[PowerPC][PCRelative] Add new pseudo instructions for PCRel TLS to fix R2 clobber issue

[NeHuang]

Summary

New pseudo instructions GETtlsADDRPCREL and GETtlsldADDRPCREL are added for properly setting REGMASK for tls_get_addr function when using PCRelative address.

[bsaleil]

llvm/lib/Target/PowerPC/PPCInstr64Bit.td

1330

Some of these pseudo-instruction definitions are the same. Could you refactor this code to use a common definition so we can define the pseudo-instructions with something like:

let Defs = [X0,X2,X4,X5,X6,X7,X8,X9,X10,X11,X12,LR8,CTR8,CR0,CR1,CR5,CR6,CR7], Size = 8 in
def GETtlsADDRPCREL: GETtlsADDRPseudo<"#GETtlsADDRPCREL">

let Defs = [X0,X4,X5,X6,X7,X8,X9,X10,X11,X12,LR8,CTR8,CR0,CR1,CR5,CR6,CR7] in
def GETtlsldADDR : GETtlsADDRPseudo<"#GETtlsldADDR">;

let Defs = [X0,X2,X4,X5,X6,X7,X8,X9,X10,X11,X12,LR8,CTR8,CR0,CR1,CR5,CR6,CR7] in
def GETtlsldADDRPCRel : GETtlsADDRPseudo<"#GETtlsldADDRPCREL">;

...

[NeHuang]

Addressed review comment to merge common code.

[NeHuang]

Address review comments to create the class.

[NeHuang]

Add the missing test case.

[bsaleil]

LGTM, thanks.

[NeHuang]

Patch committed on Nov 24th 2020.
Commit hash: 1f5c4a0d04ac0384218bf56e78e3c1d9c3c2f53f

llvm/lib/Target/PowerPC/PPCAsmPrinter.cpp

EmitToStreamer(*OutStreamer,
                    .addReg(MI->getOperand(0).getReg())
                    .addReg(MI->getOperand(1).getReg())
                    .addExpr(SymGotTlsGD));
     return;
   }
   case PPC::GETtlsADDR:
     // Transform: %x3 = GETtlsADDR %x3, @sym
     // Into: BL8_NOP_TLS __tls_get_addr(sym at tlsgd)
+  case PPC::GETtlsADDRPCREL:
   case PPC::GETtlsADDR32: {
     // Transform: %r3 = GETtlsADDR32 %r3, @sym
     // Into: BL_TLS __tls_get_addr(sym at tlsgd)@PLT
     EmitTlsCall(MI, MCSymbolRefExpr::VK_PPC_TLSGD);
     return;
   }
   case PPC::ADDIStlsldHA: {
     // Transform: %xd = ADDIStlsldHA %x2, @sym
EmitToStreamer(*OutStreamer,
                        .addReg(MI->getOperand(0).getReg())
                        .addReg(MI->getOperand(1).getReg())
                        .addExpr(SymGotTlsLD));
     return;
   }
   case PPC::GETtlsldADDR:
     // Transform: %x3 = GETtlsldADDR %x3, @sym
     // Into: BL8_NOP_TLS __tls_get_addr(sym at tlsld)
+  case PPC::GETtlsldADDRPCREL:
   case PPC::GETtlsldADDR32: {
     // Transform: %r3 = GETtlsldADDR32 %r3, @sym
     // Into: BL_TLS __tls_get_addr(sym at tlsld)@PLT
     EmitTlsCall(MI, MCSymbolRefExpr::VK_PPC_TLSLD);
     return;
   }
   case PPC::ADDISdtprelHA:
     // Transform: %xd = ADDISdtprelHA %xs, @sym

