[LoongArch][Codegen] Add support for TLSDESC

The implementation only enables when the `-enable-tlsdesc` option is
passed and the TLS model is `dynamic`.

LoongArch's GCC has the same option(-mtls-dialet=) as RISC-V.

Reviewers: heiher, MaskRay, SixWeining

Reviewed By: SixWeining, MaskRay

Pull Request: #90159

clang/lib/Driver/ToolChains/CommonArgs.cpp

@@ -737,7 +737,7 @@ bool tools::isTLSDESCEnabled(const ToolChain &TC,
   StringRef V = A->getValue();
   bool SupportedArgument = false, EnableTLSDESC = false;
   bool Unsupported = !Triple.isOSBinFormatELF();
-  if (Triple.isRISCV()) {
+  if (Triple.isLoongArch() || Triple.isRISCV()) {
     SupportedArgument = V == "desc" || V == "trad";
     EnableTLSDESC = V == "desc";
   } else if (Triple.isX86()) {

clang/test/CodeGen/LoongArch/tls-dialect.c

@@ -0,0 +1,14 @@
+// REQUIRES: loongarch-registered-target
+/// cc1 -enable-tlsdesc (due to -mtls-dialect=desc) enables TLSDESC.
+// RUN: %clang_cc1 -triple loongarch64 -S -mrelocation-model pic -pic-level 1 -enable-tlsdesc %s -o - | FileCheck %s --check-prefix=DESC
+// RUN: %clang_cc1 -triple loongarch64 -S -mrelocation-model pic -pic-level 1 %s -o - | FileCheck %s --check-prefix=NODESC
+
+__thread int x;
+
+// DESC:       %desc_pc_hi20
+// DESC-NOT:   %gd_pc_hi20
+// NODESC:     %gd_pc_hi20
+// NODESC-NOT: %desc_pc_hi20
+int use() {
+  return x;
+}

clang/test/Driver/tls-dialect.c

@@ -1,3 +1,5 @@
+// RUN: %clang -### --target=loongarch64-linux -mtls-dialect=trad %s 2>&1 | FileCheck --check-prefix=NODESC %s
+// RUN: %clang -### --target=loongarch64-linux %s 2>&1 | FileCheck --check-prefix=NODESC %s
 // RUN: %clang -### --target=riscv64-freebsd -mtls-dialect=desc %s 2>&1 | FileCheck --check-prefix=DESC %s
 // RUN: %clang -### --target=riscv64-linux -mtls-dialect=trad %s 2>&1 | FileCheck --check-prefix=NODESC %s
 // RUN: %clang -### --target=riscv64-linux %s 2>&1 | FileCheck --check-prefix=NODESC %s
@@ -9,6 +11,8 @@
 // RUN: %clang -### --target=riscv64-android %s 2>&1 | FileCheck --check-prefix=DESC %s
 
 /// LTO
+// RUN: %clang -### --target=loongarch64-linux -flto -mtls-dialect=desc %s 2>&1 | FileCheck --check-prefix=LTO-DESC %s
+// RUN: %clang -### --target=loongarch64-linux -flto %s 2>&1 | FileCheck --check-prefix=LTO-NODESC %s
 // RUN: %clang -### --target=riscv64-linux -flto -mtls-dialect=desc %s 2>&1 | FileCheck --check-prefix=LTO-DESC %s
 // RUN: %clang -### --target=riscv64-linux -flto %s 2>&1 | FileCheck --check-prefix=LTO-NODESC %s
 
@@ -18,6 +22,7 @@
 // RUN: not %clang --target=x86_64-apple-macos -mtls-dialect=desc -flto %s 2>&1 | FileCheck -check-prefix=UNSUPPORTED-TARGET %s
 
 /// Unsupported argument
+// RUN: not %clang -### --target=loongarch64-linux -mtls-dialect=gnu2 %s 2>&1 | FileCheck --check-prefix=UNSUPPORTED-ARG %s
 // RUN: not %clang -### --target=riscv64-linux -mtls-dialect=gnu2 %s 2>&1 | FileCheck --check-prefix=UNSUPPORTED-ARG %s
 
