#include "llvm/Target/TargetAsmParser.h"
#include "X86.h"
#include "X86Subtarget.h"
-#include "llvm/ADT/SmallString.h"
-#include "llvm/ADT/SmallVector.h"
-#include "llvm/ADT/StringSwitch.h"
-#include "llvm/ADT/Twine.h"
+#include "llvm/Target/TargetRegistry.h"
+#include "llvm/Target/TargetAsmParser.h"
#include "llvm/MC/MCStreamer.h"
#include "llvm/MC/MCExpr.h"
#include "llvm/MC/MCInst.h"
#include "llvm/MC/MCParser/MCAsmLexer.h"
#include "llvm/MC/MCParser/MCAsmParser.h"
#include "llvm/MC/MCParser/MCParsedAsmOperand.h"
+#include "llvm/ADT/SmallString.h"
+#include "llvm/ADT/SmallVector.h"
+#include "llvm/ADT/StringExtras.h"
+#include "llvm/ADT/StringSwitch.h"
+#include "llvm/ADT/Twine.h"
#include "llvm/Support/SourceMgr.h"
#include "llvm/Support/raw_ostream.h"
-#include "llvm/Target/TargetRegistry.h"
-#include "llvm/Target/TargetAsmParser.h"
using namespace llvm;
namespace {
bool Error(SMLoc L, const Twine &Msg) { return Parser.Error(L, Msg); }
- bool ParseRegister(unsigned &RegNo, SMLoc &StartLoc, SMLoc &EndLoc);
-
X86Operand *ParseOperand();
X86Operand *ParseMemOperand(unsigned SegReg, SMLoc StartLoc);
bool ParseDirectiveWord(unsigned Size, SMLoc L);
- bool MatchInstruction(SMLoc IDLoc,
- const SmallVectorImpl<MCParsedAsmOperand*> &Operands,
- MCInst &Inst);
+ bool MatchAndEmitInstruction(SMLoc IDLoc,
+ SmallVectorImpl<MCParsedAsmOperand*> &Operands,
+ MCStreamer &Out);
+
+ /// isSrcOp - Returns true if operand is either (%rsi) or %ds:%(rsi)
+ /// in 64bit mode or (%edi) or %es:(%edi) in 32bit mode.
+ bool isSrcOp(X86Operand &Op);
+
+ /// isDstOp - Returns true if operand is either %es:(%rdi) in 64bit mode
+ /// or %es:(%edi) in 32bit mode.
+ bool isDstOp(X86Operand &Op);
/// @name Auto-generated Matcher Functions
/// {
-
+
#define GET_ASSEMBLER_HEADER
#include "X86GenAsmMatcher.inc"
-
+
/// }
public:
- X86ATTAsmParser(const Target &T, MCAsmParser &_Parser, TargetMachine &TM)
- : TargetAsmParser(T), Parser(_Parser), TM(TM) {
+ X86ATTAsmParser(const Target &T, MCAsmParser &parser, TargetMachine &TM)
+ : TargetAsmParser(T), Parser(parser), TM(TM) {
// Initialize the set of available features.
setAvailableFeatures(ComputeAvailableFeatures(
&TM.getSubtarget<X86Subtarget>()));
}
+ virtual bool ParseRegister(unsigned &RegNo, SMLoc &StartLoc, SMLoc &EndLoc);
virtual bool ParseInstruction(StringRef Name, SMLoc NameLoc,
SmallVectorImpl<MCParsedAsmOperand*> &Operands);
class X86_32ATTAsmParser : public X86ATTAsmParser {
public:
- X86_32ATTAsmParser(const Target &T, MCAsmParser &_Parser, TargetMachine &TM)
- : X86ATTAsmParser(T, _Parser, TM) {
+ X86_32ATTAsmParser(const Target &T, MCAsmParser &Parser, TargetMachine &TM)
+ : X86ATTAsmParser(T, Parser, TM) {
Is64Bit = false;
}
};
class X86_64ATTAsmParser : public X86ATTAsmParser {
public:
- X86_64ATTAsmParser(const Target &T, MCAsmParser &_Parser, TargetMachine &TM)
- : X86ATTAsmParser(T, _Parser, TM) {
+ X86_64ATTAsmParser(const Target &T, MCAsmParser &Parser, TargetMachine &TM)
+ : X86ATTAsmParser(T, Parser, TM) {
Is64Bit = true;
}
};
} // end anonymous namespace.
