[Xtensa 8/10] Add support of the Xtensa shift/load/store/move and processor control instructions.

[andreisfr]

Summary

Add new subset of Core Instructions (not full yet). Add appropriate operands description,
modify asm parser, printer and code emitter. Modify tests to support new instructions.

[arsenm]

llvm/lib/Target/Xtensa/AsmParser/XtensaAsmParser.cpp

139–141

There are already various forms of isInt/isUInt in MathExtras.h

[andreisfr]

Register names are capitalized.

[andreisfr]

llvm/lib/Target/Xtensa/AsmParser/XtensaAsmParser.cpp

139–141

Did you mean do not create/use special functions isImm and inRange? Initially I used the same approach as in SystemZ and AArch64 architecture, because we need to check whether expression immediate and also get value of it.

[andreisfr]

Patch is updated according to latest upstream version.

[andreisfr]

Patch is updated according to LLVM upstream version and latest Xtensa backend version.

[andreisfr]

Correct instruction descriptions, format descriptions and instruction operands according to common style for *.td files. The llvm_unreachable is substituted to report_fatal_error.

[andreisfr]

Remove redundant whitespaces

[andreisfr]

Patch is updated according to LLVM upstream version.

[andreisfr]

Patch is updated according to LLVM upstream version

[andreisfr]

Minor corrections, changed argument names in the XtensaInstPrinter::printMemOperand function.

[andreisfr]

Update according to changes in MEMW, EXTW, DSYNC, ESYNC, ISYNC and RSYNC instruction descriptions

[saugustine]

This is mostly mechanical, and the tests look correct to me. So ready.

[andreisfr]

Minor source code corrections. Test file xtensa-valid.s is splitted to several files.

[barannikov88]

llvm/lib/Target/Xtensa/AsmParser/XtensaAsmParser.cpp

156

use cast<> here and below. dyn_cast is for the cases when the result is allowed to be nullptr.

411

Bug: std::string termporary is destroyed at ';' and RegName references freed memory.

[andreisfr]

Added fixes according to comments

[andreisfr]

llvm/lib/Target/Xtensa/AsmParser/XtensaAsmParser.cpp

156

Fixed

411

Thank you very much, fixed this issue.

[barannikov88]

llvm/lib/Target/Xtensa/AsmParser/XtensaAsmParser.cpp

411

The issue is still there, RegName references deleted memory. One of the possible way to resolve this issue is to declare RegName as std::string rather than StringRef.

BTW, what does this code do? I couldn't find any registers that contain only digits in their names (all start with a letter), so MatchRegisterName is bound to fail. Nor could I find a test that uses parenthesized numbers.

Are those some special registers not modeled in MC layer (and represented as numbers in assembly language)?
If so, you need to differentiate them from ordinary registers somehow because there is no guarantee that their numbers and / or encodings do not collide with the autogenerated ones for the registers that are being modeled.

As a side node, 'SR' variable should have a more descriptive name (or, at least, its name should be explained in a comment).

[andreisfr]

llvm/lib/Target/Xtensa/AsmParser/XtensaAsmParser.cpp

411

Thank you for comment. The "RegName" variable is not used outside switch cases, so StringRef object should be alive at points where "RegName" actually used, is it correct? You noted correctly that we should have special registers defined by numbers and Xtensa architecture have such registers for example https://github.com/llvm/llvm-project/blob/ff25800d4ba0b577a44dc918da7a1fb3c29fdb13/llvm/lib/Target/Xtensa/XtensaRegisterInfo.td#L76 . But test for such kind of register names is absent and will be implemented in nearest patches and we will add more explanations to this function.

[barannikov88]

llvm/lib/Target/Xtensa/AsmParser/XtensaAsmParser.cpp

411

RegName is alive, but the data it references is not.
StringRef does not own the data.
The data is owned by std::string, which is destroyed before the first use of RegName.

[andreisfr]

llvm/lib/Target/Xtensa/AsmParser/XtensaAsmParser.cpp

411

I will rewrite this code in next patches

llvm/lib/Target/Xtensa/AsmParser/XtensaAsmParser.cpp

 // Auto-generated instruction matching functions
 #define GET_ASSEMBLER_HEADER
 #include "XtensaGenAsmMatcher.inc"
 
   OperandMatchResultTy parseImmediate(OperandVector &Operands);
   OperandMatchResultTy parseRegister(OperandVector &Operands,
                                      bool AllowParens = false, bool SR = false);
   OperandMatchResultTy parseOperandWithModifier(OperandVector &Operands);
-  bool parseOperand(OperandVector &Operands);
+  bool parseOperand(OperandVector &Operands, StringRef Mnemonic,
+                    bool SR = false);
+  bool ParseInstructionWithSR(ParseInstructionInfo &Info, StringRef Name,
+                              SMLoc NameLoc, OperandVector &Operands);
   OperandMatchResultTy tryParseRegister(MCRegister &RegNo, SMLoc &StartLoc,
                                         SMLoc &EndLoc) override {
     return MatchOperand_NoMatch;
   }
 
 public:
   enum XtensaMatchResultTy {
     Match_Dummy = FIRST_TARGET_MATCH_RESULT_TY,
XtensaOperand(const XtensaOperand &o) : MCParsedAsmOperand() {
     }
   }
 
   bool isToken() const override { return Kind == Token; }
   bool isReg() const override { return Kind == Register; }
   bool isImm() const override { return Kind == Immediate; }
   bool isMem() const override { return false; }
 
   bool isImm(int64_t MinValue, int64_t MaxValue) const {
     return Kind == Immediate && inRange(getImm(), MinValue, MaxValue);
   }
 
   bool isImm8() const { return isImm(-128, 127); }
 
   bool isImm8_sh8() const {
     return isImm(-32768, 32512) &&
-           ((dyn_cast<MCConstantExpr>(getImm())->getValue() & 0xFF) == 0);
+           ((cast<MCConstantExpr>(getImm())->getValue() & 0xFF) == 0);
   }
 