llvm/lib/Target/PowerPC/PPCInstr64Bit.td

def ADDIStlsgdHA: PPCEmitTimePseudo<(outs g8rc:$rD), (ins g8rc_nox0:$reg, s16imm64:$disp),
                          [(set i64:$rD,
                            (PPCaddisTlsgdHA i64:$reg, tglobaltlsaddr:$disp))]>,
                   isPPC64;
 def ADDItlsgdL : PPCEmitTimePseudo<(outs g8rc:$rD), (ins g8rc_nox0:$reg, s16imm64:$disp),
                        "#ADDItlsgdL",
                        [(set i64:$rD,
                          (PPCaddiTlsgdL i64:$reg, tglobaltlsaddr:$disp))]>,
                  isPPC64;
-// LR8 is a true define, while the rest of the Defs are clobbers.  X3 is
+
+let hasExtraSrcRegAllocReq = 1, hasExtraDefRegAllocReq = 1 in {
+// LR8 is a true define, while the rest of the Defs are clobbers. X3 is
 // explicitly defined when this op is created, so not mentioned here.
 // This is lowered to BL8_NOP_TLS by the assembly printer, so the size must be
 // correct because the branch select pass is relying on it.
-let hasExtraSrcRegAllocReq = 1, hasExtraDefRegAllocReq = 1, Size = 8,
-    Defs = [X0,X4,X5,X6,X7,X8,X9,X10,X11,X12,LR8,CTR8,CR0,CR1,CR5,CR6,CR7] in
+let Defs = [X0,X4,X5,X6,X7,X8,X9,X10,X11,X12,LR8,CTR8,CR0,CR1,CR5,CR6,CR7], Size = 8 in
 def GETtlsADDR : PPCEmitTimePseudo<(outs g8rc:$rD), (ins g8rc:$reg, tlsgd:$sym),
                         "#GETtlsADDR",
                         [(set i64:$rD,
                           (PPCgetTlsAddr i64:$reg, tglobaltlsaddr:$sym))]>,
                  isPPC64;
+
+let Defs = [X0,X2,X4,X5,X6,X7,X8,X9,X10,X11,X12,LR8,CTR8,CR0,CR1,CR5,CR6,CR7], Size = 8 in
+def GETtlsADDRPCREL : PPCEmitTimePseudo<(outs g8rc:$rD), (ins g8rc:$reg, tlsgd:$sym),
+                             "#GETtlsADDRPCREL",
+                             [(set i64:$rD,
+                               (PPCgetTlsAddr i64:$reg, tglobaltlsaddr:$sym))]>,
+                      isPPC64;
+// LR8 is a true define, while the rest of the Defs are clobbers. X3 is
+// explicitly defined when this op is created, so not mentioned here.
+let Defs = [X0,X4,X5,X6,X7,X8,X9,X10,X11,X12,LR8,CTR8,CR0,CR1,CR5,CR6,CR7] in
+def GETtlsldADDR : PPCEmitTimePseudo<(outs g8rc:$rD), (ins g8rc:$reg, tlsgd:$sym),
+                          "#GETtlsldADDR",
+                          [(set i64:$rD,
+                            (PPCgetTlsldAddr i64:$reg, tglobaltlsaddr:$sym))]>,
+                   isPPC64;
+
+let Defs = [X0,X2,X4,X5,X6,X7,X8,X9,X10,X11,X12,LR8,CTR8,CR0,CR1,CR5,CR6,CR7] in
+def GETtlsldADDRPCREL : PPCEmitTimePseudo<(outs g8rc:$rD), (ins g8rc:$reg, tlsgd:$sym),
+                               "#GETtlsldADDRPCREL",
+                               [(set i64:$rD,
+                                 (PPCgetTlsldAddr i64:$reg, tglobaltlsaddr:$sym))]>,
+                        isPPC64;
+}
+
 // Combined op for ADDItlsgdL and GETtlsADDR, late expanded.  X3 and LR8
 // are true defines while the rest of the Defs are clobbers.
 let hasExtraSrcRegAllocReq = 1, hasExtraDefRegAllocReq = 1,
     Defs = [X0,X3,X4,X5,X6,X7,X8,X9,X10,X11,X12,LR8,CTR8,CR0,CR1,CR5,CR6,CR7]
     in
 def ADDItlsgdLADDR : PPCEmitTimePseudo<(outs g8rc:$rD),
                             (ins g8rc_nox0:$reg, s16imm64:$disp, tlsgd:$sym),
