AMDGPU/SI: Don't use fixup_si_rodata for scratch rsrc relocations

[• tstellarAMD]

Summary

We need to set the fixup type to FK_Data_4 for the
SCRATCH_RSRC_DWORD[01] symbols, since these require absolute
relocations, and fixup_si_rodata is for relative relocations.

[arsenm]

LGTM

lib/Target/AMDGPU/MCTargetDesc/SIMCCodeEmitter.cpp

253–257

ternary operator

lib/Target/AMDGPU/AMDGPUMCInstLower.cpp

for (const MachineOperand &MO : MI->explicit_operands()) {
     case MachineOperand::MO_GlobalAddress: {
       const GlobalValue *GV = MO.getGlobal();
       MCSymbol *Sym = Ctx.getOrCreateSymbol(StringRef(GV->getName()));
       MCOp = MCOperand::createExpr(MCSymbolRefExpr::create(Sym, Ctx));
       break;
     }
     case MachineOperand::MO_ExternalSymbol: {
       MCSymbol *Sym = Ctx.getOrCreateSymbol(StringRef(MO.getSymbolName()));
+      Sym->setExternal(true);
       const MCSymbolRefExpr *Expr = MCSymbolRefExpr::create(Sym, Ctx);
       MCOp = MCOperand::createExpr(Expr);
       break;
     }
     }
     OutMI.addOperand(MCOp);
   }
 }

lib/Target/AMDGPU/MCTargetDesc/SIMCCodeEmitter.cpp

uint64_t SIMCCodeEmitter::getMachineOpValue(const MCInst &MI,
                                             const MCOperand &MO,
                                        SmallVectorImpl<MCFixup> &Fixups,
                                        const MCSubtargetInfo &STI) const {
   if (MO.isReg())
     return MRI.getEncodingValue(MO.getReg());
 
   if (MO.isExpr()) {
     const MCSymbolRefExpr *Expr = cast<MCSymbolRefExpr>(MO.getExpr());
-    MCFixupKind Kind = (MCFixupKind)AMDGPU::fixup_si_rodata;
+    const MCSymbol &Sym = Expr->getSymbol();
+    MCFixupKind Kind;
+    if (Sym.isExternal())
+      Kind = FK_Data_4;
+    else
+      Kind = (MCFixupKind)AMDGPU::fixup_si_rodata;
     Fixups.push_back(MCFixup::create(4, Expr, Kind, MI.getLoc()));
   }
 
   // Figure out the operand number, needed for isSrcOperand check
   unsigned OpNo = 0;
   for (unsigned e = MI.getNumOperands(); OpNo < e; ++OpNo) {
     if (&MO == &MI.getOperand(OpNo))
       break;

test/CodeGen/AMDGPU/large-alloca-compute.ll

-; RUN: llc -march=amdgcn -mcpu=bonaire < %s | FileCheck -check-prefix=GCN -check-prefix=CI -check-prefix=ALL %s
-; RUN: llc -march=amdgcn -mcpu=carrizo < %s | FileCheck -check-prefix=GCN -check-prefix=VI -check-prefix=ALL %s
+; RUN: llc -march=amdgcn -mcpu=bonaire -show-mc-encoding < %s | FileCheck -check-prefix=GCN -check-prefix=CI -check-prefix=ALL %s
+; RUN: llc -march=amdgcn -mcpu=carrizo --show-mc-encoding < %s | FileCheck -check-prefix=GCN -check-prefix=VI -check-prefix=ALL %s
 ; RUN: llc -march=amdgcn -mcpu=bonaire -mtriple=amdgcn-unknown-amdhsa < %s -mattr=-flat-for-global | FileCheck -check-prefix=GCNHSA -check-prefix=CIHSA -check-prefix=ALL %s
 ; RUN: llc -march=amdgcn -mcpu=carrizo -mtriple=amdgcn-unknown-amdhsa -mattr=-flat-for-global < %s | FileCheck -check-prefix=GCNHSA -check-prefix=VIHSA -check-prefix=ALL %s
 
 ; FIXME: align on alloca seems to be ignored for private_segment_alignment
 
 ; ALL-LABEL: {{^}}large_alloca_compute_shader:
 
 ; GCN-DAG: s_mov_b32 s{{[0-9]+}}, SCRATCH_RSRC_DWORD0
+; GCN-DAG: ; fixup A - offset: 4, value: SCRATCH_RSRC_DWORD0, kind: FK_Data_4
 ; GCN-DAG: s_mov_b32 s{{[0-9]+}}, SCRATCH_RSRC_DWORD1
+; GCN-DAG: ; fixup A - offset: 4, value: SCRATCH_RSRC_DWORD1, kind: FK_Data_4
 ; GCN-DAG: s_mov_b32 s{{[0-9]+}}, -1
 ; CI-DAG: s_mov_b32 s{{[0-9]+}}, 0x88f000
 ; VI-DAG: s_mov_b32 s{{[0-9]+}}, 0x880000
 
 
 ; GCNHSA: .amd_kernel_code_t
 
 ; GCNHSA: compute_pgm_rsrc2_scratch_en = 1