+bool X86ATTAsmParser::isSrcOp(X86Operand &Op) {
+ unsigned basereg = Is64Bit ? X86::RSI : X86::ESI;
+
+ return (Op.isMem() &&
+ (Op.Mem.SegReg == 0 || Op.Mem.SegReg == X86::DS) &&
+ isa<MCConstantExpr>(Op.Mem.Disp) &&
+ cast<MCConstantExpr>(Op.Mem.Disp)->getValue() == 0 &&
+ Op.Mem.BaseReg == basereg && Op.Mem.IndexReg == 0);
+}
+
+bool X86ATTAsmParser::isDstOp(X86Operand &Op) {
+ unsigned basereg = Is64Bit ? X86::RDI : X86::EDI;
+
+ return Op.isMem() && Op.Mem.SegReg == X86::ES &&
+ isa<MCConstantExpr>(Op.Mem.Disp) &&
+ cast<MCConstantExpr>(Op.Mem.Disp)->getValue() == 0 &&
+ Op.Mem.BaseReg == basereg && Op.Mem.IndexReg == 0;
+}
bool X86ATTAsmParser::ParseRegister(unsigned &RegNo,
SMLoc &StartLoc, SMLoc &EndLoc) {
// validation later, so maybe there is no need for this here.
RegNo = MatchRegisterName(Tok.getString());
+ // If the match failed, try the register name as lowercase.
+ if (RegNo == 0)
+ RegNo = MatchRegisterName(LowercaseString(Tok.getString()));
+
// FIXME: This should be done using Requires<In32BitMode> and
// Requires<In64BitMode> so "eiz" usage in 64-bit instructions
// can be also checked.
if (RegNo == X86::RIZ && !Is64Bit)
return Error(Tok.getLoc(), "riz register in 64-bit mode only");
- // Parse %st(1) and "%st" as "%st(0)"
- if (RegNo == 0 && Tok.getString() == "st") {
+ // Parse "%st" as "%st(0)" and "%st(1)", which is multiple tokens.
+ if (RegNo == 0 && (Tok.getString() == "st" || Tok.getString() == "ST")) {
RegNo = X86::ST0;
EndLoc = Tok.getLoc();
Parser.Lex(); // Eat 'st'
bool X86ATTAsmParser::
ParseInstruction(StringRef Name, SMLoc NameLoc,
SmallVectorImpl<MCParsedAsmOperand*> &Operands) {
- // FIXME: Hack to recognize "sal..." and "rep..." for now. We need a way to
- // represent alternative syntaxes in the .td file, without requiring
- // instruction duplication.
- StringRef PatchedName = StringSwitch<StringRef>(Name)
- .Case("sal", "shl")
- .Case("salb", "shlb")
- .Case("sall", "shll")
- .Case("salq", "shlq")
- .Case("salw", "shlw")
- .Case("repe", "rep")
- .Case("repz", "rep")
- .Case("repnz", "repne")
- .Case("iret", "iretl")
- .Case("sysret", "sysretl")
- .Case("push", Is64Bit ? "pushq" : "pushl")
- .Case("pop", Is64Bit ? "popq" : "popl")
- .Case("pushf", Is64Bit ? "pushfq" : "pushfl")
- .Case("popf", Is64Bit ? "popfq" : "popfl")
- .Case("pushfd", "pushfl")
- .Case("popfd", "popfl")
- .Case("retl", Is64Bit ? "retl" : "ret")
- .Case("retq", Is64Bit ? "ret" : "retq")
- .Case("setz", "sete")
- .Case("setnz", "setne")
- .Case("jz", "je")
- .Case("jnz", "jne")
- .Case("jc", "jb")
- .Case("jna", "jbe")
- .Case("jnae", "jb")
- .Case("jnb", "jae")
- .Case("jnbe", "ja")
- .Case("jnc", "jae")
- .Case("jng", "jle")
- .Case("jnge", "jl")
- .Case("jnl", "jge")
- .Case("jnle", "jg")
- .Case("jpe", "jp")
- .Case("jpo", "jnp")
- // Condition code aliases for 16-bit, 32-bit, 64-bit and unspec operands.