+  bool isImm12() const { return isImm(-2048, 2047); }
+
+  bool isImm12m() const { return isImm(-2048, 2047); }
+
+  bool isOffset4m32() const {
+    return isImm(0, 60) &&
+           ((cast<MCConstantExpr>(getImm())->getValue() & 0x3) == 0);
+  }
+
+  bool isOffset8m8() const { return isImm(0, 255); }
+
+  bool isOffset8m16() const {
+    return isImm(0, 510) &&
+           ((cast<MCConstantExpr>(getImm())->getValue() & 0x1) == 0);
+  }
+
+  bool isOffset8m32() const {
+    return isImm(0, 1020) &&
+           ((cast<MCConstantExpr>(getImm())->getValue() & 0x3) == 0);
+  }
+
+  bool isUimm4() const { return isImm(0, 15); }
+
+  bool isUimm5() const { return isImm(0, 31); }
+
+  bool isImm8n_7() const { return isImm(-8, 7); }
+
+  bool isShimm1_31() const { return isImm(1, 31); }
+
+  bool isImm16_31() const { return isImm(16, 31); }
+
+  bool isImm1_16() const { return isImm(1, 16); }
+
   /// getStartLoc - Gets location of the first token of this operand
   SMLoc getStartLoc() const override { return StartLoc; }
   /// getEndLoc - Gets location of the last token of this operand
   SMLoc getEndLoc() const override { return EndLoc; }
 
   unsigned getReg() const override {
     assert(Kind == Register && "Invalid type access!");
     return Reg.RegNum;
bool XtensaAsmParser::MatchAndEmitInstruction(SMLoc IDLoc, unsigned &Opcode,
   }
   case Match_InvalidImm8:
     return Error(RefineErrorLoc(IDLoc, Operands, ErrorInfo),
                  "expected immediate in range [-128, 127]");
   case Match_InvalidImm8_sh8:
     return Error(RefineErrorLoc(IDLoc, Operands, ErrorInfo),
                  "expected immediate in range [-32768, 32512], first 8 bits "
                  "should be zero");
+  case Match_InvalidImm12:
+    return Error(RefineErrorLoc(IDLoc, Operands, ErrorInfo),
+                 "expected immediate in range [-2048, 2047]");
+  case Match_InvalidImm12m:
+    return Error(RefineErrorLoc(IDLoc, Operands, ErrorInfo),
+                 "expected immediate in range [-2048, 2047]");
+  case Match_InvalidImm1_16:
+    return Error(RefineErrorLoc(IDLoc, Operands, ErrorInfo),
+                 "expected immediate in range [1, 16]");
+  case Match_InvalidShimm1_31:
+    return Error(RefineErrorLoc(IDLoc, Operands, ErrorInfo),
+                 "expected immediate in range [1, 31]");
+  case Match_InvalidUimm4:
+    return Error(RefineErrorLoc(IDLoc, Operands, ErrorInfo),
+                 "expected immediate in range [0, 15]");
+  case Match_InvalidUimm5:
+    return Error(RefineErrorLoc(IDLoc, Operands, ErrorInfo),
+                 "expected immediate in range [0, 31]");
+  case Match_InvalidOffset8m8:
+    return Error(RefineErrorLoc(IDLoc, Operands, ErrorInfo),
+                 "expected immediate in range [0, 255]");
+  case Match_InvalidOffset8m16:
+    return Error(RefineErrorLoc(IDLoc, Operands, ErrorInfo),
+                 "expected immediate in range [0, 510], first bit "
+                 "should be zero");
+  case Match_InvalidOffset8m32:
+    return Error(RefineErrorLoc(IDLoc, Operands, ErrorInfo),
+                 "expected immediate in range [0, 1020], first 2 bits "
+                 "should be zero");
+  case Match_InvalidOffset4m32:
+    return Error(RefineErrorLoc(IDLoc, Operands, ErrorInfo),
+                 "expected immediate in range [0, 60], first 2 bits "
+                 "should be zero");
   }
 
   report_fatal_error("Unknown match type detected!");
 }
 
 bool XtensaAsmParser::parseRegister(MCRegister &RegNo, SMLoc &StartLoc,
                                     SMLoc &EndLoc) {
   const AsmToken &Tok = getParser().getTok();
OperandMatchResultTy XtensaAsmParser::parseRegister(OperandVector &Operands,
   bool HadParens = false;
   AsmToken Buf[2];
   StringRef RegName;
 
   // If this a parenthesised register name is allowed, parse it atomically
   if (AllowParens && getLexer().is(AsmToken::LParen)) {
     size_t ReadCount = getLexer().peekTokens(Buf);
     if (ReadCount == 2 && Buf[1].getKind() == AsmToken::RParen) {
+      if ((Buf[0].getKind() == AsmToken::Integer) && (!SR))
+        return MatchOperand_NoMatch;
       HadParens = true;
       getParser().Lex(); // Eat '('
     }
   }
 
+  unsigned RegNo = 0;
+
   switch (getLexer().getKind()) {
   default:
     return MatchOperand_NoMatch;
+  case AsmToken::Integer:
+    if (!SR)
+      return MatchOperand_NoMatch;
+    RegName = StringRef(std::to_string(getLexer().getTok().getIntVal()));
+    RegNo = MatchRegisterName(RegName);
+    if (RegNo == 0)
+      RegNo = MatchRegisterAltName(RegName);
+    break;
   case AsmToken::Identifier:
     RegName = getLexer().getTok().getIdentifier();
-    break;
-  }
-
-  unsigned RegNo = MatchRegisterName(RegName);
+    RegNo = MatchRegisterName(RegName);
     if (RegNo == 0)
       RegNo = MatchRegisterAltName(RegName);
+    break;
+  }
 
   if (RegNo == 0) {
     if (HadParens)
       getLexer().UnLex(Buf[0]);
     return MatchOperand_NoMatch;
   }
   if (HadParens)
     Operands.push_back(XtensaOperand::createToken("(", FirstS));

 OperandMatchResultTy
 XtensaAsmParser::parseOperandWithModifier(OperandVector &Operands) {
   return MatchOperand_ParseFail;
 }
 
