lib/Target/PowerPC/PPCTOCRegDeps.cpp

   1 //===-- PPCTOCRegDeps.cpp - Add Extra TOC Register Dependencies -----------===//
   2 //
   3 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
   4 // See https://llvm.org/LICENSE.txt for license information.
   5 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
   6 //
   7 //===----------------------------------------------------------------------===//
   8 //
   9 // When resolving an address using the ELF ABI TOC pointer, two relocations are
  10 // generally required: one for the high part and one for the low part. Only
  11 // the high part generally explicitly depends on r2 (the TOC pointer). And, so,
  12 // we might produce code like this:
  13 //
  14 // .Ltmp526:
  15 //         addis 3, 2, .LC12@toc@ha
  16 // .Ltmp1628:
  17 //         std 2, 40(1)
  18 //         ld 5, 0(27)
  19 //         ld 2, 8(27)
  20 //         ld 11, 16(27)
  21 //         ld 3, .LC12@toc@l(3)
  22 //         rldicl 4, 4, 0, 32
  23 //         mtctr 5
  24 //         bctrl
  25 //         ld 2, 40(1)
  26 //
  27 // And there is nothing wrong with this code, as such, but there is a linker bug
  28 // in binutils (https://sourceware.org/bugzilla/show_bug.cgi?id=18414) that will
  29 // misoptimize this code sequence to this:
  30 //         nop
  31 //         std     r2,40(r1)
  32 //         ld      r5,0(r27)
  33 //         ld      r2,8(r27)
  34 //         ld      r11,16(r27)
  35 //         ld      r3,-32472(r2)
  36 //         clrldi  r4,r4,32
  37 //         mtctr   r5
  38 //         bctrl
  39 //         ld      r2,40(r1)
  40 // because the linker does not know (and does not check) that the value in r2
  41 // changed in between the instruction using the .LC12@toc@ha (TOC-relative)
  42 // relocation and the instruction using the .LC12@toc@l(3) relocation.
  43 // Because it finds these instructions using the relocations (and not by
  44 // scanning the instructions), it has been asserted that there is no good way
  45 // to detect the change of r2 in between. As a result, this bug may never be
  46 // fixed (i.e. it may become part of the definition of the ABI). GCC was
  47 // updated to add extra dependencies on r2 to instructions using the @toc@l
  48 // relocations to avoid this problem, and we'll do the same here.
  49 //
  50 // This is done as a separate pass because:
  51 //  1. These extra r2 dependencies are not really properties of the
  52 //     instructions, but rather due to a linker bug, and maybe one day we'll be
  53 //     able to get rid of them when targeting linkers without this bug (and,
  54 //     thus, keeping the logic centralized here will make that
  55 //     straightforward).
  56 //  2. There are ISel-level peephole optimizations that propagate the @toc@l
  57 //     relocations to some user instructions, and so the exta dependencies do
  58 //     not apply only to a fixed set of instructions (without undesirable
  59 //     definition replication).
  60 //
  61 //===----------------------------------------------------------------------===//
  62
  63 #include "MCTargetDesc/PPCPredicates.h"
  64 #include "PPC.h"
  65 #include "PPCInstrBuilder.h"
  66 #include "PPCInstrInfo.h"
  67 #include "PPCMachineFunctionInfo.h"
  68 #include "PPCTargetMachine.h"
  69 #include "llvm/ADT/STLExtras.h"
  70 #include "llvm/ADT/Statistic.h"
  71 #include "llvm/CodeGen/MachineFrameInfo.h"
  72 #include "llvm/CodeGen/MachineFunctionPass.h"
  73 #include "llvm/CodeGen/MachineInstr.h"
  74 #include "llvm/CodeGen/MachineRegisterInfo.h"
  75 #include "llvm/MC/MCAsmInfo.h"
  76 #include "llvm/Support/Debug.h"
  77 #include "llvm/Support/ErrorHandling.h"
  78 #include "llvm/Support/TargetRegistry.h"
  79 #include "llvm/Support/raw_ostream.h"
  80
  81 using namespace llvm;
  82
  83 #define DEBUG_TYPE "ppc-toc-reg-deps"
  84
  85 namespace {
  86   // PPCTOCRegDeps pass - For simple functions without epilogue code, move
  87   // returns up, and create conditional returns, to avoid unnecessary
  88   // branch-to-blr sequences.
  89   struct PPCTOCRegDeps : public MachineFunctionPass {
  90     static char ID;
  91     PPCTOCRegDeps() : MachineFunctionPass(ID) {
  92       initializePPCTOCRegDepsPass(*PassRegistry::getPassRegistry());
  93     }
  94
  95 protected:
  96     bool hasTOCLoReloc(const MachineInstr &MI) {
  97       if (MI.getOpcode() == PPC::LDtocL ||
  98           MI.getOpcode() == PPC::ADDItocL ||
  99           MI.getOpcode() == PPC::LWZtocL)
 100         return true;
 101
 102       for (const MachineOperand &MO : MI.operands()) {
 103         if ((MO.getTargetFlags() & PPCII::MO_ACCESS_MASK) == PPCII::MO_TOC_LO)
 104           return true;
 105       }
 106
 107       return false;
 108     }
 109
 110     bool processBlock(MachineBasicBlock &MBB) {
 111       bool Changed = false;
 112
 113       const bool isPPC64 =
 114           MBB.getParent()->getSubtarget<PPCSubtarget>().isPPC64();
 115       const unsigned TOCReg = isPPC64 ? PPC::X2 : PPC::R2;
 116
 117       for (auto &MI : MBB) {
 118         if (!hasTOCLoReloc(MI))
 119           continue;
 120
 121         MI.addOperand(MachineOperand::CreateReg(TOCReg,
 122                                                 false  /*IsDef*/,
 123                                                 true  /*IsImp*/));
 124         Changed = true;
 125       }
 126
 127       return Changed;
 128     }
 129
 130 public:
 131     bool runOnMachineFunction(MachineFunction &MF) override {
 132       bool Changed = false;
 133
 134       for (MachineFunction::iterator I = MF.begin(); I != MF.end();) {
 135         MachineBasicBlock &B = *I++;
 136         if (processBlock(B))
 137           Changed = true;
 138       }
 139
 140       return Changed;
 141     }
 142
 143     void getAnalysisUsage(AnalysisUsage &AU) const override {
 144       MachineFunctionPass::getAnalysisUsage(AU);
 145     }
 146   };
 147 }
 148
 149 INITIALIZE_PASS(PPCTOCRegDeps, DEBUG_TYPE,
 150                 "PowerPC TOC Register Dependencies", false, false)
 151
 152 char PPCTOCRegDeps::ID = 0;
 153 FunctionPass*
 154 llvm::createPPCTOCRegDepsPass() { return new PPCTOCRegDeps(); }
 155