#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 ParseDirectiveWord(unsigned Size, SMLoc L);
- bool MatchInstruction(SMLoc IDLoc,
- const SmallVectorImpl<MCParsedAsmOperand*> &Operands,
- MCInst &Inst);
+ bool MatchAndEmitInstruction(SMLoc IDLoc,
+ SmallVectorImpl<MCParsedAsmOperand*> &Operands,
+ MCStreamer &Out);
/// @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(
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;
}
};
// 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) {
-
- // The "Jump if rCX Zero" form jcxz is not allowed in 64-bit mode and
- // the form jrcxz is not allowed in 32-bit mode.
- if (Is64Bit) {
- // FIXME: We can do jcxz/jecxz, we just don't have the encoding right yet.
- if (Name == "jcxz" || Name == "jecxz")
- return Error(NameLoc, Name + " cannot be encoded in 64-bit mode");
- }
-
- // 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("push", Is64Bit ? "pushq" : "pushl")
- .Case("pushf", Is64Bit ? "pushfq" : "pushfl")
- .Case("popf", Is64Bit ? "popfq" : "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")
- // FIXME: in 32-bit mode jcxz requires an AdSize prefix. In 64-bit mode
- // jecxz requires an AdSize prefix but jecxz does not have a prefix in
- // 32-bit mode.
- .Case("jecxz", "jcxz")
- .Case("jrcxz", Is64Bit ? "jcxz" : "jrcxz")
- .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")
- .Case("cmovcl", "cmovbl")
- .Case("cmovnal", "cmovbel")
- .Case("cmovnbl", "cmovael")
- .Case("cmovnbel", "cmoval")
- .Case("cmovncl", "cmovael")
- .Case("cmovngl", "cmovlel")
- .Case("cmovnl", "cmovgel")
- .Case("cmovngl", "cmovlel")
- .Case("cmovngel", "cmovll")
- .Case("cmovnll", "cmovgel")
- .Case("cmovnlel", "cmovgl")
- .Case("cmovnzl", "cmovnel")
- .Case("cmovzl", "cmovel")
- .Case("fwait", "wait")
- .Case("movzx", "movzb")
- .Default(Name);
+ StringRef PatchedName = Name;
// FIXME: Hack to recognize cmp<comparison code>{ss,sd,ps,pd}.
const MCExpr *ExtraImmOp = 0;
PatchedName = "vpclmulqdq";
}
}
+
Operands.push_back(X86Operand::CreateToken(PatchedName, NameLoc));
if (ExtraImmOp)
Operands.push_back(X86Operand::CreateImm(ExtraImmOp, NameLoc, NameLoc));
-
- // This does the actual operand parsing.
- if (getLexer().isNot(AsmToken::EndOfStatement)) {
+
+ // Determine whether this is an instruction prefix.
+ bool isPrefix =
+ Name == "lock" || Name == "rep" ||
+ Name == "repe" || Name == "repz" ||
+ Name == "repne" || Name == "repnz";
+
+
+ // 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
+ // the next one.
+ if (getLexer().isNot(AsmToken::EndOfStatement) && !isPrefix) {
// Parse '*' modifier.
if (getLexer().is(AsmToken::Star)) {
// Read the first operand.
if (X86Operand *Op = ParseOperand())
Operands.push_back(Op);
- else
+ else {
+ Parser.EatToEndOfStatement();
return true;
+ }
while (getLexer().is(AsmToken::Comma)) {
Parser.Lex(); // Eat the comma.
// Parse and remember the operand.
if (X86Operand *Op = ParseOperand())
Operands.push_back(Op);
- else
+ else {
+ Parser.EatToEndOfStatement();
return true;
+ }
+ }
+
+ if (getLexer().isNot(AsmToken::EndOfStatement)) {
+ Parser.EatToEndOfStatement();
+ return TokError("unexpected token in argument list");
}
}
- // FIXME: Hack to handle recognizing s{hr,ar,hl}? $1.
+ if (getLexer().is(AsmToken::EndOfStatement))
+ Parser.Lex(); // Consume the EndOfStatement
+
+ // Hack to allow 'movq <largeimm>, <reg>' as an alias for movabsq.
+ if ((Name == "movq" || Name == "mov") && Operands.size() == 3 &&
+ static_cast<X86Operand*>(Operands[2])->isReg() &&
+ static_cast<X86Operand*>(Operands[1])->isImm() &&
+ !static_cast<X86Operand*>(Operands[1])->isImmSExti64i32()) {
+ delete Operands[0];
+ Operands[0] = X86Operand::CreateToken("movabsq", NameLoc);
+ }
+
+ // 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("shl") || Name.startswith("sal")) &&
Operands.size() == 3) {
X86Operand *Op1 = static_cast<X86Operand*>(Operands[1]);
if (Op1->isImm() && isa<MCConstantExpr>(Op1->getImm()) &&
}
}
+ // FIXME: Hack to handle recognize "rc[lr] <op>" -> "rcl $1, <op>".
+ if ((Name.startswith("rcl") || Name.startswith("rcr")) &&
+ Operands.size() == 2) {
+ const MCExpr *One = MCConstantExpr::Create(1, getParser().getContext());
+ Operands.push_back(X86Operand::CreateImm(One, NameLoc, NameLoc));
+ std::swap(Operands[1], Operands[2]);
+ }
+
+ // FIXME: Hack to handle recognize "sh[lr]d op,op" -> "shld $1, op,op".
+ if ((Name.startswith("shld") || Name.startswith("shrd")) &&
+ Operands.size() == 3) {
+ const MCExpr *One = MCConstantExpr::Create(1, getParser().getContext());
+ Operands.insert(Operands.begin()+1,
+ X86Operand::CreateImm(One, NameLoc, NameLoc));
+ }
+
+
+ // 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));
+ }
+
+ // 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]);
+ }
+
+ // FIXME: Hack to handle "out[bwl]? %al, (%dx)" -> "outb %al, %dx".