 /// Looks at a token type and creates the relevant operand
 /// from this information, adding to Operands.
 /// If operand was parsed, returns false, else true.
-bool XtensaAsmParser::parseOperand(OperandVector &Operands) {
+bool XtensaAsmParser::parseOperand(OperandVector &Operands, StringRef Mnemonic,
+                                   bool SR) {
   // Attempt to parse token as register
-  if (parseRegister(Operands, true) == MatchOperand_Success)
+  if (parseRegister(Operands, true, SR) == MatchOperand_Success)
     return false;
 
   // Attempt to parse token as an immediate
   if (parseImmediate(Operands) == MatchOperand_Success) {
     return false;
   }
 
   // Finally we have exhausted all options and must declare defeat.
   Error(getLoc(), "unknown operand");
   return true;
 }
 
+bool XtensaAsmParser::ParseInstructionWithSR(ParseInstructionInfo &Info,
+                                             StringRef Name, SMLoc NameLoc,
+                                             OperandVector &Operands) {
+  if ((Name.startswith("wsr.") || Name.startswith("rsr.") ||
+       Name.startswith("xsr.")) &&
+      (Name.size() > 4)) {
+    // Parse case when instruction name is concatenated with SR register
+    // name, like "wsr.sar a1"
+
+    // First operand is token for instruction
+    Operands.push_back(XtensaOperand::createToken(Name.take_front(3), NameLoc));
+
+    StringRef RegName = Name.drop_front(4);
+    unsigned RegNo = MatchRegisterName(RegName);
+
+    if (RegNo == 0)
+      RegNo = MatchRegisterAltName(RegName);
+
+    if (RegNo == 0) {
+      Error(NameLoc, "invalid register name");
+      return true;
+    }
+
+    // Parse operand
+    if (parseOperand(Operands, Name))
+      return true;
+
+    SMLoc S = getLoc();
+    SMLoc E = SMLoc::getFromPointer(S.getPointer() - 1);
+    Operands.push_back(XtensaOperand::createReg(RegNo, S, E));
+  } else {
+    // First operand is token for instruction
+    Operands.push_back(XtensaOperand::createToken(Name, NameLoc));
+
+    // Parse first operand
+    if (parseOperand(Operands, Name))
+      return true;
+
+    if (!getLexer().is(AsmToken::Comma)) {
+      SMLoc Loc = getLexer().getLoc();
+      getParser().eatToEndOfStatement();
+      return Error(Loc, "unexpected token");
+    }
+
+    getLexer().Lex();
+
+    // Parse second operand
+    if (parseOperand(Operands, Name, true))
+      return true;
+  }
+
+  if (getLexer().isNot(AsmToken::EndOfStatement)) {
+    SMLoc Loc = getLexer().getLoc();
+    getParser().eatToEndOfStatement();
+    return Error(Loc, "unexpected token");
+  }
+
+  getParser().Lex(); // Consume the EndOfStatement.
+  return false;
+}
+
 bool XtensaAsmParser::ParseInstruction(ParseInstructionInfo &Info,
                                        StringRef Name, SMLoc NameLoc,
                                        OperandVector &Operands) {
+  if (Name.startswith("wsr") || Name.startswith("rsr") ||
+      Name.startswith("xsr")) {
+    return ParseInstructionWithSR(Info, Name, NameLoc, Operands);
+  }
+
   // First operand is token for instruction
   Operands.push_back(XtensaOperand::createToken(Name, NameLoc));
 
   // If there are no more operands, then finish
   if (getLexer().is(AsmToken::EndOfStatement))
     return false;
 
   // Parse first operand
-  if (parseOperand(Operands))
+  if (parseOperand(Operands, Name))
     return true;
 
   // Parse until end of statement, consuming commas between operands
   while (getLexer().is(AsmToken::Comma)) {
     // Consume comma token
     getLexer().Lex();
 
     // Parse next operand
-    if (parseOperand(Operands))
+    if (parseOperand(Operands, Name))
       return true;
   }
 
   if (getLexer().isNot(AsmToken::EndOfStatement)) {
     SMLoc Loc = getLexer().getLoc();
     getParser().eatToEndOfStatement();
     return Error(Loc, "unexpected token");
   }

llvm/lib/Target/Xtensa/MCTargetDesc/XtensaInstPrinter.h

public:
   // Override MCInstPrinter.
   void printRegName(raw_ostream &O, unsigned RegNo) const override;
   void printInst(const MCInst *MI, uint64_t Address, StringRef Annot,
                  const MCSubtargetInfo &STI, raw_ostream &O) override;
 
 private:
   // Print various types of operand.
   void printOperand(const MCInst *MI, int OpNum, raw_ostream &O);
+  void printMemOperand(const MCInst *MI, int OpNUm, raw_ostream &O);
 
   void printImm8_AsmOperand(const MCInst *MI, int OpNum, raw_ostream &O);
   void printImm8_sh8_AsmOperand(const MCInst *MI, int OpNum, raw_ostream &O);
+  void printImm12_AsmOperand(const MCInst *MI, int OpNum, raw_ostream &O);
+  void printImm12m_AsmOperand(const MCInst *MI, int OpNum, raw_ostream &O);
+  void printUimm4_AsmOperand(const MCInst *MI, int OpNum, raw_ostream &O);
+  void printUimm5_AsmOperand(const MCInst *MI, int OpNum, raw_ostream &O);
+  void printShimm1_31_AsmOperand(const MCInst *MI, int OpNum, raw_ostream &O);
+  void printImm1_16_AsmOperand(const MCInst *MI, int OpNum, raw_ostream &O);
+  void printOffset8m8_AsmOperand(const MCInst *MI, int OpNum, raw_ostream &O);
+  void printOffset8m16_AsmOperand(const MCInst *MI, int OpNum, raw_ostream &O);
+  void printOffset8m32_AsmOperand(const MCInst *MI, int OpNum, raw_ostream &O);
+  void printOffset4m32_AsmOperand(const MCInst *MI, int OpNum, raw_ostream &O);
 };
 } // end namespace llvm
 