                             "#ADDItlsgdLADDR",
                             [(set i64:$rD,
                               (PPCaddiTlsgdLAddr i64:$reg,
                                                  tglobaltlsaddr:$disp,
                                                  tglobaltlsaddr:$sym))]>,
                      isPPC64;
 def ADDIStlsldHA: PPCEmitTimePseudo<(outs g8rc:$rD), (ins g8rc_nox0:$reg, s16imm64:$disp),
                          "#ADDIStlsldHA",
                          [(set i64:$rD,
                            (PPCaddisTlsldHA i64:$reg, tglobaltlsaddr:$disp))]>,
                   isPPC64;
 def ADDItlsldL : PPCEmitTimePseudo<(outs g8rc:$rD), (ins g8rc_nox0:$reg, s16imm64:$disp),
                        "#ADDItlsldL",
                        [(set i64:$rD,
                          (PPCaddiTlsldL i64:$reg, tglobaltlsaddr:$disp))]>,
                  isPPC64;
-// LR8 is a true define, while the rest of the Defs are clobbers.  X3 is
-// explicitly defined when this op is created, so not mentioned here.
-let hasExtraSrcRegAllocReq = 1, hasExtraDefRegAllocReq = 1,
-    Defs = [X0,X4,X5,X6,X7,X8,X9,X10,X11,X12,LR8,CTR8,CR0,CR1,CR5,CR6,CR7] in
-def GETtlsldADDR : PPCEmitTimePseudo<(outs g8rc:$rD), (ins g8rc:$reg, tlsgd:$sym),
-                          "#GETtlsldADDR",
-                          [(set i64:$rD,
-                            (PPCgetTlsldAddr i64:$reg, tglobaltlsaddr:$sym))]>,
-                   isPPC64;
 // Combined op for ADDItlsldL and GETtlsADDR, late expanded.  X3 and LR8
 // are true defines, while the rest of the Defs are clobbers.
 let hasExtraSrcRegAllocReq = 1, hasExtraDefRegAllocReq = 1,
     Defs = [X0,X3,X4,X5,X6,X7,X8,X9,X10,X11,X12,LR8,CTR8,CR0,CR1,CR5,CR6,CR7]
     in
 def ADDItlsldLADDR : PPCEmitTimePseudo<(outs g8rc:$rD),
                             (ins g8rc_nox0:$reg, s16imm64:$disp, tlsgd:$sym),
                             "#ADDItlsldLADDR",
                             [(set i64:$rD,
                               (PPCaddiTlsldLAddr i64:$reg,
                                                  tglobaltlsaddr:$disp,
                                                  tglobaltlsaddr:$sym))]>,

llvm/lib/Target/PowerPC/PPCTLSDynamicCall.cpp

bool processBlock(MachineBasicBlock &MBB) {
           Opc1 = PPC::ADDItlsldL32;
           Opc2 = PPC::GETtlsldADDR32;
           break;
         case PPC::PADDI8pc:
           assert(IsPCREL && "Expecting General/Local Dynamic PCRel");
           Opc1 = PPC::PADDI8pc;
           Opc2 = MI.getOperand(2).getTargetFlags() ==
                          PPCII::MO_GOT_TLSGD_PCREL_FLAG
-                     ? PPC::GETtlsADDR
-                     : PPC::GETtlsldADDR;
+                     ? PPC::GETtlsADDRPCREL
+                     : PPC::GETtlsldADDRPCREL;
         }
 
         // We create ADJCALLSTACKUP and ADJCALLSTACKDOWN around _tls_get_addr
         // as scheduling fence to avoid it is scheduled before
         // mflr in the prologue and the address in LR is clobbered (PR25839).
         // We don't really need to save data to the stack - the clobbered
         // registers are already saved when the SDNode (e.g. PPCaddiTlsgdLAddr)
         // gets translated to the pseudo instruction (e.g. ADDItlsgdLADDR).