llvm-project/lld/MachO/LinkerOptimizationHints.cpp
Daniel Bertalan 43f10639a1
[lld-macho] Enable Linker Optimization Hints pass for arm64_32 (#148964)
The backend emits `.loh` directives for arm64_32 as well. Our pass
already handles 32-bit pointer loads correctly (there was an extraneous
sanity check for 8-byte pointer sizes, I removed that here), so we can
enable them for all arm64 subtargets, including our upcoming arm64e
support.
2025-07-16 21:29:48 +02:00

524 lines
16 KiB
C++

//===- LinkerOptimizationHints.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 "LinkerOptimizationHints.h"
#include "Arch/ARM64Common.h"
#include "lld/Common/ErrorHandler.h"
#include "llvm/ADT/ArrayRef.h"
#include "llvm/BinaryFormat/MachO.h"
#include "llvm/Support/Endian.h"
#include "llvm/Support/LEB128.h"
#include "llvm/Support/MathExtras.h"
using namespace llvm;
using namespace llvm::MachO;
using namespace llvm::support::endian;
using namespace lld;
using namespace lld::macho;
namespace {
struct Adrp {
uint32_t destRegister;
int64_t addend;
};
struct Add {
uint8_t destRegister;
uint8_t srcRegister;
uint32_t addend;
};
enum ExtendType { ZeroExtend = 1, Sign64 = 2, Sign32 = 3 };
struct Ldr {
uint8_t destRegister;
uint8_t baseRegister;
uint8_t p2Size;
bool isFloat;
ExtendType extendType;
int64_t offset;
};
} // namespace
static bool parseAdrp(uint32_t insn, Adrp &adrp) {
if ((insn & 0x9f000000) != 0x90000000)
return false;
adrp.destRegister = insn & 0x1f;
uint64_t immHi = (insn >> 5) & 0x7ffff;
uint64_t immLo = (insn >> 29) & 0x3;
adrp.addend = SignExtend64<21>(immLo | (immHi << 2)) * 4096;
return true;
}
static bool parseAdd(uint32_t insn, Add &add) {
if ((insn & 0xffc00000) != 0x91000000)
return false;
add.destRegister = insn & 0x1f;
add.srcRegister = (insn >> 5) & 0x1f;
add.addend = (insn >> 10) & 0xfff;
return true;
}
static bool parseLdr(uint32_t insn, Ldr &ldr) {
ldr.destRegister = insn & 0x1f;
ldr.baseRegister = (insn >> 5) & 0x1f;
uint8_t size = insn >> 30;
uint8_t opc = (insn >> 22) & 3;
if ((insn & 0x3fc00000) == 0x39400000) {
// LDR (immediate), LDRB (immediate), LDRH (immediate)
ldr.p2Size = size;
ldr.extendType = ZeroExtend;
ldr.isFloat = false;
} else if ((insn & 0x3f800000) == 0x39800000) {
// LDRSB (immediate), LDRSH (immediate), LDRSW (immediate)
ldr.p2Size = size;
ldr.extendType = static_cast<ExtendType>(opc);
ldr.isFloat = false;
} else if ((insn & 0x3f400000) == 0x3d400000) {
// LDR (immediate, SIMD&FP)
ldr.extendType = ZeroExtend;
ldr.isFloat = true;
if (opc == 1)
ldr.p2Size = size;
else if (size == 0 && opc == 3)
ldr.p2Size = 4;
else
return false;
} else {
return false;
}
ldr.offset = ((insn >> 10) & 0xfff) << ldr.p2Size;
return true;
}
static bool isValidAdrOffset(int32_t delta) { return isInt<21>(delta); }
static void writeAdr(void *loc, uint32_t dest, int32_t delta) {
assert(isValidAdrOffset(delta));
uint32_t opcode = 0x10000000;
uint32_t immHi = (delta & 0x001ffffc) << 3;
uint32_t immLo = (delta & 0x00000003) << 29;
write32le(loc, opcode | immHi | immLo | dest);
}
static void writeNop(void *loc) { write32le(loc, 0xd503201f); }
static bool isLiteralLdrEligible(const Ldr &ldr) {
return ldr.p2Size > 1 && isShiftedInt<19, 2>(ldr.offset);
}
static void writeLiteralLdr(void *loc, const Ldr &ldr) {
assert(isLiteralLdrEligible(ldr));
uint32_t imm19 = (ldr.offset / 4 & maskTrailingOnes<uint32_t>(19)) << 5;
uint32_t opcode;
switch (ldr.p2Size) {
case 2:
if (ldr.isFloat)
opcode = 0x1c000000;
else
opcode = ldr.extendType == Sign64 ? 0x98000000 : 0x18000000;