 #endif /* LLVM_LIB_TARGET_XTENSA_MCTARGETDESC_XTENSAINSTPRINTER_H */

llvm/lib/Target/Xtensa/MCTargetDesc/XtensaInstPrinter.cpp

void XtensaInstPrinter::printRegName(raw_ostream &O, unsigned RegNo) const {
   O << getRegisterName(RegNo);
 }
 
 void XtensaInstPrinter::printOperand(const MCInst *MI, int OpNum,
                                      raw_ostream &O) {
   printOperand(MI->getOperand(OpNum), O);
 }
 
+void XtensaInstPrinter::printMemOperand(const MCInst *MI, int OpNum,
+                                        raw_ostream &OS) {
+  OS << getRegisterName(MI->getOperand(OpNum).getReg());
+  OS << ", ";
+  printOperand(MI, OpNum + 1, OS);
+}
+
 void XtensaInstPrinter::printImm8_AsmOperand(const MCInst *MI, int OpNum,
                                              raw_ostream &O) {
   if (MI->getOperand(OpNum).isImm()) {
     int64_t Value = MI->getOperand(OpNum).getImm();
     assert(isInt<8>(Value) &&
            "Invalid argument, value must be in ranges [-128,127]");
     O << Value;
   } else {
     printOperand(MI, OpNum, O);
   }
 }
 
 void XtensaInstPrinter::printImm8_sh8_AsmOperand(const MCInst *MI, int OpNum,
                                                  raw_ostream &O) {
   if (MI->getOperand(OpNum).isImm()) {
     int64_t Value = MI->getOperand(OpNum).getImm();
     assert((isInt<16>(Value) && ((Value & 0xFF) == 0)) &&
            "Invalid argument, value must be multiples of 256 in range "
            "[-32768,32512]");
     O << Value;
   } else
     printOperand(MI, OpNum, O);
 }
+
+void XtensaInstPrinter::printImm12_AsmOperand(const MCInst *MI, int OpNum,
+                                              raw_ostream &O) {
+  if (MI->getOperand(OpNum).isImm()) {
+    int64_t Value = MI->getOperand(OpNum).getImm();
+    assert((Value >= -2048 && Value <= 2047) &&
+           "Invalid argument, value must be in ranges [-2048,2047]");
+    O << Value;
+  } else
+    printOperand(MI, OpNum, O);
+}
+
+void XtensaInstPrinter::printImm12m_AsmOperand(const MCInst *MI, int OpNum,
+                                               raw_ostream &O) {
+  if (MI->getOperand(OpNum).isImm()) {
+    int64_t Value = MI->getOperand(OpNum).getImm();
+    assert((Value >= -2048 && Value <= 2047) &&
+           "Invalid argument, value must be in ranges [-2048,2047]");
+    O << Value;
+  } else
+    printOperand(MI, OpNum, O);
+}
+
+void XtensaInstPrinter::printUimm4_AsmOperand(const MCInst *MI, int OpNum,
+                                              raw_ostream &O) {
+  if (MI->getOperand(OpNum).isImm()) {
+    int64_t Value = MI->getOperand(OpNum).getImm();
+    assert((Value >= 0 && Value <= 15) && "Invalid argument");
+    O << Value;
+  } else
+    printOperand(MI, OpNum, O);
+}
+
+void XtensaInstPrinter::printUimm5_AsmOperand(const MCInst *MI, int OpNum,
+                                              raw_ostream &O) {
+  if (MI->getOperand(OpNum).isImm()) {
+    int64_t Value = MI->getOperand(OpNum).getImm();
+    assert((Value >= 0 && Value <= 31) && "Invalid argument");
+    O << Value;
+  } else
+    printOperand(MI, OpNum, O);
+}
+
+void XtensaInstPrinter::printShimm1_31_AsmOperand(const MCInst *MI, int OpNum,
+                                                  raw_ostream &O) {
+  if (MI->getOperand(OpNum).isImm()) {
+    int64_t Value = MI->getOperand(OpNum).getImm();
+    assert((Value >= 1 && Value <= 31) &&
+           "Invalid argument, value must be in range [1,31]");
+    O << Value;
+  } else
+    printOperand(MI, OpNum, O);
+}
+
+void XtensaInstPrinter::printImm1_16_AsmOperand(const MCInst *MI, int OpNum,
+                                                raw_ostream &O) {
+  if (MI->getOperand(OpNum).isImm()) {
+    int64_t Value = MI->getOperand(OpNum).getImm();
+    assert((Value >= 1 && Value <= 16) &&
+           "Invalid argument, value must be in range [1,16]");
+    O << Value;
+  } else
+    printOperand(MI, OpNum, O);
+}
+
+void XtensaInstPrinter::printOffset8m8_AsmOperand(const MCInst *MI, int OpNum,
+                                                  raw_ostream &O) {
+  if (MI->getOperand(OpNum).isImm()) {
+    int64_t Value = MI->getOperand(OpNum).getImm();
+    assert((Value >= 0 && Value <= 255) &&
+           "Invalid argument, value must be in range [0,255]");
+    O << Value;
+  } else
+    printOperand(MI, OpNum, O);
+}
+
+void XtensaInstPrinter::printOffset8m16_AsmOperand(const MCInst *MI, int OpNum,
+                                                   raw_ostream &O) {
+  if (MI->getOperand(OpNum).isImm()) {
+    int64_t Value = MI->getOperand(OpNum).getImm();
+    assert((Value >= 0 && Value <= 510 && ((Value & 0x1) == 0)) &&
+           "Invalid argument, value must be multiples of two in range [0,510]");
+    O << Value;
+  } else
+    printOperand(MI, OpNum, O);
+}
+
+void XtensaInstPrinter::printOffset8m32_AsmOperand(const MCInst *MI, int OpNum,
+                                                   raw_ostream &O) {
+  if (MI->getOperand(OpNum).isImm()) {
+    int64_t Value = MI->getOperand(OpNum).getImm();
+    assert(
+        (Value >= 0 && Value <= 1020 && ((Value & 0x3) == 0)) &&
+        "Invalid argument, value must be multiples of four in range [0,1020]");
+    O << Value;
+  } else
+    printOperand(MI, OpNum, O);
+}
+
+void XtensaInstPrinter::printOffset4m32_AsmOperand(const MCInst *MI, int OpNum,
+                                                   raw_ostream &O) {
+  if (MI->getOperand(OpNum).isImm()) {
+    int64_t Value = MI->getOperand(OpNum).getImm();
+    assert((Value >= 0 && Value <= 60 && ((Value & 0x3) == 0)) &&
+           "Invalid argument, value must be multiples of four in range [0,60]");
+    O << Value;
+  } else
+    printOperand(MI, OpNum, O);
+}

llvm/lib/Target/Xtensa/MCTargetDesc/XtensaMCCodeEmitter.cpp

uint64_t getBinaryCodeForInstr(const MCInst &MI,
                                  const MCSubtargetInfo &STI) const;
 
