rlibs, but we'll instead
just perform ThinLTO amongst all codegen units produced by the compiler for
the local crate. This is intended to emulate a desired end point where we have
codegen units turned on by default for all crates and ThinLTO allows us to do
this without performance loss.
* In anther mode, like full LTO today, we'll optimize all upstream dependencies
in "thin" mode. Unlike today, however, this LTO step is fully parallelized so
should finish much more quickly.
There's a good bit of comments about what the implementation is doing and where
it came from, but the tl;dr; is that currently most of the support here is
copied from upstream LLVM. This code duplication is done for a number of
reasons:
* Controlling parallelism means we can use the existing jobserver support to
avoid overloading machines.
* We will likely want a slightly different form of incremental caching which
integrates with our own incremental strategy, but this is yet to be
determined.
* This buys us some flexibility about when/where we run ThinLTO, as well as
having it tailored to fit our needs for the time being.
* Finally this allows us to reuse some artifacts such as our `TargetMachine`
creation, where all our options we used today aren't necessarily supported by
upstream LLVM yet.
My hope is that we can get some experience with this copy/paste in tree and then
eventually upstream some work to LLVM itself to avoid the duplication while
still ensuring our needs are met. Otherwise I fear that maintaining these
bindings may be quite costly over the years with LLVM updates!
2017-07-23 17:14:38 +02:00
|
|
|
|
|
|
|
// Load each module's summary and merge it into one combined index
|
|
|
|
for (int i = 0; i < num_modules; i++) {
|
|
|
|
auto module = &modules[i];
|
|
|
|
StringRef buffer(module->data, module->len);
|
|
|
|
MemoryBufferRef mem_buffer(buffer, module->identifier);
|
|
|
|
|
|
|
|
Ret->ModuleMap[module->identifier] = mem_buffer;
|
|
|
|
|
2017-10-10 21:29:14 +02:00
|
|
|
if (Error Err = readModuleSummaryIndex(mem_buffer, Ret->Index, i)) {
|
|
|
|
LLVMRustSetLastError(toString(std::move(Err)).c_str());
|
|
|
|
return nullptr;
|
|
|
|
}
|
rustc: Implement ThinLTO
This commit is an implementation of LLVM's ThinLTO for consumption in rustc
itself. Currently today LTO works by merging all relevant LLVM modules into one
and then running optimization passes. "Thin" LTO operates differently by having
more sharded work and allowing parallelism opportunities between optimizing
codegen units. Further down the road Thin LTO also allows *incremental* LTO
which should enable even faster release builds without compromising on the
performance we have today.
This commit uses a `-Z thinlto` flag to gate whether ThinLTO is enabled. It then
also implements two forms of ThinLTO:
* In one mode we'll *only* perform ThinLTO over the codegen units produced in a
single compilation. That is, we won't load upstream rlibs, but we'll instead
just perform ThinLTO amongst all codegen units produced by the compiler for
the local crate. This is intended to emulate a desired end point where we have
codegen units turned on by default for all crates and ThinLTO allows us to do
this without performance loss.
* In anther mode, like full LTO today, we'll optimize all upstream dependencies
in "thin" mode. Unlike today, however, this LTO step is fully parallelized so
should finish much more quickly.
There's a good bit of comments about what the implementation is doing and where
it came from, but the tl;dr; is that currently most of the support here is
copied from upstream LLVM. This code duplication is done for a number of
reasons:
* Controlling parallelism means we can use the existing jobserver support to
avoid overloading machines.
* We will likely want a slightly different form of incremental caching which
integrates with our own incremental strategy, but this is yet to be
determined.
* This buys us some flexibility about when/where we run ThinLTO, as well as
having it tailored to fit our needs for the time being.
* Finally this allows us to reuse some artifacts such as our `TargetMachine`
creation, where all our options we used today aren't necessarily supported by
upstream LLVM yet.
My hope is that we can get some experience with this copy/paste in tree and then
eventually upstream some work to LLVM itself to avoid the duplication while
still ensuring our needs are met. Otherwise I fear that maintaining these
bindings may be quite costly over the years with LLVM updates!
2017-07-23 17:14:38 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
// Collect for each module the list of function it defines (GUID -> Summary)
|
|
|
|
Ret->Index.collectDefinedGVSummariesPerModule(Ret->ModuleToDefinedGVSummaries);
|
|
|
|
|
|
|
|
// Convert the preserved symbols set from string to GUID, this is then needed
|
2017-11-29 21:00:10 +01:00
|
|
|
// for internalization.
|
rustc: Implement ThinLTO
This commit is an implementation of LLVM's ThinLTO for consumption in rustc
itself. Currently today LTO works by merging all relevant LLVM modules into one
and then running optimization passes. "Thin" LTO operates differently by having
more sharded work and allowing parallelism opportunities between optimizing
codegen units. Further down the road Thin LTO also allows *incremental* LTO
which should enable even faster release builds without compromising on the
performance we have today.
This commit uses a `-Z thinlto` flag to gate whether ThinLTO is enabled. It then
also implements two forms of ThinLTO:
* In one mode we'll *only* perform ThinLTO over the codegen units produced in a
single compilation. That is, we won't load upstream rlibs, but we'll instead
just perform ThinLTO amongst all codegen units produced by the compiler for
the local crate. This is intended to emulate a desired end point where we have
codegen units turned on by default for all crates and ThinLTO allows us to do
this without performance loss.
* In anther mode, like full LTO today, we'll optimize all upstream dependencies
in "thin" mode. Unlike today, however, this LTO step is fully parallelized so
should finish much more quickly.
There's a good bit of comments about what the implementation is doing and where
it came from, but the tl;dr; is that currently most of the support here is
copied from upstream LLVM. This code duplication is done for a number of
reasons:
* Controlling parallelism means we can use the existing jobserver support to
avoid overloading machines.
* We will likely want a slightly different form of incremental caching which
integrates with our own incremental strategy, but this is yet to be
determined.
* This buys us some flexibility about when/where we run ThinLTO, as well as
having it tailored to fit our needs for the time being.
* Finally this allows us to reuse some artifacts such as our `TargetMachine`
creation, where all our options we used today aren't necessarily supported by
upstream LLVM yet.
My hope is that we can get some experience with this copy/paste in tree and then
eventually upstream some work to LLVM itself to avoid the duplication while
still ensuring our needs are met. Otherwise I fear that maintaining these
bindings may be quite costly over the years with LLVM updates!
2017-07-23 17:14:38 +02:00
|
|
|
for (int i = 0; i < num_symbols; i++) {
|
2017-11-29 21:00:10 +01:00
|
|
|
auto GUID = GlobalValue::getGUID(preserved_symbols[i]);
|
|
|
|
Ret->GUIDPreservedSymbols.insert(GUID);
|
rustc: Implement ThinLTO
This commit is an implementation of LLVM's ThinLTO for consumption in rustc
itself. Currently today LTO works by merging all relevant LLVM modules into one
and then running optimization passes. "Thin" LTO operates differently by having
more sharded work and allowing parallelism opportunities between optimizing
codegen units. Further down the road Thin LTO also allows *incremental* LTO
which should enable even faster release builds without compromising on the
performance we have today.
This commit uses a `-Z thinlto` flag to gate whether ThinLTO is enabled. It then
also implements two forms of ThinLTO:
* In one mode we'll *only* perform ThinLTO over the codegen units produced in a
single compilation. That is, we won't load upstream rlibs, but we'll instead
just perform ThinLTO amongst all codegen units produced by the compiler for
the local crate. This is intended to emulate a desired end point where we have
codegen units turned on by default for all crates and ThinLTO allows us to do
this without performance loss.
* In anther mode, like full LTO today, we'll optimize all upstream dependencies
in "thin" mode. Unlike today, however, this LTO step is fully parallelized so
should finish much more quickly.
There's a good bit of comments about what the implementation is doing and where
it came from, but the tl;dr; is that currently most of the support here is
copied from upstream LLVM. This code duplication is done for a number of
reasons:
* Controlling parallelism means we can use the existing jobserver support to
avoid overloading machines.
* We will likely want a slightly different form of incremental caching which
integrates with our own incremental strategy, but this is yet to be
determined.
* This buys us some flexibility about when/where we run ThinLTO, as well as
having it tailored to fit our needs for the time being.
* Finally this allows us to reuse some artifacts such as our `TargetMachine`
creation, where all our options we used today aren't necessarily supported by
upstream LLVM yet.
My hope is that we can get some experience with this copy/paste in tree and then
eventually upstream some work to LLVM itself to avoid the duplication while
still ensuring our needs are met. Otherwise I fear that maintaining these
bindings may be quite costly over the years with LLVM updates!
2017-07-23 17:14:38 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
// Collect the import/export lists for all modules from the call-graph in the
|
|
|
|
// combined index
|
|
|
|
//
|
|
|
|
// This is copied from `lib/LTO/ThinLTOCodeGenerator.cpp`
|
2018-01-30 20:53:18 +01:00
|
|
|
auto deadIsPrevailing = [&](GlobalValue::GUID G) {
|
|
|
|
return PrevailingType::Unknown;
|
|
|
|
};
|
2019-04-26 17:58:14 +02:00
|
|
|
// We don't have a complete picture in our use of ThinLTO, just our immediate
|
|
|
|
// crate, so we need `ImportEnabled = false` to limit internalization.
|
|
|
|
// Otherwise, we sometimes lose `static` values -- see #60184.
|
2019-01-16 18:59:03 +01:00
|
|
|
computeDeadSymbolsWithConstProp(Ret->Index, Ret->GUIDPreservedSymbols,
|
2019-04-26 17:58:14 +02:00
|
|
|
deadIsPrevailing, /* ImportEnabled = */ false);
|
2017-10-17 07:41:56 +02:00
|
|
|
ComputeCrossModuleImport(
|
|
|
|
Ret->Index,
|
|
|
|
Ret->ModuleToDefinedGVSummaries,
|
|
|
|
Ret->ImportLists,
|
|
|
|
Ret->ExportLists
|
|
|
|
);
|
rustc: Implement ThinLTO
This commit is an implementation of LLVM's ThinLTO for consumption in rustc
itself. Currently today LTO works by merging all relevant LLVM modules into one
and then running optimization passes. "Thin" LTO operates differently by having
more sharded work and allowing parallelism opportunities between optimizing
codegen units. Further down the road Thin LTO also allows *incremental* LTO
which should enable even faster release builds without compromising on the
performance we have today.