 // DESC:       "-cc1" {{.*}}"-enable-tlsdesc"

llvm/lib/Target/LoongArch/LoongArchExpandPseudoInsts.cpp

@@ -80,6 +80,9 @@ class LoongArchPreRAExpandPseudo : public MachineFunctionPass {
   bool expandLoadAddressTLSGD(MachineBasicBlock &MBB,
                               MachineBasicBlock::iterator MBBI,
                               MachineBasicBlock::iterator &NextMBBI);
+  bool expandLoadAddressTLSDesc(MachineBasicBlock &MBB,
+                                MachineBasicBlock::iterator MBBI,
+                                MachineBasicBlock::iterator &NextMBBI);
 };
 
 char LoongArchPreRAExpandPseudo::ID = 0;
@@ -122,6 +125,8 @@ bool LoongArchPreRAExpandPseudo::expandMI(
     return expandLoadAddressTLSLD(MBB, MBBI, NextMBBI);
   case LoongArch::PseudoLA_TLS_GD:
     return expandLoadAddressTLSGD(MBB, MBBI, NextMBBI);
+  case LoongArch::PseudoLA_TLS_DESC_PC:
+    return expandLoadAddressTLSDesc(MBB, MBBI, NextMBBI);
   }
   return false;
 }
@@ -267,6 +272,52 @@ bool LoongArchPreRAExpandPseudo::expandLoadAddressTLSGD(
                                  SecondOpcode, LoongArchII::MO_GOT_PC_LO);
 }
 
+bool LoongArchPreRAExpandPseudo::expandLoadAddressTLSDesc(
+    MachineBasicBlock &MBB, MachineBasicBlock::iterator MBBI,
+    MachineBasicBlock::iterator &NextMBBI) {
+  // Code Sequence:
+  // pcalau12i $a0, %desc_pc_hi20(sym)
+  // addi.w/d  $a0, $a0, %desc_pc_lo12(sym)
+  // ld.w/d    $ra, $a0, %desc_ld(sym)
+  // jirl      $ra, $ra, %desc_ld(sym)
+  // add.d     $dst, $a0, $tp
+  MachineFunction *MF = MBB.getParent();
+  MachineInstr &MI = *MBBI;
+  DebugLoc DL = MI.getDebugLoc();
+
+  const auto &STI = MF->getSubtarget<LoongArchSubtarget>();
+  unsigned ADD = STI.is64Bit() ? LoongArch::ADD_D : LoongArch::ADD_W;
+  unsigned ADDI = STI.is64Bit() ? LoongArch::ADDI_D : LoongArch::ADDI_W;
+  unsigned LD = STI.is64Bit() ? LoongArch::LD_D : LoongArch::LD_W;
+
+  Register DestReg = MI.getOperand(0).getReg();
+  Register ScratchReg =
+      MF->getRegInfo().createVirtualRegister(&LoongArch::GPRRegClass);
+  MachineOperand &Symbol = MI.getOperand(1);
+
+  BuildMI(MBB, MBBI, DL, TII->get(LoongArch::PCALAU12I), ScratchReg)
+      .addDisp(Symbol, 0, LoongArchII::MO_DESC_PC_HI);
+
+  BuildMI(MBB, MBBI, DL, TII->get(ADDI), LoongArch::R4)
+      .addReg(ScratchReg)
+      .addDisp(Symbol, 0, LoongArchII::MO_DESC_PC_LO);
+
+  BuildMI(MBB, MBBI, DL, TII->get(LD), LoongArch::R1)
+      .addReg(LoongArch::R4)
+      .addDisp(Symbol, 0, LoongArchII::MO_DESC_LD);
+
+  BuildMI(MBB, MBBI, DL, TII->get(LoongArch::PseudoDESC_CALL), LoongArch::R1)
+      .addReg(LoongArch::R1)
+      .addDisp(Symbol, 0, LoongArchII::MO_DESC_CALL);
+
+  BuildMI(MBB, MBBI, DL, TII->get(ADD), DestReg)
+      .addReg(LoongArch::R4)
+      .addReg(LoongArch::R2);
+
+  MI.eraseFromParent();
+  return true;
+}
+
 class LoongArchExpandPseudo : public MachineFunctionPass {
 public:
   const LoongArchInstrInfo *TII;
@@ -313,6 +364,9 @@ class LoongArchExpandPseudo : public MachineFunctionPass {
   bool expandLoadAddressTLSGDLarge(MachineBasicBlock &MBB,
                                    MachineBasicBlock::iterator MBBI,
                                    MachineBasicBlock::iterator &NextMBBI);
+  bool expandLoadAddressTLSDescPcLarge(MachineBasicBlock &MBB,
+                                       MachineBasicBlock::iterator MBBI,
+                                       MachineBasicBlock::iterator &NextMBBI);
   bool expandFunctionCALL(MachineBasicBlock &MBB,
                           MachineBasicBlock::iterator MBBI,
                           MachineBasicBlock::iterator &NextMBBI,
@@ -361,6 +415,8 @@ bool LoongArchExpandPseudo::expandMI(MachineBasicBlock &MBB,
     return expandLoadAddressTLSLDLarge(MBB, MBBI, NextMBBI);
   case LoongArch::PseudoLA_TLS_GD_LARGE:
     return expandLoadAddressTLSGDLarge(MBB, MBBI, NextMBBI);
+  case LoongArch::PseudoLA_TLS_DESC_PC_LARGE:
+    return expandLoadAddressTLSDescPcLarge(MBB, MBBI, NextMBBI);
   case LoongArch::PseudoCALL:
   case LoongArch::PseudoCALL_MEDIUM:
   case LoongArch::PseudoCALL_LARGE:
@@ -560,6 +616,58 @@ bool LoongArchExpandPseudo::expandLoadAddressTLSGDLarge(
                                 LoongArchII::MO_GD_PC_HI);
 }
 