   // Called by the TableGen code to get the binary encoding of operand
   // MO in MI.  Fixups is the list of fixups against MI.
   uint32_t getMachineOpValue(const MCInst &MI, const MCOperand &MO,
                              SmallVectorImpl<MCFixup> &Fixups,
                              const MCSubtargetInfo &STI) const;
 
+  uint32_t getMemRegEncoding(const MCInst &MI, unsigned OpNo,
+                             SmallVectorImpl<MCFixup> &Fixups,
+                             const MCSubtargetInfo &STI) const;
+
   uint32_t getImm8OpValue(const MCInst &MI, unsigned OpNo,
                           SmallVectorImpl<MCFixup> &Fixups,
                           const MCSubtargetInfo &STI) const;
 
   uint32_t getImm8_sh8OpValue(const MCInst &MI, unsigned OpNo,
                               SmallVectorImpl<MCFixup> &Fixups,
                               const MCSubtargetInfo &STI) const;
+
+  uint32_t getImm12OpValue(const MCInst &MI, unsigned OpNo,
+                           SmallVectorImpl<MCFixup> &Fixups,
+                           const MCSubtargetInfo &STI) const;
+
+  uint32_t getUimm4OpValue(const MCInst &MI, unsigned OpNo,
+                           SmallVectorImpl<MCFixup> &Fixups,
+                           const MCSubtargetInfo &STI) const;
+
+  uint32_t getUimm5OpValue(const MCInst &MI, unsigned OpNo,
+                           SmallVectorImpl<MCFixup> &Fixups,
+                           const MCSubtargetInfo &STI) const;
+
+  uint32_t getImm1_16OpValue(const MCInst &MI, unsigned OpNo,
+                             SmallVectorImpl<MCFixup> &Fixups,
+                             const MCSubtargetInfo &STI) const;
+
+  uint32_t getShimm1_31OpValue(const MCInst &MI, unsigned OpNo,
+                               SmallVectorImpl<MCFixup> &Fixups,
+                               const MCSubtargetInfo &STI) const;
 };
 } // namespace
 
 MCCodeEmitter *llvm::createXtensaMCCodeEmitter(const MCInstrInfo &MCII,
                                                MCContext &Ctx) {
   return new XtensaMCCodeEmitter(MCII, Ctx, true);
 }
 

 
 uint32_t
 XtensaMCCodeEmitter::getMachineOpValue(const MCInst &MI, const MCOperand &MO,
                                        SmallVectorImpl<MCFixup> &Fixups,
                                        const MCSubtargetInfo &STI) const {
   if (MO.isReg())
     return Ctx.getRegisterInfo()->getEncodingValue(MO.getReg());
   if (MO.isImm()) {
-    uint32_t res = static_cast<uint32_t>(MO.getImm());
-    return res;
+    uint32_t Res = static_cast<uint32_t>(MO.getImm());
+    return Res;
   }
 
   report_fatal_error("Unhandled expression!");
   return 0;
 }
 
+uint32_t
+XtensaMCCodeEmitter::getMemRegEncoding(const MCInst &MI, unsigned OpNo,
+                                       SmallVectorImpl<MCFixup> &Fixups,
+                                       const MCSubtargetInfo &STI) const {
+  assert(MI.getOperand(OpNo + 1).isImm());
+
+  uint32_t Res = static_cast<uint32_t>(MI.getOperand(OpNo + 1).getImm());
+
+  switch (MI.getOpcode()) {
+  case Xtensa::S16I:
+  case Xtensa::L16SI:
+  case Xtensa::L16UI:
+    if (Res & 0x1) {
+      report_fatal_error("Unexpected operand value!");
+    }
+    Res >>= 1;
+    break;
+  case Xtensa::S32I:
+  case Xtensa::L32I:
+    if (Res & 0x3) {
+      report_fatal_error("Unexpected operand value!");
+    }
+    Res >>= 2;
+    break;
+  }
+  
+  assert((isUInt<8>(Res)) && "Unexpected operand value!");
+
+  uint32_t OffBits = Res << 4;
+  uint32_t RegBits = getMachineOpValue(MI, MI.getOperand(OpNo), Fixups, STI);
+
+  return ((OffBits & 0xFF0) | RegBits);
+}
+
 uint32_t XtensaMCCodeEmitter::getImm8OpValue(const MCInst &MI, unsigned OpNo,
                                              SmallVectorImpl<MCFixup> &Fixups,
                                              const MCSubtargetInfo &STI) const {
   const MCOperand &MO = MI.getOperand(OpNo);
   int32_t Res = MO.getImm();
 
   assert(((Res >= -128) && (Res <= 127)) && "Unexpected operand value!");
 
   return (Res & 0xff);
 }
 
 uint32_t
 XtensaMCCodeEmitter::getImm8_sh8OpValue(const MCInst &MI, unsigned OpNo,
                                         SmallVectorImpl<MCFixup> &Fixups,
                                         const MCSubtargetInfo &STI) const {
   const MCOperand &MO = MI.getOperand(OpNo);
   int32_t Res = MO.getImm();
 
   assert(((Res >= -32768) && (Res <= 32512) && ((Res & 0xff) == 0)) &&
          "Unexpected operand value!");
 