+ if ((Name == "outb" || Name == "outw" || Name == "outl" || Name == "out") &&
+ Operands.size() == 3) {
+ 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 "f{mul*,add*,sub*,div*} $op, st(0)" the same as
// "f{mul*,add*,sub*,div*} $op"
if ((Name.startswith("fmul") || Name.startswith("fadd") ||
Operands.erase(Operands.begin() + 2);
}
+ // FIXME: Hack to handle "f{mulp,addp} st(0), $op" the same as
+ // "f{mulp,addp} $op", since they commute. We also allow fdivrp/fsubrp even
+ // though they don't commute, solely because gas does support this.
+ if ((Name=="fmulp" || Name=="faddp" || Name=="fsubrp" || Name=="fdivrp") &&
+ Operands.size() == 3 &&
+ static_cast<X86Operand*>(Operands[1])->isReg() &&
+ static_cast<X86Operand*>(Operands[1])->getReg() == X86::ST0) {
+ delete Operands[1];
+ Operands.erase(Operands.begin() + 1);
+ }
+
// FIXME: Hack to handle "imul <imm>, B" which is an alias for "imul <imm>, B,
// B".
if (Name.startswith("imul") && Operands.size() == 3 &&
Operands.push_back(X86Operand::CreateReg(Op->getReg(), Op->getStartLoc(),
Op->getEndLoc()));
}
-
+
// '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())
+ static_cast<X86Operand*>(Operands[1])->isMem()) {
+ delete Operands[0];
Operands[0] = X86Operand::CreateToken("sldtw", NameLoc);
+ }
- return false;
-}
-bool X86ATTAsmParser::ParseDirective(AsmToken DirectiveID) {
- StringRef IDVal = DirectiveID.getIdentifier();
- if (IDVal == ".word")
- return ParseDirectiveWord(2, DirectiveID.getLoc());
- return true;
-}
+ // The assembler accepts these instructions with no operand as a synonym for
+ // an instruction acting on st(1). e.g. "fxch" -> "fxch %st(1)".
+ if ((Name == "fxch" || Name == "fucom" || Name == "fucomp" ||
+ Name == "faddp" || Name == "fsubp" || Name == "fsubrp" ||
+ Name == "fmulp" || Name == "fdivp" || Name == "fdivrp") &&
+ Operands.size() == 1) {
+ Operands.push_back(X86Operand::CreateReg(MatchRegisterName("st(1)"),
+ NameLoc, NameLoc));
+ }
-/// 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;
+ // The assembler accepts these instructions with two few operands as a synonym
+ // for taking %st(1),%st(0) or X, %st(0).
+ if ((Name == "fcomi" || Name == "fucomi" || Name == "fucompi" ||
+ Name == "fcompi" ) &&
+ Operands.size() < 3) {
+ if (Operands.size() == 1)
+ Operands.push_back(X86Operand::CreateReg(MatchRegisterName("st(1)"),
+ NameLoc, NameLoc));
+ Operands.push_back(X86Operand::CreateReg(MatchRegisterName("st(0)"),
+ NameLoc, NameLoc));
+ }
- getParser().getStreamer().EmitValue(Value, Size, 0 /*addrspace*/);
+ // The assembler accepts various amounts of brokenness for fnstsw.
+ if (Name == "fnstsw" || Name == "fnstsww") {
+ if (Operands.size() == 2 &&
+ static_cast<X86Operand*>(Operands[1])->isReg()) {
+ // "fnstsw al" and "fnstsw eax" -> "fnstw"
+ unsigned Reg = static_cast<X86Operand*>(Operands[1])->Reg.RegNo;
+ if (Reg == MatchRegisterName("eax") ||
+ Reg == MatchRegisterName("al")) {
+ delete Operands[1];
+ Operands.pop_back();
+ }
+ }
- if (getLexer().is(AsmToken::EndOfStatement))
- break;
+ // "fnstw" -> "fnstw %ax"
+ if (Operands.size() == 1)
+ Operands.push_back(X86Operand::CreateReg(MatchRegisterName("ax"),
+ NameLoc, NameLoc));
+ }
- // FIXME: Improve diagnostic.
- if (getLexer().isNot(AsmToken::Comma))
- return Error(L, "unexpected token in directive");
- Parser.Lex();
- }
+ // fstp <mem> -> fstps <mem>. Without this, we'll default to fstpl due to
+ // suffix searching.
+ if (Name == "fstp" && Operands.size() == 2 &&
+ static_cast<X86Operand*>(Operands[1])->isMem()) {
+ delete Operands[0];
+ Operands[0] = X86Operand::CreateToken("fstps", NameLoc);
+ }
+
+ // FIXME: Hack to handle recognize "aa[dm]" -> "aa[dm] $0xA".
+ if ((Name.startswith("aad") || Name.startswith("aam")) &&
+ Operands.size() == 1) {
+ const MCExpr *A = MCConstantExpr::Create(0xA, getParser().getContext());
+ Operands.push_back(X86Operand::CreateImm(A, NameLoc, NameLoc));
}
- 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.
+ 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");
// 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;
unsigned NumSuccessfulMatches =
(MatchB == Match_Success) + (MatchW == Match_Success) +
(MatchL == Match_Success) + (MatchQ == Match_Success);
- if (NumSuccessfulMatches == 1)
+ if (NumSuccessfulMatches == 1) {
+ Out.EmitInstruction(Inst);
return false;
+ }
// Otherwise, the match failed, try to produce a decent error message.
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 (!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) +
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) +
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.
projects / oota-llvm.git / blobdiff