#include "AsmWriterEmitter.h"
#include "CodeGenTarget.h"
#include "Record.h"
+#include "StringToOffsetTable.h"
#include "llvm/ADT/StringExtras.h"
#include "llvm/Support/Debug.h"
#include "llvm/Support/MathExtras.h"
#include <algorithm>
-#include <sstream>
-#include <iostream>
using namespace llvm;
+
static bool isIdentChar(char C) {
return (C >= 'a' && C <= 'z') ||
(C >= 'A' && C <= 'Z') ||
LastEmitted = VarEnd;
}
}
-
- Operands.push_back(
- AsmWriterOperand("EmitComments(*MI);\n",
- AsmWriterOperand::isLiteralStatementOperand));
- AddLiteralString("\\n");
+
+ Operands.push_back(AsmWriterOperand("return;",
+ AsmWriterOperand::isLiteralStatementOperand));
}
/// MatchesAllButOneOp - If this instruction is exactly identical to the
Command = " " + Inst->Operands[0].getCode() + "\n";
- // If this is the last operand, emit a return.
- if (Inst->Operands.size() == 1)
- Command += " return;\n";
-
// Check to see if we already have 'Command' in UniqueOperandCommands.
// If not, add it.
bool FoundIt = false;
// to UniqueOperandCommands and remember that it was consumed.
std::string Command = " " + FirstInst->Operands[Op].getCode() + "\n";
- // If this is the last operand, emit a return after the code.
- if (FirstInst->Operands.size() == Op+1 &&
- // Don't early-out too soon. Other instructions in this
- // group may have more operands.
- FirstInst->Operands.size() == MaxSize) {
- Command += " return;\n";
- }
-
UniqueOperandCommands[CommandIdx] += Command;
InstOpsUsed[CommandIdx]++;
}
}
+static void UnescapeString(std::string &Str) {
+ for (unsigned i = 0; i != Str.size(); ++i) {
+ if (Str[i] == '\\' && i != Str.size()-1) {
+ switch (Str[i+1]) {
+ default: continue; // Don't execute the code after the switch.
+ case 'a': Str[i] = '\a'; break;
+ case 'b': Str[i] = '\b'; break;
+ case 'e': Str[i] = 27; break;
+ case 'f': Str[i] = '\f'; break;
+ case 'n': Str[i] = '\n'; break;
+ case 'r': Str[i] = '\r'; break;
+ case 't': Str[i] = '\t'; break;
+ case 'v': Str[i] = '\v'; break;
+ case '"': Str[i] = '\"'; break;
+ case '\'': Str[i] = '\''; break;
+ case '\\': Str[i] = '\\'; break;
+ }
+ // Nuke the second character.
+ Str.erase(Str.begin()+i+1);
+ }
+ }
+}
-void AsmWriterEmitter::run(raw_ostream &O) {
- EmitSourceFileHeader("Assembly Writer Source Fragment", O);
-
+/// EmitPrintInstruction - Generate the code for the "printInstruction" method
+/// implementation.
+void AsmWriterEmitter::EmitPrintInstruction(raw_ostream &O) {
CodeGenTarget Target;
Record *AsmWriter = Target.getAsmWriter();
std::string ClassName = AsmWriter->getValueAsString("AsmWriterClassName");
-
+
O <<
"/// printInstruction - This method is automatically generated by tablegen\n"
- "/// from the instruction set description. This method returns true if the\n"
- "/// machine instruction was sufficiently described to print it, otherwise\n"
- "/// it returns false.\n"
+ "/// from the instruction set description.\n"
"void " << Target.getName() << ClassName
<< "::printInstruction(const MachineInstr *MI) {\n";
for (CodeGenTarget::inst_iterator I = Target.inst_begin(),
E = Target.inst_end(); I != E; ++I)
- if (!I->second.AsmString.empty())
+ if (!I->second.AsmString.empty() &&
+ I->second.TheDef->getName() != "PHI")
Instructions.push_back(AsmWriterInst(I->second, AsmWriter));
// Get the instruction numbering.