break;
case 3:
opcode = ldr.isFloat ? 0x5c000000 : 0x58000000;
break;
case 4:
opcode = 0x9c000000;
break;
default:
llvm_unreachable("Invalid literal ldr size");
}
write32le(loc, opcode | imm19 | ldr.destRegister);
}
static bool isImmediateLdrEligible(const Ldr &ldr) {
// Note: We deviate from ld64's behavior, which converts to immediate loads
// only if ldr.offset < 4096, even though the offset is divided by the load's
// size in the 12-bit immediate operand. Only the unsigned offset variant is
// supported.
uint32_t size = 1 << ldr.p2Size;
return ldr.offset >= 0 && (ldr.offset % size) == 0 &&
isUInt<12>(ldr.offset >> ldr.p2Size);
}
static void writeImmediateLdr(void *loc, const Ldr &ldr) {
assert(isImmediateLdrEligible(ldr));
uint32_t opcode = 0x39000000;
if (ldr.isFloat) {
opcode |= 0x04000000;
assert(ldr.extendType == ZeroExtend);
}
opcode |= ldr.destRegister;
opcode |= ldr.baseRegister << 5;
uint8_t size, opc;
if (ldr.p2Size == 4) {
size = 0;
opc = 3;
} else {
opc = ldr.extendType;
size = ldr.p2Size;
}
uint32_t immBits = ldr.offset >> ldr.p2Size;
write32le(loc, opcode | (immBits << 10) | (opc << 22) | (size << 30));
}
// 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 bool applyAdrpAdd(uint8_t *buf, const ConcatInputSection *isec,
uint64_t offset1, uint64_t offset2) {
uint32_t ins1 = read32le(buf + offset1);
uint32_t ins2 = read32le(buf + offset2);
Adrp adrp;
Add add;
if (!parseAdrp(ins1, adrp) || !parseAdd(ins2, add))
return false;
if (adrp.destRegister != add.srcRegister)
return false;
uint64_t addr1 = isec->getVA() + offset1;
uint64_t referent = lld::macho::pageBits(addr1) + adrp.addend + add.addend;
int64_t delta = referent - addr1;
if (!isValidAdrOffset(delta))
return false;
writeAdr(buf + offset1, add.destRegister, delta);
writeNop(buf + offset2);
return true;
}
// 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) {
uint32_t ins1 = read32le(buf + offset1);
uint32_t ins2 = read32le(buf + offset2);
Adrp adrp1, adrp2;
if (!parseAdrp(ins1, adrp1) || !parseAdrp(ins2, adrp2))
return;
if (adrp1.destRegister != adrp2.destRegister)
return;
uint64_t page1 = pageBits(offset1 + isec->getVA()) + adrp1.addend;
uint64_t page2 = pageBits(offset2 + isec->getVA()) + adrp2.addend;
if (page1 != page2)
return;
writeNop(buf + 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) {
uint32_t ins1 = read32le(buf + offset1);
uint32_t ins2 = read32le(buf + offset2);
Adrp adrp;
Ldr ldr;
if (!parseAdrp(ins1, adrp) || !parseLdr(ins2, ldr))
return;
if (adrp.destRegister != ldr.baseRegister)
return;
uint64_t addr1 = isec->getVA() + offset1;
uint64_t addr2 = isec->getVA() + offset2;
uint64_t referent = pageBits(addr1) + adrp.addend + ldr.offset;
ldr.offset = referent - addr2;
if (!isLiteralLdrEligible(ldr))
return;
writeNop(buf + offset1);
writeLiteralLdr(buf + offset2, ldr);
}
// 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) {
uint32_t ins2 = read32le(buf + offset2);
Add add;
Ldr ldr;
if (parseAdd(ins2, add))
applyAdrpAdd(buf, isec, offset1, offset2);
else if (parseLdr(ins2, ldr))
applyAdrpLdr(buf, isec, offset1, 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) {
uint32_t ins1 = read32le(buf + offset1);
uint32_t ins2 = read32le(buf + offset2);
uint32_t ins3 = read32le(buf + offset3);
Adrp adrp;
Add add;
Ldr ldr;
if (!parseAdrp(ins1, adrp) || !parseAdd(ins2, add) || !parseLdr(ins3, ldr))
return;
if (adrp.destRegister != add.srcRegister)
return;
if (add.destRegister != ldr.baseRegister)
return;
// Load from the target address directly.
