[Tablegen] Fix alias instruction printer mangling opcodes with optional suffixes.

[mwahab]

Summary

Log:

[Tablegen] Fix alias instruction printer mangling opcodes with optional suffixes.

The generic printer will now properly print AsmString formats such as
"opcode${a}${b} operands".

This introduces a slight change in behaviour: a tab is printed for the first
whitespace after the opcode and suffixes part. Previously, a tab would always be
printed after the opcode.

Summary:

An instruction alias defined with InstAlias and an optional operand immediately
after the opcode, "<opcode>${a} <operands>", would get the optional operand
emitted as unprintable characters in the instrution disassembly. This wouldn't
happen if the optional operand appeared as the last item in the AsmString which
is how the current backends avoided the problem.

There don't appear to be any tests for this part of Tablegen but it passes the
pre-commit tests. Manually tested the change by enabling the generic alias
printer in the ARM backend and checking the output.

[mwahab]

Ping

[hfinkel]

Can you please upload this patch with full context? For instructions, see:

http://llvm.org/docs/Phabricator.html#requesting-a-review-via-the-web-interface

utils/TableGen/AsmWriterEmitter.cpp

1029

So the general idea here is that we always eat the first ' ' or '\t' and replace it with a '\t'. All other whitespace (including that in between operands) passes through as provided. Is that right?

[mwahab]

Uploaded patch with full context.

[mwahab]

From: [e-mail] [mailto:[e-mail]]
Sent: 23 December 2014 23:27

Comment at: utils/TableGen/AsmWriterEmitter.cpp:1029
@@ -1026,2 +1028,3 @@

O << "          printOperand(MI, unsigned(AsmString[I++]) - 1, OS);\n";

• O << " } else {\n";

+ O << " } else if (NeedOpcodeSep &&\n";
+ O << " (AsmString[I] == ' ' || AsmString[I] == '\t'))

{\n";

So the general idea here is that we always eat the first ' ' or '\t' and
replace it with a '\t'. All other whitespace (including that in between
operands) passes through as provided. Is that right?

That's right. The current printer puts a tab between the opcode and the first operand. This needs to be suppressed until after those operands which are not separated from the opcode. Whitespace after the first separator will appear as is.

http://reviews.llvm.org/D6530

EMAIL PREFERENCES

http://reviews.llvm.org/settings/panel/emailpreferences/

• IMPORTANT NOTICE: The contents of this email and any attachments are confidential and may also be privileged. If you are not the intended recipient, please notify the sender immediately and do not disclose the contents to any other person, use it for any purpose, or store or copy the information in any medium. Thank you.

ARM Limited, Registered office 110 Fulbourn Road, Cambridge CB1 9NJ, Registered in England & Wales, Company No: 2557590
ARM Holdings plc, Registered office 110 Fulbourn Road, Cambridge CB1 9NJ, Registered in England & Wales, Company No: 2548782

utils/TableGen/AsmWriterEmitter.cpp

void AsmWriterEmitter::EmitPrintAliasInstruction(raw_ostream &O) {
   O << HeaderO.str();
   O.indent(2) << "const char *AsmString;\n";
   O.indent(2) << "switch (MI->getOpcode()) {\n";
   O.indent(2) << "default: return false;\n";
   O << CasesO.str();
   O.indent(2) << "}\n\n";
 
   // Code that prints the alias, replacing the operands with the ones from the
-  // MCInst.
+  // MCInst. Follows the format given by the instruction AsmString field,
+  // slightly complicated by opcodes with optional suffixes
+  // '<opcode>('${'<operand>'}')*<separator><operands>'.
   O << "  unsigned I = 0;\n";
   O << "  while (AsmString[I] != ' ' && AsmString[I] != '\t' &&\n";
-  O << "         AsmString[I] != '\\0')\n";
+  O << "         AsmString[I] != '\\0' && AsmString[I] != '$')\n";
   O << "    ++I;\n";
   O << "  OS << '\\t' << StringRef(AsmString, I);\n";
 
   O << "  if (AsmString[I] != '\\0') {\n";
-  O << "    OS << '\\t';\n";
+  O << "    bool NeedOpcodeSep = true;\n";
   O << "    do {\n";
   O << "      if (AsmString[I] == '$') {\n";
   O << "        ++I;\n";
   O << "        if (AsmString[I] == (char)0xff) {\n";
   O << "          ++I;\n";
   O << "          int OpIdx = AsmString[I++] - 1;\n";
   O << "          int PrintMethodIdx = AsmString[I++] - 1;\n";
   O << "          printCustomAliasOperand(MI, OpIdx, PrintMethodIdx, OS);\n";
   O << "        } else\n";
   O << "          printOperand(MI, unsigned(AsmString[I++]) - 1, OS);\n";
-  O << "      } else {\n";
+  O << "      } else if (NeedOpcodeSep &&\n";
+  O << "                 (AsmString[I] == ' ' || AsmString[I] == '\t')) {\n";
+  O << "          NeedOpcodeSep = false;\n";
+  O << "          OS << '\\t';\n";
+  O << "          ++I;\n";
+  O << "      } else\n";
   O << "        OS << AsmString[I++];\n";
-  O << "      }\n";
   O << "    } while (AsmString[I] != '\\0');\n";
   O << "  }\n\n";
 
   O << "  return true;\n";
   O << "}\n\n";
 
   //////////////////////////////
   // Write out the printCustomAliasOperand function