+bool LoongArchExpandPseudo::expandLoadAddressTLSDescPcLarge(
+    MachineBasicBlock &MBB, MachineBasicBlock::iterator MBBI,
+    MachineBasicBlock::iterator &NextMBBI) {
+  // Code Sequence:
+  //
+  // pcalau12i  $a0, %desc_pc_hi20(sym)
+  // addi.d     $t8, $zero, %desc_pc_lo12(sym)
+  // lu32i.d    $t8, %desc64_pc_lo20(sym)
+  // lu52i.d    $t8, $t8, %desc64_pc_hi12(sym)
+  // add.d      $a0, $a0, $t8
+  // ld.d       $ra, $a0, %desc_ld(sym)
+  // jirl       $ra, $ra, %desc_call(sym)
+  // add.d      $dst, $a0, $tp
+
+  MachineInstr &MI = *MBBI;
+  DebugLoc DL = MI.getDebugLoc();
+  Register DestReg = MI.getOperand(0).getReg();
+  MachineOperand &Symbol = MI.getOperand(2);
+  Register ScratchReg = LoongArch::R20; // $t8
+
+  assert(MBB.getParent()->getSubtarget<LoongArchSubtarget>().is64Bit() &&
+         "Large code model requires LA64");
+
+  BuildMI(MBB, MBBI, DL, TII->get(LoongArch::PCALAU12I), LoongArch::R4)
+      .addDisp(Symbol, 0, LoongArchII::MO_DESC_PC_HI);
+  BuildMI(MBB, MBBI, DL, TII->get(LoongArch::ADDI_D), ScratchReg)
+      .addReg(LoongArch::R0)
+      .addDisp(Symbol, 0, LoongArchII::MO_DESC_PC_LO);
+  BuildMI(MBB, MBBI, DL, TII->get(LoongArch::LU32I_D), ScratchReg)
+      .addReg(ScratchReg)
+      .addDisp(Symbol, 0, LoongArchII::MO_DESC64_PC_LO);
+  BuildMI(MBB, MBBI, DL, TII->get(LoongArch::LU52I_D), ScratchReg)
+      .addReg(ScratchReg)
+      .addDisp(Symbol, 0, LoongArchII::MO_DESC64_PC_HI);
+  BuildMI(MBB, MBBI, DL, TII->get(LoongArch::ADD_D), LoongArch::R4)
+      .addReg(ScratchReg)
+      .addReg(LoongArch::R4);
+  BuildMI(MBB, MBBI, DL, TII->get(LoongArch::LD_D), LoongArch::R1)
+      .addReg(LoongArch::R4)
+      .addDisp(Symbol, 0, LoongArchII::MO_DESC_LD);
+  BuildMI(MBB, MBBI, DL, TII->get(LoongArch::PseudoDESC_CALL), LoongArch::R1)
+      .addReg(LoongArch::R1)
+      .addDisp(Symbol, 0, LoongArchII::MO_DESC_CALL);
+  BuildMI(MBB, MBBI, DL, TII->get(LoongArch::ADD_D), DestReg)
+      .addReg(LoongArch::R4)
+      .addReg(LoongArch::R2);
+
+  MI.eraseFromParent();
+
+  return true;
+}
+
 bool LoongArchExpandPseudo::expandFunctionCALL(
     MachineBasicBlock &MBB, MachineBasicBlock::iterator MBBI,
     MachineBasicBlock::iterator &NextMBBI, bool IsTailCall) {

llvm/lib/Target/LoongArch/LoongArchISelLowering.cpp

@@ -903,6 +903,24 @@ SDValue LoongArchTargetLowering::getDynamicTLSAddr(GlobalAddressSDNode *N,
   return LowerCallTo(CLI).first;
 }
 
