ARM64.cpp | Explore in Territory

//===- ARM64.cpp ----------------------------------------------------------===//
//
// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
// See https://llvm.org/LICENSE.txt for license information.
// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
//
//===----------------------------------------------------------------------===//

#include "Arch/ARM64Common.h"
#include "InputFiles.h"
#include "Symbols.h"
#include "SyntheticSections.h"
#include "Target.h"

#include "lld/Common/ErrorHandler.h"
#include "mach-o/compact_unwind_encoding.h"
#include "llvm/ADT/SmallVector.h"
#include "llvm/ADT/StringRef.h"
#include "llvm/BinaryFormat/MachO.h"
#include "llvm/Support/Endian.h"
#include "llvm/Support/LEB128.h"
#include "llvm/Support/MathExtras.h"

usingnamespacellvm;
usingnamespacellvm::MachO;
usingnamespacellvm::support::endian;
usingnamespacelld;
usingnamespacelld::macho;

namespace {

struct ARM64 : ARM64Common { … };

} // namespace

// Random notes on reloc types:
// ADDEND always pairs with BRANCH26, PAGE21, or PAGEOFF12
// POINTER_TO_GOT: ld64 supports a 4-byte pc-relative form as well as an 8-byte
// absolute version of this relocation. The semantics of the absolute relocation
// are weird -- it results in the value of the GOT slot being written, instead
// of the address. Let's not support it unless we find a real-world use case.
static constexpr std::array<RelocAttrs, 11> relocAttrsArray{ … };

static constexpr uint32_t stubCode[] = …;

void ARM64::writeStub(uint8_t *buf8, const Symbol &sym,
                      uint64_t pointerVA) const { … }

static constexpr uint32_t stubHelperHeaderCode[] = …;

void ARM64::writeStubHelperHeader(uint8_t *buf8) const { … }

static constexpr uint32_t stubHelperEntryCode[] = …;

void ARM64::writeStubHelperEntry(uint8_t *buf8, const Symbol &sym,
                                 uint64_t entryVA) const { … }

static constexpr uint32_t objcStubsFastCode[] = …;

static constexpr uint32_t objcStubsSmallCode[] = …;

void ARM64::writeObjCMsgSendStub(uint8_t *buf, Symbol *sym, uint64_t stubsAddr,
                                 uint64_t &stubOffset, uint64_t selrefVA,
                                 Symbol *objcMsgSend) const { … }

// A thunk is the relaxed variation of stubCode. We don't need the
// extra indirection through a lazy pointer because the target address
// is known at link time.
static constexpr uint32_t thunkCode[] = …;

void ARM64::populateThunk(InputSection *thunk, Symbol *funcSym) { … }

ARM64::ARM64() : … { … }

namespace {
struct Adrp { … };

struct Add { … };

enum ExtendType { … };

struct Ldr { … };
} // namespace

static bool parseAdrp(uint32_t insn, Adrp &adrp) { … }

static bool parseAdd(uint32_t insn, Add &add) { … }

static bool parseLdr(uint32_t insn, Ldr &ldr) { … }

static bool isValidAdrOffset(int32_t delta) { … }

static void writeAdr(void *loc, uint32_t dest, int32_t delta) { … }

static void writeNop(void *loc) { … }

static bool isLiteralLdrEligible(const Ldr &ldr) { … }

static void writeLiteralLdr(void *loc, const Ldr &ldr) { … }

static bool isImmediateLdrEligible(const Ldr &ldr) { … }

static void writeImmediateLdr(void *loc, const Ldr &ldr) { … }

// Transforms a pair of adrp+add instructions into an adr instruction if the
// target is within the +/- 1 MiB range allowed by the adr's 21 bit signed
// immediate offset.
//
//   adrp xN, _foo@PAGE
//   add  xM, xN, _foo@PAGEOFF
// ->
//   adr  xM, _foo
//   nop
static void applyAdrpAdd(uint8_t *buf, const ConcatInputSection *isec,
                         uint64_t offset1, uint64_t offset2) { … }

// Transforms two adrp instructions into a single adrp if their referent
// addresses are located on the same 4096 byte page.
//
//   adrp xN, _foo@PAGE
//   adrp xN, _bar@PAGE
// ->
//   adrp xN, _foo@PAGE
//   nop
static void applyAdrpAdrp(uint8_t *buf, const ConcatInputSection *isec,
                          uint64_t offset1, uint64_t offset2) { … }

// Transforms a pair of adrp+ldr (immediate) instructions into an ldr (literal)
// load from a PC-relative address if it is 4-byte aligned and within +/- 1 MiB,
// as ldr can encode a signed 19-bit offset that gets multiplied by 4.
//
//   adrp xN, _foo@PAGE
//   ldr  xM, [xN, _foo@PAGEOFF]
// ->
//   nop
//   ldr  xM, _foo
static void applyAdrpLdr(uint8_t *buf, const ConcatInputSection *isec,
                         uint64_t offset1, uint64_t offset2) { … }

// GOT loads are emitted by the compiler as a pair of adrp and ldr instructions,
// but they may be changed to adrp+add by relaxGotLoad(). This hint performs
// the AdrpLdr or AdrpAdd transformation depending on whether it was relaxed.
static void applyAdrpLdrGot(uint8_t *buf, const ConcatInputSection *isec,
                            uint64_t offset1, uint64_t offset2) { … }

// Optimizes an adrp+add+ldr sequence used for loading from a local symbol's
// address by loading directly if it's close enough, or to an adrp(p)+ldr
// sequence if it's not.
//
//   adrp x0, _foo@PAGE
//   add  x1, x0, _foo@PAGEOFF
//   ldr  x2, [x1, #off]
static void applyAdrpAddLdr(uint8_t *buf, const ConcatInputSection *isec,
                            uint64_t offset1, uint64_t offset2,
                            uint64_t offset3) { … }

// Relaxes a GOT-indirect load.
// If the referenced symbol is external and its GOT entry is within +/- 1 MiB,
// the GOT entry can be loaded with a single literal ldr instruction.
// If the referenced symbol is local and thus has been relaxed to adrp+add+ldr,
// we perform the AdrpAddLdr transformation.
static void applyAdrpLdrGotLdr(uint8_t *buf, const ConcatInputSection *isec,
                               uint64_t offset1, uint64_t offset2,
                               uint64_t offset3) { … }

static uint64_t readValue(const uint8_t *&ptr, const uint8_t *end) { … }

template <typename Callback>
static void forEachHint(ArrayRef<uint8_t> data, Callback callback) { … }

// On RISC architectures like arm64, materializing a memory address generally
// takes multiple instructions. If the referenced symbol is located close enough
// in memory, fewer instructions are needed.
//
// Linker optimization hints record where addresses are computed. After
// addresses have been assigned, if possible, we change them to a shorter
// sequence of instructions. The size of the binary is not modified; the
// eliminated instructions are replaced with NOPs. This still leads to faster
// code as the CPU can skip over NOPs quickly.
//
// LOHs are specified by the LC_LINKER_OPTIMIZATION_HINTS load command, which
// points to a sequence of ULEB128-encoded numbers. Each entry specifies a
// transformation kind, and 2 or 3 addresses where the instructions are located.
void ARM64::applyOptimizationHints(uint8_t *outBuf, const ObjFile &obj) const { … }

TargetInfo *macho::createARM64TargetInfo() { … }
llvm/lld/MachO/Arch/ARM64.cpp