- .Case("cmovcw", "cmovbw") .Case("cmovcl", "cmovbl")
- .Case("cmovcq", "cmovbq") .Case("cmovc", "cmovb")
- .Case("cmovnaw", "cmovbew").Case("cmovnal", "cmovbel")
- .Case("cmovnaq", "cmovbeq").Case("cmovna", "cmovbe")
- .Case("cmovnbw", "cmovaew").Case("cmovnbl", "cmovael")
- .Case("cmovnbq", "cmovaeq").Case("cmovnb", "cmovae")
- .Case("cmovnbew","cmovaw") .Case("cmovnbel","cmoval")
- .Case("cmovnbeq","cmovaq") .Case("cmovnbe", "cmova")
- .Case("cmovncw", "cmovaew").Case("cmovncl", "cmovael")
- .Case("cmovncq", "cmovaeq").Case("cmovnc", "cmovae")
- .Case("cmovngw", "cmovlew").Case("cmovngl", "cmovlel")
- .Case("cmovngq", "cmovleq").Case("cmovng", "cmovle")
- .Case("cmovnw", "cmovgew").Case("cmovnl", "cmovgel")
- .Case("cmovnq", "cmovgeq").Case("cmovn", "cmovge")
- .Case("cmovngw", "cmovlew").Case("cmovngl", "cmovlel")
- .Case("cmovngq", "cmovleq").Case("cmovng", "cmovle")
- .Case("cmovngew","cmovlw") .Case("cmovngel","cmovll")
- .Case("cmovngeq","cmovlq") .Case("cmovnge", "cmovl")
- .Case("cmovnlw", "cmovgew").Case("cmovnll", "cmovgel")
- .Case("cmovnlq", "cmovgeq").Case("cmovnl", "cmovge")
- .Case("cmovnlew","cmovgw") .Case("cmovnlel","cmovgl")
- .Case("cmovnleq","cmovgq") .Case("cmovnle", "cmovg")
- .Case("cmovnzw", "cmovnew").Case("cmovnzl", "cmovnel")
- .Case("cmovnzq", "cmovneq").Case("cmovnz", "cmovne")
- .Case("cmovzw", "cmovew") .Case("cmovzl", "cmovel")
- .Case("cmovzq", "cmoveq") .Case("cmovz", "cmove")
- .Case("fwait", "wait")
- .Case("movzx", "movzb")
- .Default(Name);
+ StringRef PatchedName = Name;
+ // FIXME: Hack to recognize setneb as setne.
+ if (PatchedName.startswith("set") && PatchedName.endswith("b") &&
+ PatchedName != "setb" && PatchedName != "setnb")
+ PatchedName = PatchedName.substr(0, Name.size()-1);
+
// FIXME: Hack to recognize cmp<comparison code>{ss,sd,ps,pd}.
const MCExpr *ExtraImmOp = 0;
if ((PatchedName.startswith("cmp") || PatchedName.startswith("vcmp")) &&
}
}
- // FIXME: Hack to recognize vpclmul<src1_quadword, src2_quadword>dq
- if (PatchedName.startswith("vpclmul")) {
- unsigned CLMULQuadWordSelect = StringSwitch<unsigned>(
- PatchedName.slice(7, PatchedName.size() - 2))
- .Case("lqlq", 0x00) // src1[63:0], src2[63:0]
- .Case("hqlq", 0x01) // src1[127:64], src2[63:0]
- .Case("lqhq", 0x10) // src1[63:0], src2[127:64]
- .Case("hqhq", 0x11) // src1[127:64], src2[127:64]
- .Default(~0U);
- if (CLMULQuadWordSelect != ~0U) {
- ExtraImmOp = MCConstantExpr::Create(CLMULQuadWordSelect,
- getParser().getContext());
- assert(PatchedName.endswith("dq") && "Unexpected mnemonic!");
- PatchedName = "vpclmulqdq";
- }
- }
-
Operands.push_back(X86Operand::CreateToken(PatchedName, NameLoc));
if (ExtraImmOp)
Operands.push_back(X86Operand::CreateImm(ExtraImmOp, NameLoc, NameLoc));
-
-
+
+
// Determine whether this is an instruction prefix.
bool isPrefix =
- PatchedName == "lock" || PatchedName == "rep" ||
- PatchedName == "repne";
-
-
+ Name == "lock" || Name == "rep" ||
+ Name == "repe" || Name == "repz" ||
+ Name == "repne" || Name == "repnz" ||
+ Name == "rex64" || Name == "data16";
+
+
// This does the actual operand parsing. Don't parse any more if we have a
// prefix juxtaposed with an operation like "lock incl 4(%rax)", because we
// just want to parse the "lock" as the first instruction and the "incl" as
return true;
}
}
-
+
if (getLexer().isNot(AsmToken::EndOfStatement)) {
+ SMLoc Loc = getLexer().getLoc();
Parser.EatToEndOfStatement();
- return TokError("unexpected token in argument list");
+ return Error(Loc, "unexpected token in argument list");
}
}
-
+
if (getLexer().is(AsmToken::EndOfStatement))
Parser.Lex(); // Consume the EndOfStatement
+ else if (isPrefix && getLexer().is(AsmToken::Slash))
+ Parser.Lex(); // Consume the prefix separator Slash
- // FIXME: Hack to handle recognize s{hr,ar,hl} <op>, $1. Canonicalize to
- // "shift <op>".
- if ((Name.startswith("shr") || Name.startswith("sar") ||
- Name.startswith("shl")) &&
+ // This is a terrible hack to handle "out[bwl]? %al, (%dx)" ->
+ // "outb %al, %dx". Out doesn't take a memory form, but this is a widely
+ // documented form in various unofficial manuals, so a lot of code uses it.