This commit uses a `-Z thinlto` flag to gate whether ThinLTO is enabled. It then
also implements two forms of ThinLTO:
* In one mode we'll *only* perform ThinLTO over the codegen units produced in a
single compilation. That is, we won't load upstream rlibs, but we'll instead
just perform ThinLTO amongst all codegen units produced by the compiler for
the local crate. This is intended to emulate a desired end point where we have
codegen units turned on by default for all crates and ThinLTO allows us to do
this without performance loss.
* In anther mode, like full LTO today, we'll optimize all upstream dependencies
in "thin" mode. Unlike today, however, this LTO step is fully parallelized so
should finish much more quickly.
There's a good bit of comments about what the implementation is doing and where
it came from, but the tl;dr; is that currently most of the support here is
copied from upstream LLVM. This code duplication is done for a number of
reasons:
* Controlling parallelism means we can use the existing jobserver support to
avoid overloading machines.
* We will likely want a slightly different form of incremental caching which
integrates with our own incremental strategy, but this is yet to be
determined.
* This buys us some flexibility about when/where we run ThinLTO, as well as
having it tailored to fit our needs for the time being.
* Finally this allows us to reuse some artifacts such as our `TargetMachine`
creation, where all our options we used today aren't necessarily supported by
upstream LLVM yet.
My hope is that we can get some experience with this copy/paste in tree and then
eventually upstream some work to LLVM itself to avoid the duplication while
still ensuring our needs are met. Otherwise I fear that maintaining these
bindings may be quite costly over the years with LLVM updates!
2017-07-23 17:14:38 +02:00
|
|
|
|
|
|
|
// Resolve LinkOnce/Weak symbols, this has to be computed early be cause it
|
|
|
|
// impacts the caching.
|
|
|
|
//
|
2017-11-29 21:00:10 +01:00
|
|
|
// This is copied from `lib/LTO/ThinLTOCodeGenerator.cpp` with some of this
|
|
|
|
// being lifted from `lib/LTO/LTO.cpp` as well
|
rustc: Implement ThinLTO
This commit is an implementation of LLVM's ThinLTO for consumption in rustc
itself. Currently today LTO works by merging all relevant LLVM modules into one
and then running optimization passes. "Thin" LTO operates differently by having
more sharded work and allowing parallelism opportunities between optimizing
codegen units. Further down the road Thin LTO also allows *incremental* LTO
which should enable even faster release builds without compromising on the
performance we have today.
This commit uses a `-Z thinlto` flag to gate whether ThinLTO is enabled. It then
also implements two forms of ThinLTO:
* In one mode we'll *only* perform ThinLTO over the codegen units produced in a
single compilation. That is, we won't load upstream rlibs, but we'll instead
just perform ThinLTO amongst all codegen units produced by the compiler for
the local crate. This is intended to emulate a desired end point where we have
codegen units turned on by default for all crates and ThinLTO allows us to do
this without performance loss.
* In anther mode, like full LTO today, we'll optimize all upstream dependencies
in "thin" mode. Unlike today, however, this LTO step is fully parallelized so
should finish much more quickly.
There's a good bit of comments about what the implementation is doing and where
it came from, but the tl;dr; is that currently most of the support here is
copied from upstream LLVM. This code duplication is done for a number of
reasons:
* Controlling parallelism means we can use the existing jobserver support to
avoid overloading machines.
* We will likely want a slightly different form of incremental caching which
integrates with our own incremental strategy, but this is yet to be
determined.
* This buys us some flexibility about when/where we run ThinLTO, as well as
having it tailored to fit our needs for the time being.
* Finally this allows us to reuse some artifacts such as our `TargetMachine`
creation, where all our options we used today aren't necessarily supported by
upstream LLVM yet.
My hope is that we can get some experience with this copy/paste in tree and then
eventually upstream some work to LLVM itself to avoid the duplication while
still ensuring our needs are met. Otherwise I fear that maintaining these
bindings may be quite costly over the years with LLVM updates!
2017-07-23 17:14:38 +02:00
|
|
|
DenseMap<GlobalValue::GUID, const GlobalValueSummary *> PrevailingCopy;
|
|
|
|
for (auto &I : Ret->Index) {
|
2017-10-10 21:29:14 +02:00
|
|
|
if (I.second.SummaryList.size() > 1)
|
|
|
|
PrevailingCopy[I.first] = getFirstDefinitionForLinker(I.second.SummaryList);
|
rustc: Implement ThinLTO
This commit is an implementation of LLVM's ThinLTO for consumption in rustc
itself. Currently today LTO works by merging all relevant LLVM modules into one
and then running optimization passes. "Thin" LTO operates differently by having
more sharded work and allowing parallelism opportunities between optimizing
codegen units. Further down the road Thin LTO also allows *incremental* LTO
which should enable even faster release builds without compromising on the
performance we have today.
This commit uses a `-Z thinlto` flag to gate whether ThinLTO is enabled. It then
also implements two forms of ThinLTO:
* In one mode we'll *only* perform ThinLTO over the codegen units produced in a
single compilation. That is, we won't load upstream rlibs, but we'll instead
just perform ThinLTO amongst all codegen units produced by the compiler for
the local crate. This is intended to emulate a desired end point where we have
codegen units turned on by default for all crates and ThinLTO allows us to do
this without performance loss.
* In anther mode, like full LTO today, we'll optimize all upstream dependencies
in "thin" mode. Unlike today, however, this LTO step is fully parallelized so
should finish much more quickly.
There's a good bit of comments about what the implementation is doing and where
it came from, but the tl;dr; is that currently most of the support here is
copied from upstream LLVM. This code duplication is done for a number of
reasons:
* Controlling parallelism means we can use the existing jobserver support to
avoid overloading machines.
* We will likely want a slightly different form of incremental caching which
integrates with our own incremental strategy, but this is yet to be
determined.
* This buys us some flexibility about when/where we run ThinLTO, as well as
having it tailored to fit our needs for the time being.
* Finally this allows us to reuse some artifacts such as our `TargetMachine`
creation, where all our options we used today aren't necessarily supported by
upstream LLVM yet.
My hope is that we can get some experience with this copy/paste in tree and then
eventually upstream some work to LLVM itself to avoid the duplication while
still ensuring our needs are met. Otherwise I fear that maintaining these
bindings may be quite costly over the years with LLVM updates!
2017-07-23 17:14:38 +02:00
|
|
|
}
|
|
|
|
auto isPrevailing = [&](GlobalValue::GUID GUID, const GlobalValueSummary *S) {
|
|
|
|
const auto &Prevailing = PrevailingCopy.find(GUID);
|
|
|
|
if (Prevailing == PrevailingCopy.end())
|
|
|
|
return true;
|
|
|
|
return Prevailing->second == S;
|
|
|
|
};
|
|
|
|
auto recordNewLinkage = [&](StringRef ModuleIdentifier,
|
|
|
|
GlobalValue::GUID GUID,
|
|
|
|
GlobalValue::LinkageTypes NewLinkage) {
|
Use llvm::computeLTOCacheKey to determine post-ThinLTO CGU reuse
During incremental ThinLTO compilation, we attempt to re-use the
optimized (post-ThinLTO) bitcode file for a module if it is 'safe' to do
so.
Up until now, 'safe' has meant that the set of modules that our current
modules imports from/exports to is unchanged from the previous
compilation session. See PR #67020 and PR #71131 for more details.
However, this turns out be insufficient to guarantee that it's safe
to reuse the post-LTO module (i.e. that optimizing the pre-LTO module
would produce the same result). When LLVM optimizes a module during
ThinLTO, it may look at other information from the 'module index', such
as whether a (non-imported!) global variable is used. If this
information changes between compilation runs, we may end up re-using an
optimized module that (for example) had dead-code elimination run on a
function that is now used by another module.
Fortunately, LLVM implements its own ThinLTO module cache, which is used
when ThinLTO is performed by a linker plugin (e.g. when clang is used to
compile a C proect). Using this cache directly would require extensive
refactoring of our code - but fortunately for us, LLVM provides a
function that does exactly what we need.
The function `llvm::computeLTOCacheKey` is used to compute a SHA-1 hash
from all data that might influence the result of ThinLTO on a module.
In addition to the module imports/exports that we manually track, it
also hashes information about global variables (e.g. their liveness)
which might be used during optimization. By using this function, we
shouldn't have to worry about new LLVM passes breaking our module re-use
behavior.
In LLVM, the output of this function forms part of the filename used to
store the post-ThinLTO module. To keep our current filename structure
intact, this PR just writes out the mapping 'CGU name -> Hash' to a
file. To determine if a post-LTO module should be reused, we compare
hashes from the previous session.
This should unblock PR #75199 - by sheer chance, it seems to have hit
this issue due to the particular CGU partitioning and optimization
decisions that end up getting made.
2020-09-17 23:36:13 +02:00
|
|
|
Ret->ResolvedODR[ModuleIdentifier][GUID] = NewLinkage;
|
rustc: Implement ThinLTO
This commit is an implementation of LLVM's ThinLTO for consumption in rustc
itself. Currently today LTO works by merging all relevant LLVM modules into one
and then running optimization passes. "Thin" LTO operates differently by having
more sharded work and allowing parallelism opportunities between optimizing
codegen units. Further down the road Thin LTO also allows *incremental* LTO
which should enable even faster release builds without compromising on the
performance we have today.