// nop
// nop
// ldr x2, [_foo + #off]
uint64_t addr1 = isec->getVA() + offset1;
uint64_t addr3 = isec->getVA() + offset3;
uint64_t referent = pageBits(addr1) + adrp.addend + add.addend;
Ldr literalLdr = ldr;
literalLdr.offset += referent - addr3;
if (isLiteralLdrEligible(literalLdr)) {
writeNop(buf + offset1);
writeNop(buf + offset2);
writeLiteralLdr(buf + offset3, literalLdr);
return;
}
if (applyAdrpAdd(buf, isec, offset1, offset2))
return;
// Move the target's page offset into the ldr's immediate offset.
// adrp x0, _foo@PAGE
// nop
// ldr x2, [x0, _foo@PAGEOFF + #off]
Ldr immediateLdr = ldr;
immediateLdr.baseRegister = adrp.destRegister;
immediateLdr.offset += add.addend;
if (isImmediateLdrEligible(immediateLdr)) {
writeNop(buf + offset2);
writeImmediateLdr(buf + offset3, immediateLdr);
return;
}
}
// 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) {
uint32_t ins2 = read32le(buf + offset2);
Add add;
Ldr ldr2;
if (parseAdd(ins2, add)) {
applyAdrpAddLdr(buf, isec, offset1, offset2, offset3);
} else if (parseLdr(ins2, ldr2)) {
// adrp x1, _foo@GOTPAGE
// ldr x2, [x1, _foo@GOTPAGEOFF]
// ldr x3, [x2, #off]
uint32_t ins3 = read32le(buf + offset3);
Ldr ldr3;
if (!parseLdr(ins3, ldr3))
return;
if (ldr3.baseRegister != ldr2.destRegister)
return;
applyAdrpLdr(buf, isec, offset1, offset2);
}
}
template <typename Callback>
static void forEachHint(ArrayRef<uint8_t> data, Callback callback) {
std::array<uint64_t, 3> args;
auto readNext = [&]() -> uint64_t {
unsigned int n = 0;
uint64_t value = decodeULEB128(data.data(), &n, data.end());
data = data.drop_front(n);
return value;
};
while (!data.empty()) {
uint64_t type = readNext();
if (type == 0)
break;
uint64_t argCount = readNext();
for (unsigned i = 0; i < argCount; ++i) {
uint64_t arg = readNext();
if (i < 3)
args[i] = arg;
}
// All known LOH types as of 2022-09 have 3 or fewer arguments; skip others.