+ if ((Name == "outb" || Name == "outw" || Name == "outl" || Name == "out") &&
Operands.size() == 3) {
- X86Operand *Op1 = static_cast<X86Operand*>(Operands[1]);
- if (Op1->isImm() && isa<MCConstantExpr>(Op1->getImm()) &&
- cast<MCConstantExpr>(Op1->getImm())->getValue() == 1) {
- delete Operands[1];
- Operands.erase(Operands.begin() + 1);
+ X86Operand &Op = *(X86Operand*)Operands.back();
+ if (Op.isMem() && Op.Mem.SegReg == 0 &&
+ isa<MCConstantExpr>(Op.Mem.Disp) &&
+ cast<MCConstantExpr>(Op.Mem.Disp)->getValue() == 0 &&
+ Op.Mem.BaseReg == MatchRegisterName("dx") && Op.Mem.IndexReg == 0) {
+ SMLoc Loc = Op.getEndLoc();
+ Operands.back() = X86Operand::CreateReg(Op.Mem.BaseReg, Loc, Loc);
+ delete &Op;
}
}
-
- // FIXME: Hack to handle recognize "in[bwl] <op>". Canonicalize it to
- // "inb <op>, %al".
- if ((Name == "inb" || Name == "inw" || Name == "inl") &&
- Operands.size() == 2) {
- unsigned Reg;
- if (Name[2] == 'b')
- Reg = MatchRegisterName("al");
- else if (Name[2] == 'w')
- Reg = MatchRegisterName("ax");
- else
- Reg = MatchRegisterName("eax");
- SMLoc Loc = Operands.back()->getEndLoc();
- Operands.push_back(X86Operand::CreateReg(Reg, Loc, Loc));
+ // Same hack for "in[bwl]? (%dx), %al" -> "inb %dx, %al".
+ if ((Name == "inb" || Name == "inw" || Name == "inl" || Name == "in") &&
+ Operands.size() == 3) {
+ X86Operand &Op = *(X86Operand*)Operands.begin()[1];
+ if (Op.isMem() && Op.Mem.SegReg == 0 &&
+ isa<MCConstantExpr>(Op.Mem.Disp) &&
+ cast<MCConstantExpr>(Op.Mem.Disp)->getValue() == 0 &&
+ Op.Mem.BaseReg == MatchRegisterName("dx") && Op.Mem.IndexReg == 0) {
+ SMLoc Loc = Op.getEndLoc();
+ Operands.begin()[1] = X86Operand::CreateReg(Op.Mem.BaseReg, Loc, Loc);
+ delete &Op;
+ }
}
-
- // FIXME: Hack to handle recognize "out[bwl] <op>". Canonicalize it to
- // "outb %al, <op>".
- if ((Name == "outb" || Name == "outw" || Name == "outl") &&
- Operands.size() == 2) {
- unsigned Reg;
- if (Name[3] == 'b')
- Reg = MatchRegisterName("al");
- else if (Name[3] == 'w')
- Reg = MatchRegisterName("ax");
- else
- Reg = MatchRegisterName("eax");
- SMLoc Loc = Operands.back()->getEndLoc();
- Operands.push_back(X86Operand::CreateReg(Reg, Loc, Loc));
- std::swap(Operands[1], Operands[2]);
+ // Transform "ins[bwl] %dx, %es:(%edi)" into "ins[bwl]"
+ if (Name.startswith("ins") && Operands.size() == 3 &&
+ (Name == "insb" || Name == "insw" || Name == "insl")) {
+ X86Operand &Op = *(X86Operand*)Operands.begin()[1];
+ X86Operand &Op2 = *(X86Operand*)Operands.begin()[2];
+ if (Op.isReg() && Op.getReg() == X86::DX && isDstOp(Op2)) {
+ Operands.pop_back();
+ Operands.pop_back();
+ delete &Op;
+ delete &Op2;
+ }
}
-
- // FIXME: Hack to handle "f{mul*,add*,sub*,div*} $op, st(0)" the same as
- // "f{mul*,add*,sub*,div*} $op"
- if ((Name.startswith("fmul") || Name.startswith("fadd") ||
- Name.startswith("fsub") || Name.startswith("fdiv")) &&
- Operands.size() == 3 &&
- static_cast<X86Operand*>(Operands[2])->isReg() &&
- static_cast<X86Operand*>(Operands[2])->getReg() == X86::ST0) {
- delete Operands[2];
- Operands.erase(Operands.begin() + 2);
+
+ // Transform "outs[bwl] %ds:(%esi), %dx" into "out[bwl]"
+ if (Name.startswith("outs") && Operands.size() == 3 &&
+ (Name == "outsb" || Name == "outsw" || Name == "outsl")) {
+ X86Operand &Op = *(X86Operand*)Operands.begin()[1];
+ X86Operand &Op2 = *(X86Operand*)Operands.begin()[2];
+ if (isSrcOp(Op) && Op2.isReg() && Op2.getReg() == X86::DX) {
+ Operands.pop_back();
+ Operands.pop_back();
+ delete &Op;
+ delete &Op2;
+ }
}
- // FIXME: Hack to handle "imul <imm>, B" which is an alias for "imul <imm>, B,
- // B".