This commit uses a `-Z thinlto` flag to gate whether ThinLTO is enabled. It then
also implements two forms of ThinLTO:
* In one mode we'll *only* perform ThinLTO over the codegen units produced in a
single compilation. That is, we won't load upstream rlibs, but we'll instead
just perform ThinLTO amongst all codegen units produced by the compiler for
the local crate. This is intended to emulate a desired end point where we have
codegen units turned on by default for all crates and ThinLTO allows us to do
this without performance loss.
* In anther mode, like full LTO today, we'll optimize all upstream dependencies
in "thin" mode. Unlike today, however, this LTO step is fully parallelized so
should finish much more quickly.
There's a good bit of comments about what the implementation is doing and where
it came from, but the tl;dr; is that currently most of the support here is
copied from upstream LLVM. This code duplication is done for a number of
reasons:
* Controlling parallelism means we can use the existing jobserver support to
avoid overloading machines.
* We will likely want a slightly different form of incremental caching which
integrates with our own incremental strategy, but this is yet to be
determined.
* This buys us some flexibility about when/where we run ThinLTO, as well as
having it tailored to fit our needs for the time being.
* Finally this allows us to reuse some artifacts such as our `TargetMachine`
creation, where all our options we used today aren't necessarily supported by
upstream LLVM yet.
My hope is that we can get some experience with this copy/paste in tree and then
eventually upstream some work to LLVM itself to avoid the duplication while
still ensuring our needs are met. Otherwise I fear that maintaining these
bindings may be quite costly over the years with LLVM updates!
2017-07-23 17:14:38 +02:00
|
|
|
};
|
2020-11-07 23:25:45 +01:00
|
|
|
|
2019-07-06 19:02:48 +02:00
|
|
|
thinLTOResolvePrevailingInIndex(Ret->Index, isPrevailing, recordNewLinkage,
|
|
|
|
Ret->GUIDPreservedSymbols);
|
2017-11-29 21:00:10 +01:00
|
|
|
|
|
|
|
// Here we calculate an `ExportedGUIDs` set for use in the `isExported`
|
|
|
|
// callback below. This callback below will dictate the linkage for all
|
|
|
|
// summaries in the index, and we basically just only want to ensure that dead
|
|
|
|
// symbols are internalized. Otherwise everything that's already external
|
|
|
|
// linkage will stay as external, and internal will stay as internal.
|
|
|
|
std::set<GlobalValue::GUID> ExportedGUIDs;
|
|
|
|
for (auto &List : Ret->Index) {
|
2017-12-11 07:52:58 +01:00
|
|
|
for (auto &GVS: List.second.SummaryList) {
|
2017-12-07 02:59:33 +01:00
|
|
|
if (GlobalValue::isLocalLinkage(GVS->linkage()))
|
2017-11-29 21:00:10 +01:00
|
|
|
continue;
|
|
|
|
auto GUID = GVS->getOriginalName();
|
2017-12-11 07:52:58 +01:00
|
|
|
if (GVS->flags().Live)
|
2017-11-29 21:00:10 +01:00
|
|
|
ExportedGUIDs.insert(GUID);
|
|
|
|
}
|
|
|
|
}
|
2019-12-31 14:14:31 +01:00
|
|
|
#if LLVM_VERSION_GE(10, 0)
|
|
|
|
auto isExported = [&](StringRef ModuleIdentifier, ValueInfo VI) {
|
|
|
|
const auto &ExportList = Ret->ExportLists.find(ModuleIdentifier);
|
|
|
|
return (ExportList != Ret->ExportLists.end() &&
|
|
|
|
ExportList->second.count(VI)) ||
|
|
|
|
ExportedGUIDs.count(VI.getGUID());
|
|
|
|
};
|
|
|
|
thinLTOInternalizeAndPromoteInIndex(Ret->Index, isExported, isPrevailing);
|
|
|
|
#else
|
rustc: Implement ThinLTO
This commit is an implementation of LLVM's ThinLTO for consumption in rustc
itself. Currently today LTO works by merging all relevant LLVM modules into one
and then running optimization passes. "Thin" LTO operates differently by having
more sharded work and allowing parallelism opportunities between optimizing
codegen units. Further down the road Thin LTO also allows *incremental* LTO
which should enable even faster release builds without compromising on the
performance we have today.
This commit uses a `-Z thinlto` flag to gate whether ThinLTO is enabled. It then
also implements two forms of ThinLTO:
* In one mode we'll *only* perform ThinLTO over the codegen units produced in a
single compilation. That is, we won't load upstream rlibs, but we'll instead
just perform ThinLTO amongst all codegen units produced by the compiler for
the local crate. This is intended to emulate a desired end point where we have
codegen units turned on by default for all crates and ThinLTO allows us to do
this without performance loss.
* In anther mode, like full LTO today, we'll optimize all upstream dependencies
in "thin" mode. Unlike today, however, this LTO step is fully parallelized so
should finish much more quickly.
There's a good bit of comments about what the implementation is doing and where
it came from, but the tl;dr; is that currently most of the support here is
copied from upstream LLVM. This code duplication is done for a number of
reasons:
* Controlling parallelism means we can use the existing jobserver support to
avoid overloading machines.
* We will likely want a slightly different form of incremental caching which
integrates with our own incremental strategy, but this is yet to be
determined.
* This buys us some flexibility about when/where we run ThinLTO, as well as
having it tailored to fit our needs for the time being.
* Finally this allows us to reuse some artifacts such as our `TargetMachine`
creation, where all our options we used today aren't necessarily supported by
upstream LLVM yet.
My hope is that we can get some experience with this copy/paste in tree and then
eventually upstream some work to LLVM itself to avoid the duplication while
still ensuring our needs are met. Otherwise I fear that maintaining these
bindings may be quite costly over the years with LLVM updates!
2017-07-23 17:14:38 +02:00
|
|
|
auto isExported = [&](StringRef ModuleIdentifier, GlobalValue::GUID GUID) {
|
|
|
|
const auto &ExportList = Ret->ExportLists.find(ModuleIdentifier);
|
|
|
|
return (ExportList != Ret->ExportLists.end() &&
|
|
|
|
ExportList->second.count(GUID)) ||
|
2017-11-29 21:00:10 +01:00
|
|
|
ExportedGUIDs.count(GUID);
|
rustc: Implement ThinLTO
This commit is an implementation of LLVM's ThinLTO for consumption in rustc
itself. Currently today LTO works by merging all relevant LLVM modules into one
and then running optimization passes. "Thin" LTO operates differently by having
more sharded work and allowing parallelism opportunities between optimizing
codegen units. Further down the road Thin LTO also allows *incremental* LTO
which should enable even faster release builds without compromising on the
performance we have today.
This commit uses a `-Z thinlto` flag to gate whether ThinLTO is enabled. It then
also implements two forms of ThinLTO:
* In one mode we'll *only* perform ThinLTO over the codegen units produced in a
single compilation. That is, we won't load upstream rlibs, but we'll instead
just perform ThinLTO amongst all codegen units produced by the compiler for
the local crate. This is intended to emulate a desired end point where we have
codegen units turned on by default for all crates and ThinLTO allows us to do
this without performance loss.
* In anther mode, like full LTO today, we'll optimize all upstream dependencies
in "thin" mode. Unlike today, however, this LTO step is fully parallelized so
should finish much more quickly.
There's a good bit of comments about what the implementation is doing and where
it came from, but the tl;dr; is that currently most of the support here is
copied from upstream LLVM. This code duplication is done for a number of
reasons:
* Controlling parallelism means we can use the existing jobserver support to
avoid overloading machines.
* We will likely want a slightly different form of incremental caching which
integrates with our own incremental strategy, but this is yet to be
determined.
* This buys us some flexibility about when/where we run ThinLTO, as well as
having it tailored to fit our needs for the time being.
* Finally this allows us to reuse some artifacts such as our `TargetMachine`
creation, where all our options we used today aren't necessarily supported by
upstream LLVM yet.
My hope is that we can get some experience with this copy/paste in tree and then
eventually upstream some work to LLVM itself to avoid the duplication while
still ensuring our needs are met. Otherwise I fear that maintaining these
bindings may be quite costly over the years with LLVM updates!
2017-07-23 17:14:38 +02:00
|
|
|
};
|
|
|
|
thinLTOInternalizeAndPromoteInIndex(Ret->Index, isExported);
|
2019-12-31 14:14:31 +01:00
|
|
|
#endif
|
rustc: Implement ThinLTO
This commit is an implementation of LLVM's ThinLTO for consumption in rustc
itself. Currently today LTO works by merging all relevant LLVM modules into one
and then running optimization passes. "Thin" LTO operates differently by having
more sharded work and allowing parallelism opportunities between optimizing
codegen units. Further down the road Thin LTO also allows *incremental* LTO
which should enable even faster release builds without compromising on the
performance we have today.
This commit uses a `-Z thinlto` flag to gate whether ThinLTO is enabled. It then
also implements two forms of ThinLTO:
* In one mode we'll *only* perform ThinLTO over the codegen units produced in a
single compilation. That is, we won't load upstream rlibs, but we'll instead
just perform ThinLTO amongst all codegen units produced by the compiler for
the local crate. This is intended to emulate a desired end point where we have
codegen units turned on by default for all crates and ThinLTO allows us to do
this without performance loss.
* In anther mode, like full LTO today, we'll optimize all upstream dependencies
in "thin" mode. Unlike today, however, this LTO step is fully parallelized so
should finish much more quickly.
There's a good bit of comments about what the implementation is doing and where
it came from, but the tl;dr; is that currently most of the support here is
copied from upstream LLVM. This code duplication is done for a number of
reasons:
* Controlling parallelism means we can use the existing jobserver support to
avoid overloading machines.