+SDValue LoongArchTargetLowering::getTLSDescAddr(GlobalAddressSDNode *N,
+                                                SelectionDAG &DAG, unsigned Opc,
+                                                bool Large) const {
+  SDLoc DL(N);
+  EVT Ty = getPointerTy(DAG.getDataLayout());
+  const GlobalValue *GV = N->getGlobal();
+
+  // This is not actually used, but is necessary for successfully matching the
+  // PseudoLA_*_LARGE nodes.
+  SDValue Tmp = DAG.getConstant(0, DL, Ty);
+
+  // Use a PC-relative addressing mode to access the global dynamic GOT address.
+  // This generates the pattern (PseudoLA_TLS_DESC_PC{,LARGE} sym).
+  SDValue Addr = DAG.getTargetGlobalAddress(GV, DL, Ty, 0, 0);
+  return Large ? SDValue(DAG.getMachineNode(Opc, DL, Ty, Tmp, Addr), 0)
+               : SDValue(DAG.getMachineNode(Opc, DL, Ty, Addr), 0);
+}
+
 SDValue
 LoongArchTargetLowering::lowerGlobalTLSAddress(SDValue Op,
                                                SelectionDAG &DAG) const {
@@ -916,42 +934,46 @@ LoongArchTargetLowering::lowerGlobalTLSAddress(SDValue Op,
   GlobalAddressSDNode *N = cast<GlobalAddressSDNode>(Op);
   assert(N->getOffset() == 0 && "unexpected offset in global node");
 
-  SDValue Addr;
+  bool IsDesc = DAG.getTarget().useTLSDESC();
+
   switch (getTargetMachine().getTLSModel(N->getGlobal())) {
   case TLSModel::GeneralDynamic:
     // In this model, application code calls the dynamic linker function
     // __tls_get_addr to locate TLS offsets into the dynamic thread vector at
     // runtime.
-    Addr = getDynamicTLSAddr(N, DAG,
-                             Large ? LoongArch::PseudoLA_TLS_GD_LARGE
-                                   : LoongArch::PseudoLA_TLS_GD,
-                             Large);
+    if (!IsDesc)
+      return getDynamicTLSAddr(N, DAG,
+                               Large ? LoongArch::PseudoLA_TLS_GD_LARGE
+                                     : LoongArch::PseudoLA_TLS_GD,
+                               Large);
     break;
   case TLSModel::LocalDynamic:
     // Same as GeneralDynamic, except for assembly modifiers and relocation
     // records.
-    Addr = getDynamicTLSAddr(N, DAG,
-                             Large ? LoongArch::PseudoLA_TLS_LD_LARGE
-                                   : LoongArch::PseudoLA_TLS_LD,
-                             Large);
+    if (!IsDesc)
+      return getDynamicTLSAddr(N, DAG,
+                               Large ? LoongArch::PseudoLA_TLS_LD_LARGE
+                                     : LoongArch::PseudoLA_TLS_LD,
+                               Large);
     break;
   case TLSModel::InitialExec:
     // This model uses the GOT to resolve TLS offsets.
-    Addr = getStaticTLSAddr(N, DAG,
+    return getStaticTLSAddr(N, DAG,
                             Large ? LoongArch::PseudoLA_TLS_IE_LARGE
                                   : LoongArch::PseudoLA_TLS_IE,
                             Large);
-    break;
   case TLSModel::LocalExec:
     // This model is used when static linking as the TLS offsets are resolved
     // during program linking.
     //
     // This node doesn't need an extra argument for the large code model.
-    Addr = getStaticTLSAddr(N, DAG, LoongArch::PseudoLA_TLS_LE);
-    break;
+    return getStaticTLSAddr(N, DAG, LoongArch::PseudoLA_TLS_LE);
   }
 
-  return Addr;
+  return getTLSDescAddr(N, DAG,
+                        Large ? LoongArch::PseudoLA_TLS_DESC_PC_LARGE
+                              : LoongArch::PseudoLA_TLS_DESC_PC,
+                        Large);
 }
 
 template <unsigned N>

llvm/lib/Target/LoongArch/LoongArchISelLowering.h

@@ -263,6 +263,8 @@ class LoongArchTargetLowering : public TargetLowering {
                            unsigned Opc, bool Large = false) const;
   SDValue getDynamicTLSAddr(GlobalAddressSDNode *N, SelectionDAG &DAG,
                             unsigned Opc, bool Large = false) const;
+  SDValue getTLSDescAddr(GlobalAddressSDNode *N, SelectionDAG &DAG,
+                         unsigned Opc, bool Large = false) const;
   SDValue lowerGlobalAddress(SDValue Op, SelectionDAG &DAG) const;
   SDValue lowerBlockAddress(SDValue Op, SelectionDAG &DAG) const;
   SDValue lowerJumpTable(SDValue Op, SelectionDAG &DAG) const;