- if (Name.startswith("imul") && Operands.size() == 3 &&
- static_cast<X86Operand*>(Operands[1])->isImm() &&
- static_cast<X86Operand*>(Operands.back())->isReg()) {
- X86Operand *Op = static_cast<X86Operand*>(Operands.back());
- Operands.push_back(X86Operand::CreateReg(Op->getReg(), Op->getStartLoc(),
- Op->getEndLoc()));
+ // Transform "movs[bwl] %ds:(%esi), %es:(%edi)" into "movs[bwl]"
+ if (Name.startswith("movs") && Operands.size() == 3 &&
+ (Name == "movsb" || Name == "movsw" || Name == "movsl" ||
+ (Is64Bit && Name == "movsq"))) {
+ X86Operand &Op = *(X86Operand*)Operands.begin()[1];
+ X86Operand &Op2 = *(X86Operand*)Operands.begin()[2];
+ if (isSrcOp(Op) && isDstOp(Op2)) {
+ Operands.pop_back();
+ Operands.pop_back();
+ delete &Op;
+ delete &Op2;
+ }
}
-
- // 'sldt <mem>' can be encoded with either sldtw or sldtq with the same
- // effect (both store to a 16-bit mem). Force to sldtw to avoid ambiguity
- // errors, since its encoding is the most compact.
- if (Name == "sldt" && Operands.size() == 2 &&
- static_cast<X86Operand*>(Operands[1])->isMem()) {
- delete Operands[0];
- Operands[0] = X86Operand::CreateToken("sldtw", NameLoc);
+ // Transform "lods[bwl] %ds:(%esi),{%al,%ax,%eax,%rax}" into "lods[bwl]"
+ if (Name.startswith("lods") && Operands.size() == 3 &&
+ (Name == "lods" || Name == "lodsb" || Name == "lodsw" ||
+ Name == "lodsl" || (Is64Bit && Name == "lodsq"))) {
+ X86Operand *Op1 = static_cast<X86Operand*>(Operands[1]);
+ X86Operand *Op2 = static_cast<X86Operand*>(Operands[2]);
+ if (isSrcOp(*Op1) && Op2->isReg()) {
+ const char *ins;
+ unsigned reg = Op2->getReg();
+ bool isLods = Name == "lods";
+ if (reg == X86::AL && (isLods || Name == "lodsb"))
+ ins = "lodsb";
+ else if (reg == X86::AX && (isLods || Name == "lodsw"))
+ ins = "lodsw";
+ else if (reg == X86::EAX && (isLods || Name == "lodsl"))
+ ins = "lodsl";
+ else if (reg == X86::RAX && (isLods || Name == "lodsq"))
+ ins = "lodsq";
+ else
+ ins = NULL;
+ if (ins != NULL) {
+ Operands.pop_back();
+ Operands.pop_back();
+ delete Op1;
+ delete Op2;
+ if (Name != ins)
+ static_cast<X86Operand*>(Operands[0])->setTokenValue(ins);
+ }
+ }
}
-
- // The assembler accepts "xchgX <reg>, <mem>" and "xchgX <mem>, <reg>" as
- // synonyms. Our tables only have the "<reg>, <mem>" form, so if we see the
- // other operand order, swap them.
- if (Name == "xchgb" || Name == "xchgw" || Name == "xchgl" || Name == "xchgq"||
- Name == "xchg")
- if (Operands.size() == 3 &&
- static_cast<X86Operand*>(Operands[1])->isMem() &&
- static_cast<X86Operand*>(Operands[2])->isReg()) {
- std::swap(Operands[1], Operands[2]);
+ // Transform "stos[bwl] {%al,%ax,%eax,%rax},%es:(%edi)" into "stos[bwl]"
+ if (Name.startswith("stos") && Operands.size() == 3 &&
+ (Name == "stos" || Name == "stosb" || Name == "stosw" ||
+ Name == "stosl" || (Is64Bit && Name == "stosq"))) {
+ X86Operand *Op1 = static_cast<X86Operand*>(Operands[1]);
+ X86Operand *Op2 = static_cast<X86Operand*>(Operands[2]);
+ if (isDstOp(*Op2) && Op1->isReg()) {
+ const char *ins;
+ unsigned reg = Op1->getReg();
+ bool isStos = Name == "stos";
+ if (reg == X86::AL && (isStos || Name == "stosb"))
+ ins = "stosb";
+ else if (reg == X86::AX && (isStos || Name == "stosw"))
+ ins = "stosw";
+ else if (reg == X86::EAX && (isStos || Name == "stosl"))
+ ins = "stosl";
+ else if (reg == X86::RAX && (isStos || Name == "stosq"))
+ ins = "stosq";
+ else
+ ins = NULL;
+ if (ins != NULL) {
+ Operands.pop_back();
+ Operands.pop_back();
+ delete Op1;
+ delete Op2;
+ if (Name != ins)
+ static_cast<X86Operand*>(Operands[0])->setTokenValue(ins);
+ }
}
+ }
- // The assembler accepts "testX <reg>, <mem>" and "testX <mem>, <reg>" as
- // synonyms. Our tables only have the "<mem>, <reg>" form, so if we see the
- // other operand order, swap them.