CGIAWIMap.insert(std::make_pair(Instructions[i].CGI, &Instructions[i]));
// Build an aggregate string, and build a table of offsets into it.
- std::map<std::string, unsigned> StringOffset;
- std::string AggregateString;
- AggregateString.push_back(0); // "\0"
- AggregateString.push_back(0); // "\0"
+ StringToOffsetTable StringTable;
/// OpcodeInfo - This encodes the index of the string to use for the first
/// chunk of the output as well as indices used for operand printing.
unsigned Idx;
if (AWI == 0) {
// Something not handled by the asmwriter printer.
- Idx = 0;
+ Idx = ~0U;
} else if (AWI->Operands[0].OperandType !=
AsmWriterOperand::isLiteralTextOperand ||
AWI->Operands[0].Str.empty()) {
// Something handled by the asmwriter printer, but with no leading string.
- Idx = 1;
+ Idx = StringTable.GetOrAddStringOffset("");
} else {
- unsigned &Entry = StringOffset[AWI->Operands[0].Str];
- if (Entry == 0) {
- // Add the string to the aggregate if this is the first time found.
- MaxStringIdx = Entry = AggregateString.size();
- std::string Str = AWI->Operands[0].Str;
- UnescapeString(Str);
- AggregateString += Str;
- AggregateString += '\0';
- }
- Idx = Entry;
-
+ std::string Str = AWI->Operands[0].Str;
+ UnescapeString(Str);
+ Idx = StringTable.GetOrAddStringOffset(Str);
+ MaxStringIdx = std::max(MaxStringIdx, Idx);
+
// Nuke the string from the operand list. It is now handled!
AWI->Operands.erase(AWI->Operands.begin());
}
- OpcodeInfo.push_back(Idx);
+
+ // Bias offset by one since we want 0 as a sentinel.
+ OpcodeInfo.push_back(Idx+1);
}
// Figure out how many bits we used for the string index.
- unsigned AsmStrBits = Log2_32_Ceil(MaxStringIdx+1);
+ unsigned AsmStrBits = Log2_32_Ceil(MaxStringIdx+2);
// To reduce code size, we compactify common instructions into a few bits
// in the opcode-indexed table.
std::vector<std::vector<std::string> > TableDrivenOperandPrinters;
- bool isFirst = true;
while (1) {
std::vector<std::string> UniqueOperandCommands;
-
- // For the first operand check, add a default value for instructions with
- // just opcode strings to use.
- if (isFirst) {
- UniqueOperandCommands.push_back(" return;\n");
- isFirst = false;
- }
-
std::vector<unsigned> InstIdxs;
std::vector<unsigned> NumInstOpsHandled;
FindUniqueOperandCommands(UniqueOperandCommands, InstIdxs,
// If we don't have enough bits for this operand, don't include it.
if (NumBits > BitsLeft) {
- DOUT << "Not enough bits to densely encode " << NumBits
- << " more bits\n";
+ DEBUG(errs() << "Not enough bits to densely encode " << NumBits
+ << " more bits\n");
break;
}
O << " };\n\n";
// Emit the string itself.
- O << " const char *AsmStrs = \n \"";
- unsigned CharsPrinted = 0;
- EscapeString(AggregateString);
- for (unsigned i = 0, e = AggregateString.size(); i != e; ++i) {
- if (CharsPrinted > 70) {
- O << "\"\n \"";
- CharsPrinted = 0;
- }
- O << AggregateString[i];
- ++CharsPrinted;
-
- // Print escape sequences all together.