llvm/lib/Target/LoongArch/LoongArchInstrInfo.cpp

@@ -530,6 +530,12 @@ LoongArchInstrInfo::getSerializableDirectMachineOperandTargetFlags() const {
       {MO_IE_PC_LO, "loongarch-ie-pc-lo"},
       {MO_IE_PC64_LO, "loongarch-ie-pc64-lo"},
       {MO_IE_PC64_HI, "loongarch-ie-pc64-hi"},
+      {MO_DESC_PC_HI, "loongarch-desc-pc-hi"},
+      {MO_DESC_PC_LO, "loongarch-desc-pc-lo"},
+      {MO_DESC64_PC_LO, "loongarch-desc64-pc-lo"},
+      {MO_DESC64_PC_HI, "loongarch-desc64-pc-hi"},
+      {MO_DESC_LD, "loongarch-desc-ld"},
+      {MO_DESC_CALL, "loongarch-desc-call"},
       {MO_LD_PC_HI, "loongarch-ld-pc-hi"},
       {MO_GD_PC_HI, "loongarch-gd-pc-hi"}};
   return ArrayRef(TargetFlags);

llvm/lib/Target/LoongArch/LoongArchInstrInfo.td

@@ -1607,6 +1607,13 @@ def PseudoLA_TLS_GD_LARGE : Pseudo<(outs GPR:$dst),
 } // Defs = [R20], Size = 20
 }
 
+// Used for expand PseudoLA_TLS_DESC_* instructions.
+let isCall = 1, isBarrier = 1, hasSideEffects = 0, mayStore = 0, mayLoad = 0,
+    Defs = [R4], Uses = [R4] in
+def PseudoDESC_CALL : Pseudo<(outs GPR:$rd), (ins GPR:$rj, simm16_lsl2:$imm16)>,
+                      PseudoInstExpansion<(JIRL GPR:$rd, GPR:$rj,
+                                           simm16_lsl2:$imm16)>;
+
 // TLSDESC
 let hasSideEffects = 0, mayLoad = 1, mayStore = 0, isCodeGenOnly = 0,
     isAsmParserOnly = 1, Defs = [R1] in {

llvm/lib/Target/LoongArch/LoongArchMCInstLower.cpp

@@ -98,6 +98,24 @@ static MCOperand lowerSymbolOperand(const MachineOperand &MO, MCSymbol *Sym,
   case LoongArchII::MO_CALL36:
     Kind = LoongArchMCExpr::VK_LoongArch_CALL36;
     break;
+  case LoongArchII::MO_DESC_PC_HI:
+    Kind = LoongArchMCExpr::VK_LoongArch_TLS_DESC_PC_HI20;
+    break;
+  case LoongArchII::MO_DESC_PC_LO:
+    Kind = LoongArchMCExpr::VK_LoongArch_TLS_DESC_PC_LO12;
+    break;
+  case LoongArchII::MO_DESC64_PC_LO:
+    Kind = LoongArchMCExpr::VK_LoongArch_TLS_DESC64_PC_LO20;
+    break;
+  case LoongArchII::MO_DESC64_PC_HI:
+    Kind = LoongArchMCExpr::VK_LoongArch_TLS_DESC64_PC_HI12;
+    break;
+  case LoongArchII::MO_DESC_LD:
+    Kind = LoongArchMCExpr::VK_LoongArch_TLS_DESC_LD;
+    break;
+  case LoongArchII::MO_DESC_CALL:
+    Kind = LoongArchMCExpr::VK_LoongArch_TLS_DESC_CALL;
+    break;
     // TODO: Handle more target-flags.
   }
 

llvm/lib/Target/LoongArch/MCTargetDesc/LoongArchBaseInfo.h

@@ -47,7 +47,13 @@ enum {
   MO_IE_PC64_HI,
   MO_LD_PC_HI,
   MO_GD_PC_HI,
-  MO_CALL36
+  MO_CALL36,
+  MO_DESC_PC_HI,
+  MO_DESC_PC_LO,
+  MO_DESC64_PC_HI,
+  MO_DESC64_PC_LO,
+  MO_DESC_LD,
+  MO_DESC_CALL,
   // TODO: Add more flags.
 };
 } // end namespace LoongArchII