- if (Name == "testb" || Name == "testw" || Name == "testl" || Name == "testq"||
- Name == "test")
- if (Operands.size() == 3 &&
- static_cast<X86Operand*>(Operands[1])->isReg() &&
- static_cast<X86Operand*>(Operands[2])->isMem()) {
- std::swap(Operands[1], Operands[2]);
+ // FIXME: Hack to handle recognize s{hr,ar,hl} $1, <op>. Canonicalize to
+ // "shift <op>".
+ if ((Name.startswith("shr") || Name.startswith("sar") ||
+ Name.startswith("shl") || Name.startswith("sal") ||
+ Name.startswith("rcl") || Name.startswith("rcr") ||
+ Name.startswith("rol") || Name.startswith("ror")) &&
+ Operands.size() == 3) {
+ X86Operand *Op1 = static_cast<X86Operand*>(Operands[1]);
+ if (Op1->isImm() && isa<MCConstantExpr>(Op1->getImm()) &&
+ cast<MCConstantExpr>(Op1->getImm())->getValue() == 1) {
+ delete Operands[1];
+ Operands.erase(Operands.begin() + 1);
}
+ }
- return false;
-}
-
-bool X86ATTAsmParser::ParseDirective(AsmToken DirectiveID) {
- StringRef IDVal = DirectiveID.getIdentifier();
- if (IDVal == ".word")
- return ParseDirectiveWord(2, DirectiveID.getLoc());
- return true;
-}
-
-/// ParseDirectiveWord
-/// ::= .word [ expression (, expression)* ]
-bool X86ATTAsmParser::ParseDirectiveWord(unsigned Size, SMLoc L) {
- if (getLexer().isNot(AsmToken::EndOfStatement)) {
- for (;;) {
- const MCExpr *Value;
- if (getParser().ParseExpression(Value))
- return true;
-
- getParser().getStreamer().EmitValue(Value, Size, 0 /*addrspace*/);
-
- if (getLexer().is(AsmToken::EndOfStatement))
- break;
-
- // FIXME: Improve diagnostic.
- if (getLexer().isNot(AsmToken::Comma))
- return Error(L, "unexpected token in directive");
- Parser.Lex();
+ // Transforms "int $3" into "int3" as a size optimization. We can't write an
+ // instalias with an immediate operand yet.
+ if (Name == "int" && Operands.size() == 2) {
+ X86Operand *Op1 = static_cast<X86Operand*>(Operands[1]);
+ if (Op1->isImm() && isa<MCConstantExpr>(Op1->getImm()) &&
+ cast<MCConstantExpr>(Op1->getImm())->getValue() == 3) {
+ delete Operands[1];
+ Operands.erase(Operands.begin() + 1);
+ static_cast<X86Operand*>(Operands[0])->setTokenValue("int3");
}
}
- Parser.Lex();
return false;
}
-
-bool
-X86ATTAsmParser::MatchInstruction(SMLoc IDLoc,
- const SmallVectorImpl<MCParsedAsmOperand*>
- &Operands,
- MCInst &Inst) {
+bool X86ATTAsmParser::
+MatchAndEmitInstruction(SMLoc IDLoc,
+ SmallVectorImpl<MCParsedAsmOperand*> &Operands,
+ MCStreamer &Out) {
assert(!Operands.empty() && "Unexpect empty operand list!");
+ X86Operand *Op = static_cast<X86Operand*>(Operands[0]);
+ assert(Op->isToken() && "Leading operand should always be a mnemonic!");
+
+ // First, handle aliases that expand to multiple instructions.
+ // FIXME: This should be replaced with a real .td file alias mechanism.
+ // Also, MatchInstructionImpl should do actually *do* the EmitInstruction
+ // call.