   return (Res & 0xffff);
 }
 
+uint32_t
+XtensaMCCodeEmitter::getImm12OpValue(const MCInst &MI, unsigned OpNo,
+                                     SmallVectorImpl<MCFixup> &Fixups,
+                                     const MCSubtargetInfo &STI) const {
+  const MCOperand &MO = MI.getOperand(OpNo);
+  int32_t Res = MO.getImm();
+
+  assert(((Res >= -2048) && (Res <= 2047)) && "Unexpected operand value!");
+
+  return (Res & 0xfff);
+}
+
+uint32_t
+XtensaMCCodeEmitter::getUimm4OpValue(const MCInst &MI, unsigned OpNo,
+                                     SmallVectorImpl<MCFixup> &Fixups,
+                                     const MCSubtargetInfo &STI) const {
+  const MCOperand &MO = MI.getOperand(OpNo);
+  uint32_t Res = static_cast<uint32_t>(MO.getImm());
+
+  assert((Res <= 15) && "Unexpected operand value!");
+
+  return Res & 0xf;
+}
+
+uint32_t
+XtensaMCCodeEmitter::getUimm5OpValue(const MCInst &MI, unsigned OpNo,
+                                     SmallVectorImpl<MCFixup> &Fixups,
+                                     const MCSubtargetInfo &STI) const {
+  const MCOperand &MO = MI.getOperand(OpNo);
+  uint32_t Res = static_cast<uint32_t>(MO.getImm());
+
+  assert((Res <= 31) && "Unexpected operand value!");
+
+  return (Res & 0x1f);
+}
+
+uint32_t
+XtensaMCCodeEmitter::getShimm1_31OpValue(const MCInst &MI, unsigned OpNo,
+                                         SmallVectorImpl<MCFixup> &Fixups,
+                                         const MCSubtargetInfo &STI) const {
+  const MCOperand &MO = MI.getOperand(OpNo);
+  uint32_t Res = static_cast<uint32_t>(MO.getImm());
+
+  assert(((Res >= 1) && (Res <= 31)) && "Unexpected operand value!");
+
+  return ((32 - Res) & 0x1f);
+}
+
+uint32_t
+XtensaMCCodeEmitter::getImm1_16OpValue(const MCInst &MI, unsigned OpNo,
+                                       SmallVectorImpl<MCFixup> &Fixups,
+                                       const MCSubtargetInfo &STI) const {
+  const MCOperand &MO = MI.getOperand(OpNo);
+  uint32_t Res = static_cast<uint32_t>(MO.getImm());
+
+  assert(((Res >= 1) && (Res <= 16)) && "Unexpected operand value!");
+
+  return (Res - 1);
+}
 #include "XtensaGenMCCodeEmitter.inc"

llvm/lib/Target/Xtensa/XtensaInstrInfo.td

def NEG : RRR_Inst<0x00, 0x00, 0x06, (outs AR:$r), (ins AR:$t),
                   "neg\t$r, $t",
                   [(set AR:$r, (ineg AR:$t))]> {
   let s = 0x00;
 }
 
 //===----------------------------------------------------------------------===//
 // Move instructions
 //===----------------------------------------------------------------------===//
+def MOVI : RRI8_Inst<0x02, (outs AR:$t), (ins imm12m:$imm),
+                    "movi\t$t, $imm",
+                    [(set AR:$t, imm12m:$imm)]> {
+  bits<12> imm;
+
+  let imm8{7-0} = imm{7-0};
+  let s{3-0} = imm{11-8};
+  let r = 0xa;
+}
 
-// Conditional move
 def MOVEQZ : RRR_Inst<0x00, 0x03, 0x08, (outs AR:$r), (ins AR:$s, AR:$t),
                      "moveqz\t$r, $s, $t", []>;
 def MOVNEZ : RRR_Inst<0x00, 0x03, 0x09, (outs AR:$r), (ins AR:$s, AR:$t),
                      "movnez\t$r, $s, $t", []>;
 def MOVLTZ : RRR_Inst<0x00, 0x03, 0x0A, (outs AR:$r), (ins AR:$s, AR:$t),
                      "movltz\t$r, $s, $t", []>;
 def MOVGEZ : RRR_Inst<0x00, 0x03, 0x0B, (outs AR:$r), (ins AR:$s, AR:$t),
                      "movgez\t$r, $s, $t", []>;
 