* We will likely want a slightly different form of incremental caching which
integrates with our own incremental strategy, but this is yet to be
determined.
* This buys us some flexibility about when/where we run ThinLTO, as well as
having it tailored to fit our needs for the time being.
* Finally this allows us to reuse some artifacts such as our `TargetMachine`
creation, where all our options we used today aren't necessarily supported by
upstream LLVM yet.
My hope is that we can get some experience with this copy/paste in tree and then
eventually upstream some work to LLVM itself to avoid the duplication while
still ensuring our needs are met. Otherwise I fear that maintaining these
bindings may be quite costly over the years with LLVM updates!
2017-07-23 17:14:38 +02:00
|
|
|
|
|
|
|
return Ret.release();
|
|
|
|
}
|
|
|
|
|
|
|
|
extern "C" void
|
|
|
|
LLVMRustFreeThinLTOData(LLVMRustThinLTOData *Data) {
|
|
|
|
delete Data;
|
|
|
|
}
|
|
|
|
|
|
|
|
// Below are the various passes that happen *per module* when doing ThinLTO.
|
|
|
|
//
|
|
|
|
// In other words, these are the functions that are all run concurrently
|
|
|
|
// with one another, one per module. The passes here correspond to the analysis
|
|
|
|
// passes in `lib/LTO/ThinLTOCodeGenerator.cpp`, currently found in the
|
|
|
|
// `ProcessThinLTOModule` function. Here they're split up into separate steps
|
|
|
|
// so rustc can save off the intermediate bytecode between each step.
|
|
|
|
|
2020-06-26 03:52:41 +02:00
|
|
|
#if LLVM_VERSION_GE(11, 0)
|
|
|
|
static bool
|
|
|
|
clearDSOLocalOnDeclarations(Module &Mod, TargetMachine &TM) {
|
|
|
|
// When linking an ELF shared object, dso_local should be dropped. We
|
|
|
|
// conservatively do this for -fpic.
|
|
|
|
bool ClearDSOLocalOnDeclarations =
|
|
|
|
TM.getTargetTriple().isOSBinFormatELF() &&
|
|
|
|
TM.getRelocationModel() != Reloc::Static &&
|
|
|
|
Mod.getPIELevel() == PIELevel::Default;
|
|
|
|
return ClearDSOLocalOnDeclarations;
|
|
|
|
}
|
|
|
|
#endif
|
|
|
|
|
rustc: Implement ThinLTO
This commit is an implementation of LLVM's ThinLTO for consumption in rustc
itself. Currently today LTO works by merging all relevant LLVM modules into one
and then running optimization passes. "Thin" LTO operates differently by having
more sharded work and allowing parallelism opportunities between optimizing
codegen units. Further down the road Thin LTO also allows *incremental* LTO
which should enable even faster release builds without compromising on the
performance we have today.
This commit uses a `-Z thinlto` flag to gate whether ThinLTO is enabled. It then
also implements two forms of ThinLTO:
* In one mode we'll *only* perform ThinLTO over the codegen units produced in a
single compilation. That is, we won't load upstream rlibs, but we'll instead
just perform ThinLTO amongst all codegen units produced by the compiler for
the local crate. This is intended to emulate a desired end point where we have
codegen units turned on by default for all crates and ThinLTO allows us to do
this without performance loss.
* In anther mode, like full LTO today, we'll optimize all upstream dependencies
in "thin" mode. Unlike today, however, this LTO step is fully parallelized so
should finish much more quickly.
There's a good bit of comments about what the implementation is doing and where
it came from, but the tl;dr; is that currently most of the support here is
copied from upstream LLVM. This code duplication is done for a number of
reasons:
* Controlling parallelism means we can use the existing jobserver support to
avoid overloading machines.
* We will likely want a slightly different form of incremental caching which
integrates with our own incremental strategy, but this is yet to be
determined.
* This buys us some flexibility about when/where we run ThinLTO, as well as
having it tailored to fit our needs for the time being.
* Finally this allows us to reuse some artifacts such as our `TargetMachine`
creation, where all our options we used today aren't necessarily supported by
upstream LLVM yet.
My hope is that we can get some experience with this copy/paste in tree and then
eventually upstream some work to LLVM itself to avoid the duplication while
still ensuring our needs are met. Otherwise I fear that maintaining these
bindings may be quite costly over the years with LLVM updates!
2017-07-23 17:14:38 +02:00
|
|
|
extern "C" bool
|
2020-06-26 03:52:41 +02:00
|
|
|
LLVMRustPrepareThinLTORename(const LLVMRustThinLTOData *Data, LLVMModuleRef M,
|
|
|
|
LLVMTargetMachineRef TM) {
|
rustc: Implement ThinLTO
This commit is an implementation of LLVM's ThinLTO for consumption in rustc
itself. Currently today LTO works by merging all relevant LLVM modules into one
and then running optimization passes. "Thin" LTO operates differently by having
more sharded work and allowing parallelism opportunities between optimizing
codegen units. Further down the road Thin LTO also allows *incremental* LTO
which should enable even faster release builds without compromising on the
performance we have today.
This commit uses a `-Z thinlto` flag to gate whether ThinLTO is enabled. It then
also implements two forms of ThinLTO:
* In one mode we'll *only* perform ThinLTO over the codegen units produced in a
single compilation. That is, we won't load upstream rlibs, but we'll instead
just perform ThinLTO amongst all codegen units produced by the compiler for
the local crate. This is intended to emulate a desired end point where we have
codegen units turned on by default for all crates and ThinLTO allows us to do
this without performance loss.
* In anther mode, like full LTO today, we'll optimize all upstream dependencies
in "thin" mode. Unlike today, however, this LTO step is fully parallelized so
should finish much more quickly.
There's a good bit of comments about what the implementation is doing and where
it came from, but the tl;dr; is that currently most of the support here is
copied from upstream LLVM. This code duplication is done for a number of
reasons:
* Controlling parallelism means we can use the existing jobserver support to
avoid overloading machines.
* We will likely want a slightly different form of incremental caching which
integrates with our own incremental strategy, but this is yet to be
determined.
* This buys us some flexibility about when/where we run ThinLTO, as well as
having it tailored to fit our needs for the time being.
* Finally this allows us to reuse some artifacts such as our `TargetMachine`
creation, where all our options we used today aren't necessarily supported by
upstream LLVM yet.
My hope is that we can get some experience with this copy/paste in tree and then
eventually upstream some work to LLVM itself to avoid the duplication while
still ensuring our needs are met. Otherwise I fear that maintaining these
bindings may be quite costly over the years with LLVM updates!
2017-07-23 17:14:38 +02:00
|
|
|
Module &Mod = *unwrap(M);
|
2020-06-26 03:52:41 +02:00
|
|
|
TargetMachine &Target = *unwrap(TM);
|
|
|
|
|
|
|
|
#if LLVM_VERSION_GE(11, 0)
|
|
|
|
bool ClearDSOLocal = clearDSOLocalOnDeclarations(Mod, Target);
|
|
|
|
bool error = renameModuleForThinLTO(Mod, Data->Index, ClearDSOLocal);
|
|
|
|
#else
|
|
|
|
bool error = renameModuleForThinLTO(Mod, Data->Index);
|
|
|
|
#endif
|
|
|
|
|
|
|
|
if (error) {
|
rustc: Implement ThinLTO
This commit is an implementation of LLVM's ThinLTO for consumption in rustc
itself. Currently today LTO works by merging all relevant LLVM modules into one
and then running optimization passes. "Thin" LTO operates differently by having
more sharded work and allowing parallelism opportunities between optimizing
codegen units. Further down the road Thin LTO also allows *incremental* LTO
which should enable even faster release builds without compromising on the
performance we have today.
This commit uses a `-Z thinlto` flag to gate whether ThinLTO is enabled. It then
also implements two forms of ThinLTO:
* In one mode we'll *only* perform ThinLTO over the codegen units produced in a
single compilation. That is, we won't load upstream rlibs, but we'll instead
just perform ThinLTO amongst all codegen units produced by the compiler for
the local crate. This is intended to emulate a desired end point where we have
codegen units turned on by default for all crates and ThinLTO allows us to do
this without performance loss.
* In anther mode, like full LTO today, we'll optimize all upstream dependencies
in "thin" mode. Unlike today, however, this LTO step is fully parallelized so
should finish much more quickly.
There's a good bit of comments about what the implementation is doing and where
it came from, but the tl;dr; is that currently most of the support here is
copied from upstream LLVM. This code duplication is done for a number of
reasons:
* Controlling parallelism means we can use the existing jobserver support to
avoid overloading machines.
* We will likely want a slightly different form of incremental caching which
integrates with our own incremental strategy, but this is yet to be
determined.
* This buys us some flexibility about when/where we run ThinLTO, as well as
having it tailored to fit our needs for the time being.
* Finally this allows us to reuse some artifacts such as our `TargetMachine`
creation, where all our options we used today aren't necessarily supported by
upstream LLVM yet.
My hope is that we can get some experience with this copy/paste in tree and then
eventually upstream some work to LLVM itself to avoid the duplication while
still ensuring our needs are met. Otherwise I fear that maintaining these
bindings may be quite costly over the years with LLVM updates!
2017-07-23 17:14:38 +02:00
|
|
|
LLVMRustSetLastError("renameModuleForThinLTO failed");
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
return true;
|
|
|
|
}
|
|
|
|
|
|
|
|
extern "C" bool
|
|
|
|
LLVMRustPrepareThinLTOResolveWeak(const LLVMRustThinLTOData *Data, LLVMModuleRef M) {
|
|
|
|
Module &Mod = *unwrap(M);
|
|
|
|
const auto &DefinedGlobals = Data->ModuleToDefinedGVSummaries.lookup(Mod.getModuleIdentifier());
|
2018-11-09 22:09:32 +01:00
|
|
|
thinLTOResolvePrevailingInModule(Mod, DefinedGlobals);
|
rustc: Implement ThinLTO
This commit is an implementation of LLVM's ThinLTO for consumption in rustc
itself. Currently today LTO works by merging all relevant LLVM modules into one
and then running optimization passes. "Thin" LTO operates differently by having
more sharded work and allowing parallelism opportunities between optimizing
codegen units. Further down the road Thin LTO also allows *incremental* LTO
which should enable even faster release builds without compromising on the
performance we have today.