+ if (Op->getToken() == "fstsw" || Op->getToken() == "fstcw" ||
+ Op->getToken() == "fstsww" || Op->getToken() == "fstcww" ||
+ Op->getToken() == "finit" || Op->getToken() == "fsave" ||
+ Op->getToken() == "fstenv" || Op->getToken() == "fclex") {
+ MCInst Inst;
+ Inst.setOpcode(X86::WAIT);
+ Out.EmitInstruction(Inst);
+
+ const char *Repl =
+ StringSwitch<const char*>(Op->getToken())
+ .Case("finit", "fninit")
+ .Case("fsave", "fnsave")
+ .Case("fstcw", "fnstcw")
+ .Case("fstcww", "fnstcw")
+ .Case("fstenv", "fnstenv")
+ .Case("fstsw", "fnstsw")
+ .Case("fstsww", "fnstsw")
+ .Case("fclex", "fnclex")
+ .Default(0);
+ assert(Repl && "Unknown wait-prefixed instruction");
+ delete Operands[0];
+ Operands[0] = X86Operand::CreateToken(Repl, IDLoc);
+ }
bool WasOriginallyInvalidOperand = false;
unsigned OrigErrorInfo;
-
+ MCInst Inst;
+
// First, try a direct match.
switch (MatchInstructionImpl(Operands, Inst, OrigErrorInfo)) {
case Match_Success:
+ Out.EmitInstruction(Inst);
return false;
case Match_MissingFeature:
Error(IDLoc, "instruction requires a CPU feature not currently enabled");
return true;
+ case Match_ConversionFail:
+ return Error(IDLoc, "unable to convert operands to instruction");
case Match_InvalidOperand:
WasOriginallyInvalidOperand = true;
break;
// type. However, that requires substantially more matcher support than the
// following hack.
- X86Operand *Op = static_cast<X86Operand*>(Operands[0]);
- assert(Op->isToken() && "Leading operand should always be a mnemonic!");
-
// Change the operand to point to a temporary token.
StringRef Base = Op->getToken();
SmallString<16> Tmp;
Tmp += ' ';
Op->setTokenValue(Tmp.str());
+ // If this instruction starts with an 'f', then it is a floating point stack
+ // instruction. These come in up to three forms for 32-bit, 64-bit, and
+ // 80-bit floating point, which use the suffixes s,l,t respectively.
+ //
+ // Otherwise, we assume that this may be an integer instruction, which comes
+ // in 8/16/32/64-bit forms using the b,w,l,q suffixes respectively.
+ const char *Suffixes = Base[0] != 'f' ? "bwlq" : "slt\0";
+
// Check for the various suffix matches.
- Tmp[Base.size()] = 'b';
- unsigned BErrorInfo, WErrorInfo, LErrorInfo, QErrorInfo;
- MatchResultTy MatchB = MatchInstructionImpl(Operands, Inst, BErrorInfo);
- Tmp[Base.size()] = 'w';
- MatchResultTy MatchW = MatchInstructionImpl(Operands, Inst, WErrorInfo);
- Tmp[Base.size()] = 'l';
- MatchResultTy MatchL = MatchInstructionImpl(Operands, Inst, LErrorInfo);
- Tmp[Base.size()] = 'q';
- MatchResultTy MatchQ = MatchInstructionImpl(Operands, Inst, QErrorInfo);
+ Tmp[Base.size()] = Suffixes[0];
+ unsigned ErrorInfoIgnore;
+ MatchResultTy Match1, Match2, Match3, Match4;
+
+ Match1 = MatchInstructionImpl(Operands, Inst, ErrorInfoIgnore);
+ Tmp[Base.size()] = Suffixes[1];
+ Match2 = MatchInstructionImpl(Operands, Inst, ErrorInfoIgnore);
+ Tmp[Base.size()] = Suffixes[2];
+ Match3 = MatchInstructionImpl(Operands, Inst, ErrorInfoIgnore);
+ Tmp[Base.size()] = Suffixes[3];
+ Match4 = MatchInstructionImpl(Operands, Inst, ErrorInfoIgnore);
// Restore the old token.
Op->setTokenValue(Base);
// instruction will already have been filled in correctly, since the failing
// matches won't have modified it).
unsigned NumSuccessfulMatches =
- (MatchB == Match_Success) + (MatchW == Match_Success) +
- (MatchL == Match_Success) + (MatchQ == Match_Success);
- if (NumSuccessfulMatches == 1)
+ (Match1 == Match_Success) + (Match2 == Match_Success) +
+ (Match3 == Match_Success) + (Match4 == Match_Success);
+ if (NumSuccessfulMatches == 1) {
+ Out.EmitInstruction(Inst);
return false;
+ }
// Otherwise, the match failed, try to produce a decent error message.