 //===----------------------------------------------------------------------===//
+// Shift instructions
+//===----------------------------------------------------------------------===//
+
+let Uses = [SAR] in {
+  def SLL : RRR_Inst<0x00, 0x01, 0x0A, (outs AR:$r), (ins AR:$s),
+                    "sll\t$r, $s", []> {
+    let t = 0x00;
+  }
+
+  def SRA : RRR_Inst<0x00, 0x01, 0x0B, (outs AR:$r), (ins AR:$t),
+                    "sra\t$r, $t", []> {
+    let s = 0x00;
+  }
+
+  def SRC : RRR_Inst<0x00, 0x01, 0x08, (outs AR:$r), (ins AR:$s, AR:$t),
+                    "src\t$r, $s, $t", []>;
+
+  def SRL : RRR_Inst<0x00, 0x01, 0x09, (outs AR:$r), (ins AR:$t),
+                    "srl\t$r, $t", []> {
+    let s = 0x00;
+  }
+}
+
+let Defs = [SAR] in {
+  def SSL : RRR_Inst<0x00, 0x00, 0x04, (outs), (ins AR:$s),
+                    "ssl\t$s", []> {
+    let r = 0x01;
+    let t = 0x00;
+  }
+
+  def SSR : RRR_Inst<0x00, 0x00, 0x04, (outs), (ins AR:$s),
+                    "ssr\t$s", []> {
+    let r = 0x00;
+    let t = 0x00;
+  }
+}
+
+def EXTUI : RRR_Inst<0x00, 0x04, 0x00, (outs AR:$r), (ins AR:$t, uimm5:$imm1, imm1_16:$imm2),
+                    "extui\t$r, $t, $imm1, $imm2", []> {
+  bits<5> imm1;
+  bits<4> imm2;
+
+  let s = imm1{3-0};
+  let Inst{16} = imm1{4};
+  let Inst{23-20} = imm2;
+}
+
+def SRAI : RRR_Inst<0x00, 0x01, 0x02, (outs AR:$r), (ins AR:$t, uimm5:$sa),
+                   "srai\t$r, $t, $sa",
+                   [(set AR:$r, (sra AR:$t, uimm5:$sa))]> {
+  bits<5> sa;
+
+  let Inst{20} = sa{4};
+  let s = sa{3-0};
+}
+
+def SRLI : RRR_Inst<0x00, 0x01, 0x04, (outs AR:$r), (ins AR:$t, uimm4:$sa),
+                   "srli\t$r, $t, $sa",
+                   [(set AR:$r, (srl AR:$t, uimm4:$sa))]> {
+  bits<4> sa;
+
+  let s = sa;
+}
+
+def SLLI : RRR_Inst<0x00, 0x01, 0x00, (outs AR:$r), (ins AR:$s, shimm1_31:$sa),
+                   "slli\t$r, $s, $sa",
+                   [(set AR:$r, (shl AR:$s, shimm1_31:$sa))]> {
+  bits<5> sa;
+
+  let Inst{20} = sa{4};
+  let t = sa{3-0};
+}
+
+def SSA8L : RRR_Inst<0x00, 0x00, 0x04, (outs), (ins AR:$s),
+                    "ssa8l\t$s", []> {
+  let r = 0x2;
+  let t = 0x0;
+}
+
+def SSAI : RRR_Inst<0x00, 0x00, 0x04, (outs), (ins uimm5:$imm),
+                   "ssai\t$imm", []> {
+  bits<5> imm;
+
+  let r = 0x04;
+  let s = imm{3-0};
+  let t{3-1} = 0;
+  let t{0} = imm{4};
+}
+
+//===----------------------------------------------------------------------===//
+// Load and store instructions
+//===----------------------------------------------------------------------===//
+
+// Load instructions
+let mayLoad = 1 in {
+
+  class Load_RRI8<bits<4> oper, string instrAsm, SDPatternOperator opNode,
+        ComplexPattern addrOp, Operand memOp>
+    : RRI8_Inst<0x02, (outs AR:$t), (ins memOp:$addr),
+                instrAsm#"\t$t, $addr",
+               [(set AR:$t, (opNode addrOp:$addr))]> {
+    bits<12> addr;
+
+    let r = oper;
+    let imm8{7-0} = addr{11-4};
+    let s{3-0} = addr{3-0};
+  }
+}
+
+def L8UI  : Load_RRI8<0x00, "l8ui", zextloadi8, addr_ish1, mem8>;
+def L16SI : Load_RRI8<0x09, "l16si", sextloadi16, addr_ish2, mem16>;
+def L16UI : Load_RRI8<0x01, "l16ui", zextloadi16, addr_ish2, mem16>;
+def L32I  : Load_RRI8<0x02, "l32i", load, addr_ish4, mem32>;
+
+// Store instructions
+let mayStore = 1 in {
+  class Store_II8<bits<4> oper, string instrAsm, SDPatternOperator opNode,
+        ComplexPattern addrOp, Operand memOp>
+    : RRI8_Inst<0x02, (outs), (ins AR:$t, memOp:$addr),
+                instrAsm#"\t$t, $addr",
+               [(opNode AR:$t, addrOp:$addr)]> {
+    bits<12> addr;
+
+    let r = oper;
+    let imm8{7-0} = addr{11-4};
+    let s{3-0} = addr{3-0};
+  }
+}
+
+def S8I  : Store_II8<0x04, "s8i", truncstorei8, addr_ish1, mem8>;
+def S16I : Store_II8<0x05, "s16i", truncstorei16, addr_ish2, mem16>;
+def S32I : Store_II8<0x06, "s32i", store, addr_ish4, mem32>;
+
+//===----------------------------------------------------------------------===//
 // Mem barrier instructions
 //===----------------------------------------------------------------------===//
+
 def MEMW :  RRR_Inst<0x00, 0x00, 0x00, (outs), (ins),
                     "memw", []> {
   let r = 0x2;
   let t = 0x0c;
   let s = 0x0;
 }
 
 def EXTW : RRR_Inst<0x00, 0x00, 0x00, (outs), (ins),

 }
 
 def NOP : RRR_Inst<0x00, 0x00, 0x00, (outs), (ins),
                   "nop", []> {
   let r = 0x02;
   let s = 0x00;
   let t = 0x0f;
 }
+
+def WSR : RSR_Inst<0x00, 0x03, 0x01, (outs SR:$sr), (ins AR:$t),
+                  "wsr\t$t, $sr", []>;
+
+def RSR : RSR_Inst<0x00, 0x03, 0x00, (outs AR:$t), (ins SR:$sr),
+                  "rsr\t$t, $sr", []>;
+
+def XSR : RSR_Inst<0x00, 0x01, 0x06, (outs AR:$ard, SR:$srd), (ins AR:$t, SR:$sr),
+                  "xsr\t$t, $sr", []> {
+  let Constraints = "$ard = $t, $srd = $sr";
+}

