mirror of
https://github.com/RPCS3/llvm-mirror.git
synced 2024-10-22 12:33:33 +02:00
b47927165c
In testing, we've found yet another miscompile caused by the new tables. And this one is even less clear how to fix (we could teach it to fold a 16-bit load instead of the 32-bit load it wants, or block folding entirely). Also, the approach to excluding instructions seems increasingly to not scale well. I have left a more detailed analysis on the review log for the original patch (https://reviews.llvm.org/D32684) along with suggested path forward. I will land an additional test case that I wrote which covers the code that was miscompiling (folding into the output of `pextrw`) in a subsequent commit to keep this a pure revert. For each commit reverted here, I've restricted the revert to the non-test code touching the x86 fold table emission until the last commit where I did revert the test updates. This means the *new* test cases added for `insertps` and `xchg` remain untouched (and continue to pass). Reverted commits: r304540: [X86] Don't fold into memory operands into insertps in the ... r304347: [TableGen] Adapt more places to getValueAsString now ... r304163: [X86] Don't fold away the memory operand of an xchg. r304123: Don't capture a temporary std::string in a StringRef. r304122: Resubmit "[X86] Adding new LLVM TableGen backend that ..." Original commit was in r304088, and after a string of fixes was reverted previously in r304121 to fix build bots, and then re-landed in r304122. llvm-svn: 304762
90 lines
4.0 KiB
C++
90 lines
4.0 KiB
C++
//===- TableGenBackends.h - Declarations for LLVM TableGen Backends -------===//
|
|
//
|
|
// The LLVM Compiler Infrastructure
|
|
//
|
|
// This file is distributed under the University of Illinois Open Source
|
|
// License. See LICENSE.TXT for details.
|
|
//
|
|
//===----------------------------------------------------------------------===//
|
|
//
|
|
// This file contains the declarations for all of the LLVM TableGen
|
|
// backends. A "TableGen backend" is just a function. See below for a
|
|
// precise description.
|
|
//
|
|
//===----------------------------------------------------------------------===//
|
|
|
|
#ifndef LLVM_UTILS_TABLEGEN_TABLEGENBACKENDS_H
|
|
#define LLVM_UTILS_TABLEGEN_TABLEGENBACKENDS_H
|
|
|
|
// A TableGen backend is a function that looks like
|
|
//
|
|
// EmitFoo(RecordKeeper &RK, raw_ostream &OS /*, anything else you need */ )
|
|
//
|
|
// What you do inside of that function is up to you, but it will usually
|
|
// involve generating C++ code to the provided raw_ostream.
|
|
//
|
|
// The RecordKeeper is just a top-level container for an in-memory
|
|
// representation of the data encoded in the TableGen file. What a TableGen
|
|
// backend does is walk around that in-memory representation and generate
|
|
// stuff based on the information it contains.
|
|
//
|
|
// The in-memory representation is a node-graph (think of it like JSON but
|
|
// with a richer ontology of types), where the nodes are subclasses of
|
|
// Record. The methods `getClass`, `getDef` are the basic interface to
|
|
// access the node-graph. RecordKeeper also provides a handy method
|
|
// `getAllDerivedDefinitions`. Consult "include/llvm/TableGen/Record.h" for
|
|
// the exact interfaces provided by Record's and RecordKeeper.
|
|
//
|
|
// A common pattern for TableGen backends is for the EmitFoo function to
|
|
// instantiate a class which holds some context for the generation process,
|
|
// and then have most of the work happen in that class's methods. This
|
|
// pattern partly has historical roots in the previous TableGen backend API
|
|
// that involved a class and an invocation like `FooEmitter(RK).run(OS)`.
|
|
//
|
|
// Remember to wrap private things in an anonymous namespace. For most
|
|
// backends, this means that the EmitFoo function is the only thing not in
|
|
// the anonymous namespace.
|
|
|
|
|
|
// FIXME: Reorganize TableGen so that build dependencies can be more
|
|
// accurately expressed. Currently, touching any of the emitters (or
|
|
// anything that they transitively depend on) causes everything dependent
|
|
// on TableGen to be rebuilt (this includes all the targets!). Perhaps have
|
|
// a standalone TableGen binary and have the backends be loadable modules
|
|
// of some sort; then the dependency could be expressed as being on the
|
|
// module, and all the modules would have a common dependency on the
|
|
// TableGen binary with as few dependencies as possible on the rest of
|
|
// LLVM.
|
|
|
|
|
|
namespace llvm {
|
|
|
|
class raw_ostream;
|
|
class RecordKeeper;
|
|
|
|
void EmitIntrinsics(RecordKeeper &RK, raw_ostream &OS, bool TargetOnly = false);
|
|
void EmitAsmMatcher(RecordKeeper &RK, raw_ostream &OS);
|
|
void EmitAsmWriter(RecordKeeper &RK, raw_ostream &OS);
|
|
void EmitCallingConv(RecordKeeper &RK, raw_ostream &OS);
|
|
void EmitCodeEmitter(RecordKeeper &RK, raw_ostream &OS);
|
|
void EmitDAGISel(RecordKeeper &RK, raw_ostream &OS);
|
|
void EmitDFAPacketizer(RecordKeeper &RK, raw_ostream &OS);
|
|
void EmitDisassembler(RecordKeeper &RK, raw_ostream &OS);
|
|
void EmitFastISel(RecordKeeper &RK, raw_ostream &OS);
|
|
void EmitInstrInfo(RecordKeeper &RK, raw_ostream &OS);
|
|
void EmitPseudoLowering(RecordKeeper &RK, raw_ostream &OS);
|
|
void EmitRegisterInfo(RecordKeeper &RK, raw_ostream &OS);
|
|
void EmitSubtarget(RecordKeeper &RK, raw_ostream &OS);
|
|
void EmitMapTable(RecordKeeper &RK, raw_ostream &OS);
|
|
void EmitOptParser(RecordKeeper &RK, raw_ostream &OS);
|
|
void EmitCTags(RecordKeeper &RK, raw_ostream &OS);
|
|
void EmitAttributes(RecordKeeper &RK, raw_ostream &OS);
|
|
void EmitSearchableTables(RecordKeeper &RK, raw_ostream &OS);
|
|
void EmitGlobalISel(RecordKeeper &RK, raw_ostream &OS);
|
|
void EmitX86EVEX2VEXTables(RecordKeeper &RK, raw_ostream &OS);
|
|
void EmitRegisterBank(RecordKeeper &RK, raw_ostream &OS);
|
|
|
|
} // End llvm namespace
|
|
|
|
#endif
|