This commit uses a `-Z thinlto` flag to gate whether ThinLTO is enabled. It then
also implements two forms of ThinLTO:
* In one mode we'll *only* perform ThinLTO over the codegen units produced in a
single compilation. That is, we won't load upstream rlibs, but we'll instead
just perform ThinLTO amongst all codegen units produced by the compiler for
the local crate. This is intended to emulate a desired end point where we have
codegen units turned on by default for all crates and ThinLTO allows us to do
this without performance loss.
* In anther mode, like full LTO today, we'll optimize all upstream dependencies
in "thin" mode. Unlike today, however, this LTO step is fully parallelized so
should finish much more quickly.
There's a good bit of comments about what the implementation is doing and where
it came from, but the tl;dr; is that currently most of the support here is
copied from upstream LLVM. This code duplication is done for a number of
reasons:
* Controlling parallelism means we can use the existing jobserver support to
avoid overloading machines.
* We will likely want a slightly different form of incremental caching which
integrates with our own incremental strategy, but this is yet to be
determined.
* This buys us some flexibility about when/where we run ThinLTO, as well as
having it tailored to fit our needs for the time being.
* Finally this allows us to reuse some artifacts such as our `TargetMachine`
creation, where all our options we used today aren't necessarily supported by
upstream LLVM yet.
My hope is that we can get some experience with this copy/paste in tree and then
eventually upstream some work to LLVM itself to avoid the duplication while
still ensuring our needs are met. Otherwise I fear that maintaining these
bindings may be quite costly over the years with LLVM updates!
2017-07-23 17:14:38 +02:00
|
|
|
return true;
|
|
|
|
}
|
|
|
|
|
|
|
|
extern "C" bool
|
|
|
|
LLVMRustPrepareThinLTOInternalize(const LLVMRustThinLTOData *Data, LLVMModuleRef M) {
|
|
|
|
Module &Mod = *unwrap(M);
|
|
|
|
const auto &DefinedGlobals = Data->ModuleToDefinedGVSummaries.lookup(Mod.getModuleIdentifier());
|
|
|
|
thinLTOInternalizeModule(Mod, DefinedGlobals);
|
|
|
|
return true;
|
|
|
|
}
|
|
|
|
|
|
|
|
extern "C" bool
|
2020-06-26 03:52:41 +02:00
|
|
|
LLVMRustPrepareThinLTOImport(const LLVMRustThinLTOData *Data, LLVMModuleRef M,
|
|
|
|
LLVMTargetMachineRef TM) {
|
rustc: Implement ThinLTO
This commit is an implementation of LLVM's ThinLTO for consumption in rustc
itself. Currently today LTO works by merging all relevant LLVM modules into one
and then running optimization passes. "Thin" LTO operates differently by having
more sharded work and allowing parallelism opportunities between optimizing
codegen units. Further down the road Thin LTO also allows *incremental* LTO
which should enable even faster release builds without compromising on the
performance we have today.
This commit uses a `-Z thinlto` flag to gate whether ThinLTO is enabled. It then
also implements two forms of ThinLTO:
* In one mode we'll *only* perform ThinLTO over the codegen units produced in a
single compilation. That is, we won't load upstream rlibs, but we'll instead
just perform ThinLTO amongst all codegen units produced by the compiler for
the local crate. This is intended to emulate a desired end point where we have
codegen units turned on by default for all crates and ThinLTO allows us to do
this without performance loss.
* In anther mode, like full LTO today, we'll optimize all upstream dependencies
in "thin" mode. Unlike today, however, this LTO step is fully parallelized so
should finish much more quickly.
There's a good bit of comments about what the implementation is doing and where
it came from, but the tl;dr; is that currently most of the support here is
copied from upstream LLVM. This code duplication is done for a number of
reasons:
* Controlling parallelism means we can use the existing jobserver support to
avoid overloading machines.
* We will likely want a slightly different form of incremental caching which
integrates with our own incremental strategy, but this is yet to be
determined.
* This buys us some flexibility about when/where we run ThinLTO, as well as
having it tailored to fit our needs for the time being.
* Finally this allows us to reuse some artifacts such as our `TargetMachine`
creation, where all our options we used today aren't necessarily supported by
upstream LLVM yet.
My hope is that we can get some experience with this copy/paste in tree and then
eventually upstream some work to LLVM itself to avoid the duplication while
still ensuring our needs are met. Otherwise I fear that maintaining these
bindings may be quite costly over the years with LLVM updates!
2017-07-23 17:14:38 +02:00
|
|
|
Module &Mod = *unwrap(M);
|
2020-06-26 03:52:41 +02:00
|
|
|
TargetMachine &Target = *unwrap(TM);
|
2018-07-18 21:12:53 +02:00
|
|
|
|
rustc: Implement ThinLTO
This commit is an implementation of LLVM's ThinLTO for consumption in rustc
itself. Currently today LTO works by merging all relevant LLVM modules into one
and then running optimization passes. "Thin" LTO operates differently by having
more sharded work and allowing parallelism opportunities between optimizing
codegen units. Further down the road Thin LTO also allows *incremental* LTO
which should enable even faster release builds without compromising on the
performance we have today.
This commit uses a `-Z thinlto` flag to gate whether ThinLTO is enabled. It then
also implements two forms of ThinLTO:
* In one mode we'll *only* perform ThinLTO over the codegen units produced in a
single compilation. That is, we won't load upstream rlibs, but we'll instead
just perform ThinLTO amongst all codegen units produced by the compiler for
the local crate. This is intended to emulate a desired end point where we have
codegen units turned on by default for all crates and ThinLTO allows us to do
this without performance loss.
* In anther mode, like full LTO today, we'll optimize all upstream dependencies
in "thin" mode. Unlike today, however, this LTO step is fully parallelized so
should finish much more quickly.
There's a good bit of comments about what the implementation is doing and where
it came from, but the tl;dr; is that currently most of the support here is
copied from upstream LLVM. This code duplication is done for a number of
reasons:
* Controlling parallelism means we can use the existing jobserver support to
avoid overloading machines.
* We will likely want a slightly different form of incremental caching which
integrates with our own incremental strategy, but this is yet to be
determined.
* This buys us some flexibility about when/where we run ThinLTO, as well as
having it tailored to fit our needs for the time being.
* Finally this allows us to reuse some artifacts such as our `TargetMachine`
creation, where all our options we used today aren't necessarily supported by
upstream LLVM yet.
My hope is that we can get some experience with this copy/paste in tree and then
eventually upstream some work to LLVM itself to avoid the duplication while
still ensuring our needs are met. Otherwise I fear that maintaining these
bindings may be quite costly over the years with LLVM updates!
2017-07-23 17:14:38 +02:00
|
|
|
const auto &ImportList = Data->ImportLists.lookup(Mod.getModuleIdentifier());
|
|
|
|
auto Loader = [&](StringRef Identifier) {
|
|
|
|
const auto &Memory = Data->ModuleMap.lookup(Identifier);
|
|
|
|
auto &Context = Mod.getContext();
|
2018-07-18 21:12:53 +02:00
|
|
|
auto MOrErr = getLazyBitcodeModule(Memory, Context, true, true);
|
|
|
|
|
|
|
|
if (!MOrErr)
|
2018-07-29 13:20:06 +02:00
|
|
|
return MOrErr;
|
2018-07-18 21:12:53 +02:00
|
|
|
|
|
|
|
// The rest of this closure is a workaround for
|
|
|
|
// https://bugs.llvm.org/show_bug.cgi?id=38184 where during ThinLTO imports
|
|
|
|
// we accidentally import wasm custom sections into different modules,
|
|
|
|
// duplicating them by in the final output artifact.
|
|
|
|
//
|
|
|
|
// The issue is worked around here by manually removing the
|
|
|
|
// `wasm.custom_sections` named metadata node from any imported module. This
|
|
|
|
// we know isn't used by any optimization pass so there's no need for it to
|
|
|
|
// be imported.
|
|
|
|
//
|
|
|
|
// Note that the metadata is currently lazily loaded, so we materialize it
|
|
|
|
// here before looking up if there's metadata inside. The `FunctionImporter`
|
|
|
|
// will immediately materialize metadata anyway after an import, so this
|
|
|
|
// shouldn't be a perf hit.
|
|
|
|
if (Error Err = (*MOrErr)->materializeMetadata()) {
|
|
|
|
Expected<std::unique_ptr<Module>> Ret(std::move(Err));
|
2018-07-29 13:20:06 +02:00
|
|
|
return Ret;
|
2018-07-18 21:12:53 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
auto *WasmCustomSections = (*MOrErr)->getNamedMetadata("wasm.custom_sections");
|
|
|
|
if (WasmCustomSections)
|
|
|
|
WasmCustomSections->eraseFromParent();
|
|
|
|
|
2018-07-29 13:20:06 +02:00
|
|
|
return MOrErr;
|
rustc: Implement ThinLTO
This commit is an implementation of LLVM's ThinLTO for consumption in rustc
itself. Currently today LTO works by merging all relevant LLVM modules into one
and then running optimization passes. "Thin" LTO operates differently by having
more sharded work and allowing parallelism opportunities between optimizing
codegen units. Further down the road Thin LTO also allows *incremental* LTO
which should enable even faster release builds without compromising on the
performance we have today.