- if (AggregateString[i] == '\\') {
- assert(i+1 < AggregateString.size() && "Incomplete escape sequence!");
- if (isdigit(AggregateString[i+1])) {
- assert(isdigit(AggregateString[i+2]) && isdigit(AggregateString[i+3]) &&
- "Expected 3 digit octal escape!");
- O << AggregateString[++i];
- O << AggregateString[++i];
- O << AggregateString[++i];
- CharsPrinted += 3;
- } else {
- O << AggregateString[++i];
- ++CharsPrinted;
- }
- }
- }
- O << "\";\n\n";
-
- O << " processDebugLoc(MI->getDebugLoc());\n\n";
+ O << " const char *AsmStrs = \n";
+ StringTable.EmitString(O);
+ O << ";\n\n";
O << "\n#ifndef NO_ASM_WRITER_BOILERPLATE\n";
O << " if (MI->getOpcode() == TargetInstrInfo::INLINEASM) {\n"
- << " O << \"\\t\";\n"
<< " printInlineAsm(MI);\n"
<< " return;\n"
<< " } else if (MI->isLabel()) {\n"
<< " printLabel(MI);\n"
<< " return;\n"
- << " } else if (MI->getOpcode() == TargetInstrInfo::DECLARE) {\n"
- << " printDeclare(MI);\n"
- << " return;\n"
<< " } else if (MI->getOpcode() == TargetInstrInfo::IMPLICIT_DEF) {\n"
<< " printImplicitDef(MI);\n"
<< " return;\n"
+ << " } else if (MI->getOpcode() == TargetInstrInfo::KILL) {\n"
+ << " printKill(MI);\n"
+ << " return;\n"
<< " }\n\n";
O << "\n#endif\n";
O << " // Emit the opcode for the instruction.\n"
<< " unsigned Bits = OpInfo[MI->getOpcode()];\n"
<< " assert(Bits != 0 && \"Cannot print this instruction.\");\n"
- << " O << AsmStrs+(Bits & " << (1 << AsmStrBits)-1 << ");\n\n";
+ << " O << AsmStrs+(Bits & " << (1 << AsmStrBits)-1 << ")-1;\n\n";
// Output the table driven operand information.
BitsLeft = 32-AsmStrBits;
// elements in the vector.
std::reverse(Instructions.begin(), Instructions.end());
+
+ // Now that we've emitted all of the operand info that fit into 32 bits, emit
+ // information for those instructions that are left. This is a less dense
+ // encoding, but we expect the main 32-bit table to handle the majority of
+ // instructions.
if (!Instructions.empty()) {
// Find the opcode # of inline asm.
O << " switch (MI->getOpcode()) {\n";
O << " return;\n";
}
- O << " return;\n";
O << "}\n";
}
+
+
+void AsmWriterEmitter::EmitGetRegisterName(raw_ostream &O) {
+ CodeGenTarget Target;
+ Record *AsmWriter = Target.getAsmWriter();
+ std::string ClassName = AsmWriter->getValueAsString("AsmWriterClassName");
+ const std::vector<CodeGenRegister> &Registers = Target.getRegisters();
+
+ StringToOffsetTable StringTable;
+ O <<
+ "\n\n/// getRegisterName - This method is automatically generated by tblgen\n"
+ "/// from the register set description. This returns the assembler name\n"
+ "/// for the specified register.\n"
+ "const char *" << Target.getName() << ClassName
+ << "::getRegisterName(unsigned RegNo) {\n"
+ << " assert(RegNo && RegNo < " << (Registers.size()+1)
+ << " && \"Invalid register number!\");\n"
+ << "\n"
+ << " static const unsigned RegAsmOffset[] = {";
+ for (unsigned i = 0, e = Registers.size(); i != e; ++i) {
+ const CodeGenRegister &Reg = Registers[i];
+
+ std::string AsmName = Reg.TheDef->getValueAsString("AsmName");
+ if (AsmName.empty())
+ AsmName = Reg.getName();
+
+
+ if ((i % 14) == 0)
+ O << "\n ";
+
+ O << StringTable.GetOrAddStringOffset(AsmName) << ", ";
+ }
+ O << "0\n"
+ << " };\n"
+ << "\n";
+
+ O << " const char *AsmStrs =\n";
+ StringTable.EmitString(O);
+ O << ";\n";
+
+ O << " return AsmStrs+RegAsmOffset[RegNo-1];\n"
+ << "}\n";
+}
+
+
+void AsmWriterEmitter::run(raw_ostream &O) {
+ EmitSourceFileHeader("Assembly Writer Source Fragment", O);
+
+ EmitPrintInstruction(O);
+ EmitGetRegisterName(O);
+}
+