if (argCount > 3)
continue;
callback(type, ArrayRef(args.data(), argCount));
}
}
// 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 macho::applyOptimizationHints(uint8_t *outBuf, const ObjFile &obj) {
ArrayRef<uint8_t> data = obj.getOptimizationHints();
if (data.empty())
return;
const ConcatInputSection *section = nullptr;
uint64_t sectionAddr = 0;
uint8_t *buf = nullptr;
auto findSection = [&](uint64_t addr) {
if (section && addr >= sectionAddr &&
addr < sectionAddr + section->getSize())
return true;
if (obj.sections.empty())
return false;
auto secIt = std::prev(llvm::upper_bound(
obj.sections, addr,
[](uint64_t off, const Section *sec) { return off < sec->addr; }));
const Section *sec = *secIt;
if (sec->subsections.empty())
return false;
auto subsecIt = std::prev(llvm::upper_bound(
sec->subsections, addr - sec->addr,
[](uint64_t off, Subsection subsec) { return off < subsec.offset; }));
const Subsection &subsec = *subsecIt;
const ConcatInputSection *isec =
dyn_cast_or_null<ConcatInputSection>(subsec.isec);
if (!isec || isec->shouldOmitFromOutput())
return false;
section = isec;
sectionAddr = subsec.offset + sec->addr;
buf = outBuf + section->outSecOff + section->parent->fileOff;
return true;
};
auto isValidOffset = [&](uint64_t offset) {
if (offset < sectionAddr || offset >= sectionAddr + section->getSize()) {
error(toString(&obj) +
": linker optimization hint spans multiple sections");
return false;
}
return true;
};
bool hasAdrpAdrp = false;
forEachHint(data, [&](uint64_t kind, ArrayRef<uint64_t> args) {
if (kind == LOH_ARM64_ADRP_ADRP) {
hasAdrpAdrp = true;
return;
}
if (!findSection(args[0]))
return;
switch (kind) {
case LOH_ARM64_ADRP_ADD:
if (isValidOffset(args[1]))
applyAdrpAdd(buf, section, args[0] - sectionAddr,
args[1] - sectionAddr);
break;
case LOH_ARM64_ADRP_LDR:
if (isValidOffset(args[1]))
applyAdrpLdr(buf, section, args[0] - sectionAddr,
args[1] - sectionAddr);
break;
case LOH_ARM64_ADRP_LDR_GOT:
if (isValidOffset(args[1]))
applyAdrpLdrGot(buf, section, args[0] - sectionAddr,
args[1] - sectionAddr);
break;
case LOH_ARM64_ADRP_ADD_LDR:
if (isValidOffset(args[1]) && isValidOffset(args[2]))
applyAdrpAddLdr(buf, section, args[0] - sectionAddr,
args[1] - sectionAddr, args[2] - sectionAddr);
break;
case LOH_ARM64_ADRP_LDR_GOT_LDR:
if (isValidOffset(args[1]) && isValidOffset(args[2]))
applyAdrpLdrGotLdr(buf, section, args[0] - sectionAddr,
args[1] - sectionAddr, args[2] - sectionAddr);
break;
case LOH_ARM64_ADRP_ADD_STR:
case LOH_ARM64_ADRP_LDR_GOT_STR:
// TODO: Implement these
break;
}
});
if (!hasAdrpAdrp)
return;
// AdrpAdrp optimization hints are performed in a second pass because they
// might interfere with other transformations. For instance, consider the
// following input:
//
// adrp x0, _foo@PAGE
// add x1, x0, _foo@PAGEOFF
// adrp x0, _bar@PAGE
// add x2, x0, _bar@PAGEOFF
//
// If we perform the AdrpAdrp relaxation first, we get:
//
// adrp x0, _foo@PAGE
// add x1, x0, _foo@PAGEOFF
// nop
// add x2, x0, _bar@PAGEOFF
//
// If we then apply AdrpAdd to the first two instructions, the add will have a
// garbage value in x0:
//
// adr x1, _foo
// nop
// nop
// add x2, x0, _bar@PAGEOFF
forEachHint(data, [&](uint64_t kind, ArrayRef<uint64_t> args) {
if (kind != LOH_ARM64_ADRP_ADRP)
return;
if (!findSection(args[0]))
return;
if (isValidOffset(args[1]))
applyAdrpAdrp(buf, section, args[0] - sectionAddr, args[1] - sectionAddr);
});
}