This commit uses a `-Z thinlto` flag to gate whether ThinLTO is enabled. It then
also implements two forms of ThinLTO:
* In one mode we'll *only* perform ThinLTO over the codegen units produced in a
single compilation. That is, we won't load upstream rlibs, but we'll instead
just perform ThinLTO amongst all codegen units produced by the compiler for
the local crate. This is intended to emulate a desired end point where we have
codegen units turned on by default for all crates and ThinLTO allows us to do
this without performance loss.
* In anther mode, like full LTO today, we'll optimize all upstream dependencies
in "thin" mode. Unlike today, however, this LTO step is fully parallelized so
should finish much more quickly.
There's a good bit of comments about what the implementation is doing and where
it came from, but the tl;dr; is that currently most of the support here is
copied from upstream LLVM. This code duplication is done for a number of
reasons:
* Controlling parallelism means we can use the existing jobserver support to
avoid overloading machines.
* We will likely want a slightly different form of incremental caching which
integrates with our own incremental strategy, but this is yet to be
determined.
* This buys us some flexibility about when/where we run ThinLTO, as well as
having it tailored to fit our needs for the time being.
* Finally this allows us to reuse some artifacts such as our `TargetMachine`
creation, where all our options we used today aren't necessarily supported by
upstream LLVM yet.
My hope is that we can get some experience with this copy/paste in tree and then
eventually upstream some work to LLVM itself to avoid the duplication while
still ensuring our needs are met. Otherwise I fear that maintaining these
bindings may be quite costly over the years with LLVM updates!
2017-07-23 17:14:38 +02:00
|
|
|
};
|
2020-06-26 03:52:41 +02:00
|
|
|
#if LLVM_VERSION_GE(11, 0)
|
|
|
|
bool ClearDSOLocal = clearDSOLocalOnDeclarations(Mod, Target);
|
|
|
|
FunctionImporter Importer(Data->Index, Loader, ClearDSOLocal);
|
|
|
|
#else
|
rustc: Implement ThinLTO
This commit is an implementation of LLVM's ThinLTO for consumption in rustc
itself. Currently today LTO works by merging all relevant LLVM modules into one
and then running optimization passes. "Thin" LTO operates differently by having
more sharded work and allowing parallelism opportunities between optimizing
codegen units. Further down the road Thin LTO also allows *incremental* LTO
which should enable even faster release builds without compromising on the
performance we have today.
This commit uses a `-Z thinlto` flag to gate whether ThinLTO is enabled. It then
also implements two forms of ThinLTO:
* In one mode we'll *only* perform ThinLTO over the codegen units produced in a
single compilation. That is, we won't load upstream rlibs, but we'll instead
just perform ThinLTO amongst all codegen units produced by the compiler for
the local crate. This is intended to emulate a desired end point where we have
codegen units turned on by default for all crates and ThinLTO allows us to do
this without performance loss.
* In anther mode, like full LTO today, we'll optimize all upstream dependencies
in "thin" mode. Unlike today, however, this LTO step is fully parallelized so
should finish much more quickly.
There's a good bit of comments about what the implementation is doing and where
it came from, but the tl;dr; is that currently most of the support here is
copied from upstream LLVM. This code duplication is done for a number of
reasons:
* Controlling parallelism means we can use the existing jobserver support to
avoid overloading machines.
* We will likely want a slightly different form of incremental caching which
integrates with our own incremental strategy, but this is yet to be
determined.
* This buys us some flexibility about when/where we run ThinLTO, as well as
having it tailored to fit our needs for the time being.
* Finally this allows us to reuse some artifacts such as our `TargetMachine`
creation, where all our options we used today aren't necessarily supported by
upstream LLVM yet.
My hope is that we can get some experience with this copy/paste in tree and then
eventually upstream some work to LLVM itself to avoid the duplication while
still ensuring our needs are met. Otherwise I fear that maintaining these
bindings may be quite costly over the years with LLVM updates!
2017-07-23 17:14:38 +02:00
|
|
|
FunctionImporter Importer(Data->Index, Loader);
|
2020-06-26 03:52:41 +02:00
|
|
|
#endif
|
rustc: Implement ThinLTO
This commit is an implementation of LLVM's ThinLTO for consumption in rustc
itself. Currently today LTO works by merging all relevant LLVM modules into one
and then running optimization passes. "Thin" LTO operates differently by having
more sharded work and allowing parallelism opportunities between optimizing
codegen units. Further down the road Thin LTO also allows *incremental* LTO
which should enable even faster release builds without compromising on the
performance we have today.
This commit uses a `-Z thinlto` flag to gate whether ThinLTO is enabled. It then
also implements two forms of ThinLTO:
* In one mode we'll *only* perform ThinLTO over the codegen units produced in a
single compilation. That is, we won't load upstream rlibs, but we'll instead
just perform ThinLTO amongst all codegen units produced by the compiler for
the local crate. This is intended to emulate a desired end point where we have
codegen units turned on by default for all crates and ThinLTO allows us to do
this without performance loss.
* In anther mode, like full LTO today, we'll optimize all upstream dependencies
in "thin" mode. Unlike today, however, this LTO step is fully parallelized so
should finish much more quickly.
There's a good bit of comments about what the implementation is doing and where
it came from, but the tl;dr; is that currently most of the support here is
copied from upstream LLVM. This code duplication is done for a number of
reasons:
* Controlling parallelism means we can use the existing jobserver support to
avoid overloading machines.
* We will likely want a slightly different form of incremental caching which
integrates with our own incremental strategy, but this is yet to be
determined.
* This buys us some flexibility about when/where we run ThinLTO, as well as
having it tailored to fit our needs for the time being.
* Finally this allows us to reuse some artifacts such as our `TargetMachine`
creation, where all our options we used today aren't necessarily supported by
upstream LLVM yet.
My hope is that we can get some experience with this copy/paste in tree and then
eventually upstream some work to LLVM itself to avoid the duplication while
still ensuring our needs are met. Otherwise I fear that maintaining these
bindings may be quite costly over the years with LLVM updates!
2017-07-23 17:14:38 +02:00
|
|
|
Expected<bool> Result = Importer.importFunctions(Mod, ImportList);
|
|
|
|
if (!Result) {
|
|
|
|
LLVMRustSetLastError(toString(Result.takeError()).c_str());
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
return true;
|
|
|
|
}
|
|
|
|
|
2018-08-17 15:25:46 +02:00
|
|
|
extern "C" typedef void (*LLVMRustModuleNameCallback)(void*, // payload
|
|
|
|
const char*, // importing module name
|
|
|
|
const char*); // imported module name
|
|
|
|
|
|
|
|
// Calls `module_name_callback` for each module import done by ThinLTO.
|
|
|
|
// The callback is provided with regular null-terminated C strings.
|
|
|
|
extern "C" void
|
Use llvm::computeLTOCacheKey to determine post-ThinLTO CGU reuse
During incremental ThinLTO compilation, we attempt to re-use the
optimized (post-ThinLTO) bitcode file for a module if it is 'safe' to do
so.
Up until now, 'safe' has meant that the set of modules that our current
modules imports from/exports to is unchanged from the previous
compilation session. See PR #67020 and PR #71131 for more details.
However, this turns out be insufficient to guarantee that it's safe
to reuse the post-LTO module (i.e. that optimizing the pre-LTO module
would produce the same result). When LLVM optimizes a module during
ThinLTO, it may look at other information from the 'module index', such
as whether a (non-imported!) global variable is used. If this
information changes between compilation runs, we may end up re-using an
optimized module that (for example) had dead-code elimination run on a
function that is now used by another module.
Fortunately, LLVM implements its own ThinLTO module cache, which is used
when ThinLTO is performed by a linker plugin (e.g. when clang is used to
compile a C proect). Using this cache directly would require extensive
refactoring of our code - but fortunately for us, LLVM provides a
function that does exactly what we need.
The function `llvm::computeLTOCacheKey` is used to compute a SHA-1 hash
from all data that might influence the result of ThinLTO on a module.
In addition to the module imports/exports that we manually track, it
also hashes information about global variables (e.g. their liveness)
which might be used during optimization. By using this function, we
shouldn't have to worry about new LLVM passes breaking our module re-use
behavior.
In LLVM, the output of this function forms part of the filename used to
store the post-ThinLTO module. To keep our current filename structure
intact, this PR just writes out the mapping 'CGU name -> Hash' to a
file. To determine if a post-LTO module should be reused, we compare
hashes from the previous session.
This should unblock PR #75199 - by sheer chance, it seems to have hit
this issue due to the particular CGU partitioning and optimization
decisions that end up getting made.
2020-09-17 23:36:13 +02:00
|
|
|
LLVMRustGetThinLTOModules(const LLVMRustThinLTOData *data,
|
2018-08-17 15:25:46 +02:00
|
|
|
LLVMRustModuleNameCallback module_name_callback,
|
|
|
|
void* callback_payload) {
|
|
|
|
for (const auto& importing_module : data->ImportLists) {
|
|
|
|
const std::string importing_module_id = importing_module.getKey().str();
|
|
|
|
const auto& imports = importing_module.getValue();
|
|
|
|
for (const auto& imported_module : imports) {
|
|
|
|
const std::string imported_module_id = imported_module.getKey().str();
|
|
|
|
module_name_callback(callback_payload,
|
|
|
|
importing_module_id.c_str(),
|
|
|
|
imported_module_id.c_str());
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
rustc: Implement ThinLTO
This commit is an implementation of LLVM's ThinLTO for consumption in rustc
itself. Currently today LTO works by merging all relevant LLVM modules into one
and then running optimization passes. "Thin" LTO operates differently by having
more sharded work and allowing parallelism opportunities between optimizing
codegen units. Further down the road Thin LTO also allows *incremental* LTO
which should enable even faster release builds without compromising on the
performance we have today.
