Summary:
Without these, the generic branch relaxation pass will underestimate the
range required for branches spanning these and we can end up with
"fixup value out of range" errors rather than relaxing the branches.
Some of the instructions in the expansion may end up being compressed
but exactly determining that is awkward, and these conservative values
should be safe, if slightly suboptimal in rare cases.
Reviewers: asb, lenary, luismarques, lewis-revill
Reviewed By: asb, luismarques
Subscribers: hiraditya, rbar, johnrusso, simoncook, sabuasal, niosHD, kito-cheng, shiva0217, MaskRay, zzheng, edward-jones, rogfer01, MartinMosbeck, brucehoult, the_o, rkruppe, jfb, PkmX, jocewei, psnobl, benna, Jim, s.egerton, pzheng, sameer.abuasal, apazos, evandro, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D77443
bool RISCVExpandAtomicPseudo::expandMI(MachineBasicBlock &MBB,
MachineBasicBlock::iterator MBBI,
MachineBasicBlock::iterator &NextMBBI) {
+ // RISCVInstrInfo::getInstSizeInBytes hard-codes the number of expanded
+ // instructions for each pseudo, and must be updated when adding new pseudos
+ // or changing existing ones.
switch (MBBI->getOpcode()) {
case RISCV::PseudoAtomicLoadNand32:
return expandAtomicBinOp(MBB, MBBI, AtomicRMWInst::Nand, false, 32,
bool RISCVExpandPseudo::expandMI(MachineBasicBlock &MBB,
MachineBasicBlock::iterator MBBI,
MachineBasicBlock::iterator &NextMBBI) {
+ // RISCVInstrInfo::getInstSizeInBytes hard-codes the number of expanded
+ // instructions for each pseudo, and must be updated when adding new pseudos
+ // or changing existing ones.
switch (MBBI->getOpcode()) {
case RISCV::PseudoLLA:
return expandLoadLocalAddress(MBB, MBBI, NextMBBI);
case TargetOpcode::KILL:
case TargetOpcode::DBG_VALUE:
return 0;
+ // These values are determined based on RISCVExpandAtomicPseudoInsts,
+ // RISCVExpandPseudoInsts and RISCVMCCodeEmitter, depending on where the
+ // pseudos are expanded.
case RISCV::PseudoCALLReg:
case RISCV::PseudoCALL:
case RISCV::PseudoJump:
case RISCV::PseudoLA_TLS_IE:
case RISCV::PseudoLA_TLS_GD:
return 8;
+ case RISCV::PseudoAtomicLoadNand32:
+ case RISCV::PseudoAtomicLoadNand64:
+ return 20;
+ case RISCV::PseudoMaskedAtomicSwap32:
+ case RISCV::PseudoMaskedAtomicLoadAdd32:
+ case RISCV::PseudoMaskedAtomicLoadSub32:
+ return 28;
+ case RISCV::PseudoMaskedAtomicLoadNand32:
+ return 32;
+ case RISCV::PseudoMaskedAtomicLoadMax32:
+ case RISCV::PseudoMaskedAtomicLoadMin32:
+ return 44;
+ case RISCV::PseudoMaskedAtomicLoadUMax32:
+ case RISCV::PseudoMaskedAtomicLoadUMin32:
+ return 36;
+ case RISCV::PseudoCmpXchg32:
+ case RISCV::PseudoCmpXchg64:
+ return 16;
+ case RISCV::PseudoMaskedCmpXchg32:
+ return 32;
case TargetOpcode::INLINEASM:
case TargetOpcode::INLINEASM_BR: {
const MachineFunction &MF = *MI.getParent()->getParent();