[llvm] [MC][SystemZ] Introduce Target Specific HLASM Streamer for z/OS (PR #130535)

[Ulrich Weigand via llvm-commits]

================
@@ -34,12 +35,27 @@ void SystemZHLASMInstPrinter::printInst(const MCInst *MI, uint64_t Address,
   raw_string_ostream RSO(Str);
   printInstruction(MI, Address, RSO);
   // Eat the first tab character and replace it with a space since it is
-  // hardcoded in AsmWriterEmitter::EmitPrintInstruction
-  // TODO: introduce a line prefix member to AsmWriter to avoid this problem
+  // hardcoded in AsmWriterEmitter::EmitPrintInstruction.
+  // TODO Introduce a line prefix member to AsmWriter to avoid this problem.
   if (!Str.empty() && Str.front() == '\t')
     O << " " << Str.substr(1, Str.length());
   else
     O << Str;
 
   printAnnotation(O, Annot);
 }
+
+void SystemZHLASMInstPrinter::printPCRelOperand(const MCInst *MI, int OpNum,
+                                                raw_ostream &O) {
+  const MCOperand &MO = MI->getOperand(OpNum);
+  if (MO.isImm()) {
+    WithMarkup M = markup(O, Markup::Immediate);
+    O << "0x";
+    O.write_hex(MO.getImm());
+  } else {
+    // Don't print @PLT.
+    // TODO May need to do something more here.
+    const MCSymbolRefExpr &SRE = cast<MCSymbolRefExpr>(*(MO.getExpr()));
----------------
uweigand wrote:

I think more to the point is that this isn't really an assembler difference as such - it should rather be a *codegen* difference to begin with.  On z/OS (no matter which assembler is used), there *is* no PLT, so z/OS codegen shouldn't generate @PLT symbols at all, whether via VariantKind or some other way.

https://github.com/llvm/llvm-project/pull/130535