This commit uses a `-Z thinlto` flag to gate whether ThinLTO is enabled. It then
also implements two forms of ThinLTO:
* In one mode we'll *only* perform ThinLTO over the codegen units produced in a
single compilation. That is, we won't load upstream rlibs, but we'll instead
just perform ThinLTO amongst all codegen units produced by the compiler for
the local crate. This is intended to emulate a desired end point where we have
codegen units turned on by default for all crates and ThinLTO allows us to do
this without performance loss.
* In anther mode, like full LTO today, we'll optimize all upstream dependencies
in "thin" mode. Unlike today, however, this LTO step is fully parallelized so
should finish much more quickly.
There's a good bit of comments about what the implementation is doing and where
it came from, but the tl;dr; is that currently most of the support here is
copied from upstream LLVM. This code duplication is done for a number of
reasons:
* Controlling parallelism means we can use the existing jobserver support to
avoid overloading machines.
* We will likely want a slightly different form of incremental caching which
integrates with our own incremental strategy, but this is yet to be
determined.
* This buys us some flexibility about when/where we run ThinLTO, as well as
having it tailored to fit our needs for the time being.
* Finally this allows us to reuse some artifacts such as our `TargetMachine`
creation, where all our options we used today aren't necessarily supported by
upstream LLVM yet.
My hope is that we can get some experience with this copy/paste in tree and then
eventually upstream some work to LLVM itself to avoid the duplication while
still ensuring our needs are met. Otherwise I fear that maintaining these
bindings may be quite costly over the years with LLVM updates!
2017-07-23 17:14:38 +02:00
|
|
|
// This struct and various functions are sort of a hack right now, but the
|
|
|
|
// problem is that we've got in-memory LLVM modules after we generate and
|
|
|
|
// optimize all codegen-units for one compilation in rustc. To be compatible
|
|
|
|
// with the LTO support above we need to serialize the modules plus their
|
|
|
|
// ThinLTO summary into memory.
|
|
|
|
//
|
|
|
|
// This structure is basically an owned version of a serialize module, with
|
|
|
|
// a ThinLTO summary attached.
|
|
|
|
struct LLVMRustThinLTOBuffer {
|
|
|
|
std::string data;
|
|
|
|
};
|
|
|
|
|
|
|
|
extern "C" LLVMRustThinLTOBuffer*
|
|
|
|
LLVMRustThinLTOBufferCreate(LLVMModuleRef M) {
|
2019-12-31 14:08:25 +01:00
|
|
|
#if LLVM_VERSION_GE(10, 0)
|
|
|
|
auto Ret = std::make_unique<LLVMRustThinLTOBuffer>();
|
|
|
|
#else
|
rustc: Implement ThinLTO
This commit is an implementation of LLVM's ThinLTO for consumption in rustc
itself. Currently today LTO works by merging all relevant LLVM modules into one
and then running optimization passes. "Thin" LTO operates differently by having
more sharded work and allowing parallelism opportunities between optimizing
codegen units. Further down the road Thin LTO also allows *incremental* LTO
which should enable even faster release builds without compromising on the
performance we have today.
This commit uses a `-Z thinlto` flag to gate whether ThinLTO is enabled. It then
also implements two forms of ThinLTO:
* In one mode we'll *only* perform ThinLTO over the codegen units produced in a
single compilation. That is, we won't load upstream rlibs, but we'll instead
just perform ThinLTO amongst all codegen units produced by the compiler for
the local crate. This is intended to emulate a desired end point where we have
codegen units turned on by default for all crates and ThinLTO allows us to do
this without performance loss.
* In anther mode, like full LTO today, we'll optimize all upstream dependencies
in "thin" mode. Unlike today, however, this LTO step is fully parallelized so
should finish much more quickly.
There's a good bit of comments about what the implementation is doing and where
it came from, but the tl;dr; is that currently most of the support here is
copied from upstream LLVM. This code duplication is done for a number of
reasons:
* Controlling parallelism means we can use the existing jobserver support to
avoid overloading machines.
* We will likely want a slightly different form of incremental caching which
integrates with our own incremental strategy, but this is yet to be
determined.
* This buys us some flexibility about when/where we run ThinLTO, as well as
having it tailored to fit our needs for the time being.
* Finally this allows us to reuse some artifacts such as our `TargetMachine`
creation, where all our options we used today aren't necessarily supported by
upstream LLVM yet.
My hope is that we can get some experience with this copy/paste in tree and then
eventually upstream some work to LLVM itself to avoid the duplication while
still ensuring our needs are met. Otherwise I fear that maintaining these
bindings may be quite costly over the years with LLVM updates!
2017-07-23 17:14:38 +02:00
|
|
|
auto Ret = llvm::make_unique<LLVMRustThinLTOBuffer>();
|
2019-12-31 14:08:25 +01:00
|
|
|
#endif
|
rustc: Implement ThinLTO
This commit is an implementation of LLVM's ThinLTO for consumption in rustc
itself. Currently today LTO works by merging all relevant LLVM modules into one
and then running optimization passes. "Thin" LTO operates differently by having
more sharded work and allowing parallelism opportunities between optimizing
codegen units. Further down the road Thin LTO also allows *incremental* LTO
which should enable even faster release builds without compromising on the
performance we have today.
This commit uses a `-Z thinlto` flag to gate whether ThinLTO is enabled. It then
also implements two forms of ThinLTO:
* In one mode we'll *only* perform ThinLTO over the codegen units produced in a
single compilation. That is, we won't load upstream rlibs, but we'll instead
just perform ThinLTO amongst all codegen units produced by the compiler for
the local crate. This is intended to emulate a desired end point where we have
codegen units turned on by default for all crates and ThinLTO allows us to do
this without performance loss.
* In anther mode, like full LTO today, we'll optimize all upstream dependencies
in "thin" mode. Unlike today, however, this LTO step is fully parallelized so
should finish much more quickly.
There's a good bit of comments about what the implementation is doing and where
it came from, but the tl;dr; is that currently most of the support here is
copied from upstream LLVM. This code duplication is done for a number of
reasons:
* Controlling parallelism means we can use the existing jobserver support to
avoid overloading machines.
* We will likely want a slightly different form of incremental caching which
integrates with our own incremental strategy, but this is yet to be
determined.
* This buys us some flexibility about when/where we run ThinLTO, as well as
having it tailored to fit our needs for the time being.
* Finally this allows us to reuse some artifacts such as our `TargetMachine`
creation, where all our options we used today aren't necessarily supported by
upstream LLVM yet.
My hope is that we can get some experience with this copy/paste in tree and then
eventually upstream some work to LLVM itself to avoid the duplication while
still ensuring our needs are met. Otherwise I fear that maintaining these
bindings may be quite costly over the years with LLVM updates!
2017-07-23 17:14:38 +02:00
|
|
|
{
|
|
|
|
raw_string_ostream OS(Ret->data);
|
|
|
|
{
|
|
|
|
legacy::PassManager PM;
|
|
|
|
PM.add(createWriteThinLTOBitcodePass(OS));
|
|
|
|
PM.run(*unwrap(M));
|
|
|
|
}
|
|
|
|
}
|
|
|
|
return Ret.release();
|
|
|
|
}
|
|
|
|
|
|
|
|
extern "C" void
|
|
|
|
LLVMRustThinLTOBufferFree(LLVMRustThinLTOBuffer *Buffer) {
|
|
|
|
delete Buffer;
|
|
|
|
}
|
|
|
|
|
|
|
|
extern "C" const void*
|
|
|
|
LLVMRustThinLTOBufferPtr(const LLVMRustThinLTOBuffer *Buffer) {
|
|
|
|
return Buffer->data.data();
|
|
|
|
}
|
|
|
|
|
|
|
|
extern "C" size_t
|
|
|
|
LLVMRustThinLTOBufferLen(const LLVMRustThinLTOBuffer *Buffer) {
|
|
|
|
return Buffer->data.length();
|
|
|
|
}
|
|
|
|
|
|
|
|
// This is what we used to parse upstream bitcode for actual ThinLTO
|
|
|
|
// processing. We'll call this once per module optimized through ThinLTO, and
|
|
|
|
// it'll be called concurrently on many threads.
|
|
|
|
extern "C" LLVMModuleRef
|
2019-02-11 16:46:04 +01:00
|
|
|
LLVMRustParseBitcodeForLTO(LLVMContextRef Context,
|
|
|
|
const char *data,
|
|
|
|
size_t len,
|
|
|
|
const char *identifier) {
|
rustc: Implement ThinLTO
This commit is an implementation of LLVM's ThinLTO for consumption in rustc
itself. Currently today LTO works by merging all relevant LLVM modules into one
and then running optimization passes. "Thin" LTO operates differently by having
more sharded work and allowing parallelism opportunities between optimizing
codegen units. Further down the road Thin LTO also allows *incremental* LTO
which should enable even faster release builds without compromising on the
performance we have today.
This commit uses a `-Z thinlto` flag to gate whether ThinLTO is enabled. It then
also implements two forms of ThinLTO:
* In one mode we'll *only* perform ThinLTO over the codegen units produced in a
single compilation. That is, we won't load upstream rlibs, but we'll instead
just perform ThinLTO amongst all codegen units produced by the compiler for
the local crate. This is intended to emulate a desired end point where we have
codegen units turned on by default for all crates and ThinLTO allows us to do
this without performance loss.
* In anther mode, like full LTO today, we'll optimize all upstream dependencies
in "thin" mode. Unlike today, however, this LTO step is fully parallelized so
should finish much more quickly.
There's a good bit of comments about what the implementation is doing and where
it came from, but the tl;dr; is that currently most of the support here is
copied from upstream LLVM. This code duplication is done for a number of
reasons:
* Controlling parallelism means we can use the existing jobserver support to
avoid overloading machines.
* We will likely want a slightly different form of incremental caching which
integrates with our own incremental strategy, but this is yet to be
determined.
* This buys us some flexibility about when/where we run ThinLTO, as well as
having it tailored to fit our needs for the time being.