if (NumSuccessfulMatches > 1) {
char MatchChars[4];
unsigned NumMatches = 0;
- if (MatchB == Match_Success)
- MatchChars[NumMatches++] = 'b';
- if (MatchW == Match_Success)
- MatchChars[NumMatches++] = 'w';
- if (MatchL == Match_Success)
- MatchChars[NumMatches++] = 'l';
- if (MatchQ == Match_Success)
- MatchChars[NumMatches++] = 'q';
+ if (Match1 == Match_Success) MatchChars[NumMatches++] = Suffixes[0];
+ if (Match2 == Match_Success) MatchChars[NumMatches++] = Suffixes[1];
+ if (Match3 == Match_Success) MatchChars[NumMatches++] = Suffixes[2];
+ if (Match4 == Match_Success) MatchChars[NumMatches++] = Suffixes[3];
SmallString<126> Msg;
raw_svector_ostream OS(Msg);
Error(IDLoc, OS.str());
return true;
}
-
+
// Okay, we know that none of the variants matched successfully.
-
+
// If all of the instructions reported an invalid mnemonic, then the original
// mnemonic was invalid.
- if ((MatchB == Match_MnemonicFail) && (MatchW == Match_MnemonicFail) &&
- (MatchL == Match_MnemonicFail) && (MatchQ == Match_MnemonicFail)) {
+ if ((Match1 == Match_MnemonicFail) && (Match2 == Match_MnemonicFail) &&
+ (Match3 == Match_MnemonicFail) && (Match4 == Match_MnemonicFail)) {
if (!WasOriginallyInvalidOperand) {
- Error(IDLoc, "invalid instruction mnemonic '" + Base + "'");
+ Error(IDLoc, "invalid instruction mnemonic '" + Base + "'");
return true;
}
// Recover location info for the operand if we know which was the problem.
SMLoc ErrorLoc = IDLoc;
if (OrigErrorInfo != ~0U) {
+ if (OrigErrorInfo >= Operands.size())
+ return Error(IDLoc, "too few operands for instruction");
+
ErrorLoc = ((X86Operand*)Operands[OrigErrorInfo])->getStartLoc();
if (ErrorLoc == SMLoc()) ErrorLoc = IDLoc;
}
- Error(ErrorLoc, "invalid operand for instruction");
- return true;
+ return Error(ErrorLoc, "invalid operand for instruction");
}
-
+
// If one instruction matched with a missing feature, report this as a
// missing feature.
- if ((MatchB == Match_MissingFeature) + (MatchW == Match_MissingFeature) +
- (MatchL == Match_MissingFeature) + (MatchQ == Match_MissingFeature) == 1){
+ if ((Match1 == Match_MissingFeature) + (Match2 == Match_MissingFeature) +
+ (Match3 == Match_MissingFeature) + (Match4 == Match_MissingFeature) == 1){
Error(IDLoc, "instruction requires a CPU feature not currently enabled");
return true;
}
-
+
// If one instruction matched with an invalid operand, report this as an
// operand failure.
- if ((MatchB == Match_InvalidOperand) + (MatchW == Match_InvalidOperand) +
- (MatchL == Match_InvalidOperand) + (MatchQ == Match_InvalidOperand) == 1){
+ if ((Match1 == Match_InvalidOperand) + (Match2 == Match_InvalidOperand) +
+ (Match3 == Match_InvalidOperand) + (Match4 == Match_InvalidOperand) == 1){
Error(IDLoc, "invalid operand for instruction");
return true;
}
-
+
// If all of these were an outright failure, report it in a useless way.
// FIXME: We should give nicer diagnostics about the exact failure.
Error(IDLoc, "unknown use of instruction mnemonic without a size suffix");
}
+bool X86ATTAsmParser::ParseDirective(AsmToken DirectiveID) {
+ StringRef IDVal = DirectiveID.getIdentifier();
+ if (IDVal == ".word")
+ return ParseDirectiveWord(2, DirectiveID.getLoc());
+ return true;
+}
+
+/// ParseDirectiveWord
+/// ::= .word [ expression (, expression)* ]
+bool X86ATTAsmParser::ParseDirectiveWord(unsigned Size, SMLoc L) {
+ if (getLexer().isNot(AsmToken::EndOfStatement)) {
+ for (;;) {
+ const MCExpr *Value;
+ if (getParser().ParseExpression(Value))
+ return true;
+
+ getParser().getStreamer().EmitValue(Value, Size, 0 /*addrspace*/);
+
+ if (getLexer().is(AsmToken::EndOfStatement))
+ break;
+
+ // FIXME: Improve diagnostic.
+ if (getLexer().isNot(AsmToken::Comma))
+ return Error(L, "unexpected token in directive");
+ Parser.Lex();
+ }
+ }
+
+ Parser.Lex();
+ return false;
+}
+
+
+
+
extern "C" void LLVMInitializeX86AsmLexer();
// Force static initialization.