* Finally this allows us to reuse some artifacts such as our `TargetMachine`
creation, where all our options we used today aren't necessarily supported by
upstream LLVM yet.
My hope is that we can get some experience with this copy/paste in tree and then
eventually upstream some work to LLVM itself to avoid the duplication while
still ensuring our needs are met. Otherwise I fear that maintaining these
bindings may be quite costly over the years with LLVM updates!
2017-07-23 17:14:38 +02:00
|
|
|
StringRef Data(data, len);
|
|
|
|
MemoryBufferRef Buffer(Data, identifier);
|
|
|
|
unwrap(Context)->enableDebugTypeODRUniquing();
|
|
|
|
Expected<std::unique_ptr<Module>> SrcOrError =
|
|
|
|
parseBitcodeFile(Buffer, *unwrap(Context));
|
|
|
|
if (!SrcOrError) {
|
|
|
|
LLVMRustSetLastError(toString(SrcOrError.takeError()).c_str());
|
|
|
|
return nullptr;
|
|
|
|
}
|
|
|
|
return wrap(std::move(*SrcOrError).release());
|
|
|
|
}
|
|
|
|
|
2020-04-23 20:45:55 +02:00
|
|
|
// Find the bitcode section in the object file data and return it as a slice.
|
|
|
|
// Fail if the bitcode section is present but empty.
|
|
|
|
//
|
|
|
|
// On success, the return value is the pointer to the start of the slice and
|
|
|
|
// `out_len` is filled with the (non-zero) length. On failure, the return value
|
|
|
|
// is `nullptr` and `out_len` is set to zero.
|
|
|
|
extern "C" const char*
|
|
|
|
LLVMRustGetBitcodeSliceFromObjectData(const char *data,
|
|
|
|
size_t len,
|
|
|
|
size_t *out_len) {
|
|
|
|
*out_len = 0;
|
|
|
|
|
|
|
|
StringRef Data(data, len);
|
|
|
|
MemoryBufferRef Buffer(Data, ""); // The id is unused.
|
|
|
|
|
|
|
|
Expected<MemoryBufferRef> BitcodeOrError =
|
|
|
|
object::IRObjectFile::findBitcodeInMemBuffer(Buffer);
|
|
|
|
if (!BitcodeOrError) {
|
|
|
|
LLVMRustSetLastError(toString(BitcodeOrError.takeError()).c_str());
|
|
|
|
return nullptr;
|
|
|
|
}
|
|
|
|
|
|
|
|
*out_len = BitcodeOrError->getBufferSize();
|
|
|
|
return BitcodeOrError->getBufferStart();
|
|
|
|
}
|
|
|
|
|
2017-12-16 17:20:54 +01:00
|
|
|
// Rewrite all `DICompileUnit` pointers to the `DICompileUnit` specified. See
|
|
|
|
// the comment in `back/lto.rs` for why this exists.
|
|
|
|
extern "C" void
|
|
|
|
LLVMRustThinLTOGetDICompileUnit(LLVMModuleRef Mod,
|
|
|
|
DICompileUnit **A,
|
|
|
|
DICompileUnit **B) {
|
|
|
|
Module *M = unwrap(Mod);
|
|
|
|
DICompileUnit **Cur = A;
|
|
|
|
DICompileUnit **Next = B;
|
|
|
|
for (DICompileUnit *CU : M->debug_compile_units()) {
|
|
|
|
*Cur = CU;
|
|
|
|
Cur = Next;
|
|
|
|
Next = nullptr;
|
|
|
|
if (Cur == nullptr)
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
// Rewrite all `DICompileUnit` pointers to the `DICompileUnit` specified. See
|
|
|
|
// the comment in `back/lto.rs` for why this exists.
|
|
|
|
extern "C" void
|
|
|
|
LLVMRustThinLTOPatchDICompileUnit(LLVMModuleRef Mod, DICompileUnit *Unit) {
|
|
|
|
Module *M = unwrap(Mod);
|
|
|
|
|
|
|
|
// If the original source module didn't have a `DICompileUnit` then try to
|
|
|
|
// merge all the existing compile units. If there aren't actually any though
|
|
|
|
// then there's not much for us to do so return.
|
|
|
|
if (Unit == nullptr) {
|
|
|
|
for (DICompileUnit *CU : M->debug_compile_units()) {
|
|
|
|
Unit = CU;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
if (Unit == nullptr)
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
// Use LLVM's built-in `DebugInfoFinder` to find a bunch of debuginfo and
|
2020-06-26 03:52:41 +02:00
|
|
|
// process it recursively. Note that we used to specifically iterate over
|
|
|
|
// instructions to ensure we feed everything into it, but `processModule`
|
|
|
|
// started doing this the same way in LLVM 7 (commit d769eb36ab2b8).
|
2017-12-16 17:20:54 +01:00
|
|
|
DebugInfoFinder Finder;
|
|
|
|
Finder.processModule(*M);
|
|
|
|
|
|
|
|
// After we've found all our debuginfo, rewrite all subprograms to point to
|
|
|
|
// the same `DICompileUnit`.
|
|
|
|
for (auto &F : Finder.subprograms()) {
|
|
|
|
F->replaceUnit(Unit);
|
|
|
|
}
|
|
|
|
|
|
|
|
// Erase any other references to other `DICompileUnit` instances, the verifier
|
|
|
|
// will later ensure that we don't actually have any other stale references to
|
|
|
|
// worry about.
|
|
|
|
auto *MD = M->getNamedMetadata("llvm.dbg.cu");
|
|
|
|
MD->clearOperands();
|
|
|
|
MD->addOperand(Unit);
|
|
|
|
}
|
Use llvm::computeLTOCacheKey to determine post-ThinLTO CGU reuse
During incremental ThinLTO compilation, we attempt to re-use the
optimized (post-ThinLTO) bitcode file for a module if it is 'safe' to do
so.
Up until now, 'safe' has meant that the set of modules that our current
modules imports from/exports to is unchanged from the previous
compilation session. See PR #67020 and PR #71131 for more details.
However, this turns out be insufficient to guarantee that it's safe
to reuse the post-LTO module (i.e. that optimizing the pre-LTO module
would produce the same result). When LLVM optimizes a module during
ThinLTO, it may look at other information from the 'module index', such
as whether a (non-imported!) global variable is used. If this
information changes between compilation runs, we may end up re-using an
optimized module that (for example) had dead-code elimination run on a
function that is now used by another module.
Fortunately, LLVM implements its own ThinLTO module cache, which is used
when ThinLTO is performed by a linker plugin (e.g. when clang is used to
compile a C proect). Using this cache directly would require extensive
refactoring of our code - but fortunately for us, LLVM provides a
function that does exactly what we need.
The function `llvm::computeLTOCacheKey` is used to compute a SHA-1 hash
from all data that might influence the result of ThinLTO on a module.
In addition to the module imports/exports that we manually track, it
also hashes information about global variables (e.g. their liveness)
which might be used during optimization. By using this function, we
shouldn't have to worry about new LLVM passes breaking our module re-use
behavior.
In LLVM, the output of this function forms part of the filename used to
store the post-ThinLTO module. To keep our current filename structure
intact, this PR just writes out the mapping 'CGU name -> Hash' to a
file. To determine if a post-LTO module should be reused, we compare
hashes from the previous session.
This should unblock PR #75199 - by sheer chance, it seems to have hit
this issue due to the particular CGU partitioning and optimization
decisions that end up getting made.
2020-09-17 23:36:13 +02:00
|
|
|
|
|
|
|
// Computes the LTO cache key for the provided 'ModId' in the given 'Data',
|
|
|
|
// storing the result in 'KeyOut'.
|
|
|
|
// Currently, this cache key is a SHA-1 hash of anything that could affect
|
|
|
|
// the result of optimizing this module (e.g. module imports, exports, liveness
|
|
|
|
// of access globals, etc).
|
|
|
|
// The precise details are determined by LLVM in `computeLTOCacheKey`, which is
|
|
|
|
// used during the normal linker-plugin incremental thin-LTO process.
|
|
|
|
extern "C" void
|
|
|
|
LLVMRustComputeLTOCacheKey(RustStringRef KeyOut, const char *ModId, LLVMRustThinLTOData *Data) {
|
|
|
|
SmallString<40> Key;
|
|
|
|
llvm::lto::Config conf;
|
|
|
|
const auto &ImportList = Data->ImportLists.lookup(ModId);
|
|
|
|
const auto &ExportList = Data->ExportLists.lookup(ModId);
|
|
|
|
const auto &ResolvedODR = Data->ResolvedODR.lookup(ModId);
|
|
|
|
const auto &DefinedGlobals = Data->ModuleToDefinedGVSummaries.lookup(ModId);
|
|
|
|
std::set<GlobalValue::GUID> CfiFunctionDefs;
|
|
|
|
std::set<GlobalValue::GUID> CfiFunctionDecls;
|
|
|
|
|
|
|
|
// Based on the 'InProcessThinBackend' constructor in LLVM
|
|
|
|
for (auto &Name : Data->Index.cfiFunctionDefs())
|
|
|
|
CfiFunctionDefs.insert(
|
|
|
|
GlobalValue::getGUID(GlobalValue::dropLLVMManglingEscape(Name)));
|
|
|
|
for (auto &Name : Data->Index.cfiFunctionDecls())
|
|
|
|
CfiFunctionDecls.insert(
|
|
|
|
GlobalValue::getGUID(GlobalValue::dropLLVMManglingEscape(Name)));
|
|
|
|
|
|
|
|
llvm::computeLTOCacheKey(Key, conf, Data->Index, ModId,
|
|
|
|
ImportList, ExportList, ResolvedODR, DefinedGlobals, CfiFunctionDefs, CfiFunctionDecls
|
|
|
|
);
|
|
|
|
|
|
|
|
LLVMRustStringWriteImpl(KeyOut, Key.c_str(), Key.size());
|
|
|
|
}
|