llvm/lib/Target/Xtensa/XtensaOperands.td

 
 // imm8_sh8 predicate - Immediate in the range [-32768,32512] with (bits[7-0] == 0)
 // imm8 value left shifted by 8 bits
 def Imm8_sh8_AsmOperand : ImmAsmOperand<"Imm8_sh8">;
 def imm8_sh8 : Immediate<i32, [{ return Imm >= -32768 && Imm <= 32512 && ((Imm & 0xFF) == 0); }],
                         "Imm8_sh8_AsmOperand"> {
   let EncoderMethod = "getImm8_sh8OpValue";
 }
+
+// imm12 predicate - Immediate in the range [-2048,2047]
+def Imm12_AsmOperand : ImmAsmOperand<"Imm12">;
+def imm12 : Immediate<i32, [{ return Imm >= -2048 && Imm <= 2047; }], "Imm12_AsmOperand"> {
+  let EncoderMethod = "getImm12OpValue";
+}
+
+// imm12m predicate - Immediate for MOV operation
+def Imm12m_AsmOperand : ImmAsmOperand<"Imm12m">;
+def imm12m : Immediate<i32, [{ return Imm >= -2048 && Imm <= 2047; }], "Imm12m_AsmOperand"> {
+  let EncoderMethod = "getImm12OpValue";
+}
+
+// uimm4 predicate - Immediate in the range [0,15]
+def Uimm4_AsmOperand : ImmAsmOperand<"Uimm4">;
+def uimm4 : Immediate<i32, [{ return Imm >= 0 && Imm <= 15; }], "Uimm4_AsmOperand"> {
+  let EncoderMethod = "getUimm4OpValue";
+}
+
+// uimm5 predicate - Immediate in the range [0,31]
+def Uimm5_AsmOperand : ImmAsmOperand<"Uimm5">;
+def uimm5 : Immediate<i32, [{ return Imm >= 0 && Imm <= 31; }], "Uimm5_AsmOperand"> {
+  let EncoderMethod = "getUimm5OpValue";
+}
+
+// imm1_16 predicate - Immediate in the range [1,16]
+def Imm1_16_AsmOperand : ImmAsmOperand<"Imm1_16">;
+def imm1_16 : Immediate<i32, [{ return Imm >= 1 && Imm <= 16; }], "Imm1_16_AsmOperand"> {
+  let EncoderMethod = "getImm1_16OpValue";
+}
+
+// shimm1_31 predicate - Immediate in the range [1,31]
+def Shimm1_31_AsmOperand : ImmAsmOperand<"Shimm1_31">;
+def shimm1_31 : Immediate<i32, [{ return Imm >= 1 && Imm <= 31; }], "Shimm1_31_AsmOperand"> {
+  let EncoderMethod = "getShimm1_31OpValue";
+}
+
+// Memory offset 0..255 for 8-bit memory accesses
+def Offset8m8_AsmOperand : ImmAsmOperand<"Offset8m8">;
+def offset8m8 : Immediate<i32,
+    [{ return Imm >= 0 && Imm <= 255; }],
+    "Offset8m8_AsmOperand">;
+
+// Memory offset 0..510 for 16-bit memory accesses
+def Offset8m16_AsmOperand : ImmAsmOperand<"Offset8m16">;
+def offset8m16 : Immediate<i32,
+    [{ return Imm >= 0 && Imm <= 510 && (Imm & 0x1 == 0); }],
+    "Offset8m16_AsmOperand">;
+
+// Memory offset 0..1020 for 32-bit memory accesses
+def Offset8m32_AsmOperand : ImmAsmOperand<"Offset8m32">;
+def offset8m32 : Immediate<i32,
+    [{ return Imm >= 0 && Imm <= 1020 && (Imm & 0x3 == 0); }],
+    "Offset8m32_AsmOperand">;
+
+// Memory offset 0..60 for 32-bit memory accesses
+def Offset4m32_AsmOperand : ImmAsmOperand<"Offset4m32">;
+def offset4m32 : Immediate<i32,
+    [{ return Imm >= 0 && Imm <= 60 && (Imm & 0x3 == 0); }],
+    "Offset4m32_AsmOperand">;
+//===----------------------------------------------------------------------===//
+// Memory address operands
+//===----------------------------------------------------------------------===//
+
+class mem<Operand offset> : Operand<i32> {
+  let MIOperandInfo = (ops AR, offset);
+  let EncoderMethod = "getMemRegEncoding";
+  let OperandType = "OPERAND_MEMORY";
+  let PrintMethod = "printMemOperand";
+}
+
+def mem8   : mem<offset8m8>;
+
+def mem16  : mem<offset8m16>;
+
+def mem32  : mem<offset8m32>;
+
+def mem32n : mem<offset4m32>;
+
+//Add patterns for future use in stack addressing mode
+def addr_ish1 : ComplexPattern<iPTR, 2, "selectMemRegAddrISH1", [frameindex]>;
+def addr_ish2 : ComplexPattern<iPTR, 2, "selectMemRegAddrISH2", [frameindex]>;
+def addr_ish4 : ComplexPattern<iPTR, 2, "selectMemRegAddrISH4", [frameindex]>;

llvm/test/MC/Xtensa/Core/invalid.s

 # RUN: not llvm-mc -triple xtensa %s 2>&1 | FileCheck %s
 
+LBL0:
+
 # Out of range immediates
 
-LBL0:
+# imm12m
+movi a1, 3000
+# CHECK: :[[#@LINE-1]]:10: error: expected immediate in range [-2048, 2047]
 
 # imm8
 addi a1, a2, 300
 # CHECK: :[[#@LINE-1]]:14: error: expected immediate in range [-128, 127]
 
+# imm8
 addi a1, a2, -129
 # CHECK: :[[#@LINE-1]]:14: error: expected immediate in range [-128, 127]
 
 # imm8_sh8
 addmi a1, a2, 33
 # CHECK: :[[#@LINE-1]]:15: error: expected immediate in range [-32768, 32512], first 8 bits should be zero
 
+# shimm1_31
+slli a1, a2, 0
+# CHECK: :[[#@LINE-1]]:14: error: expected immediate in range [1, 31]
+
+# uimm4
+srli a1, a2, 16
+# CHECK: :[[#@LINE-1]]:14: error: expected immediate in range [0, 15]
+
+# uimm5
+srai a2, a3, 32
+# CHECK: :[[#@LINE-1]]:14: error: expected immediate in range [0, 31]
+
+# imm1_16
+extui a1, a3, 1, 17
+# CHECK: :[[#@LINE-1]]:18: error: expected immediate in range [1, 16]
+
+# offset8m8
+s8i a1, a2, 300
+# CHECK: :[[#@LINE-1]]:13: error: expected immediate in range [0, 255]
+
+# offset16m8
+l16si a1, a2, 512
+# CHECK: :[[#@LINE-1]]:15: error: expected immediate in range [0, 510], first bit should be zero
+
+# offset32m8
+l32i a1, a2, 1024
+# CHECK: :[[#@LINE-1]]:14: error: expected immediate in range [0, 1020], first 2 bits should be zero
+
 # Invalid number of operands
 addi a1, a2
 # CHECK: :[[#@LINE-1]]:1: error: too few operands for instruction
 addi a1, a2, 4, 4
 # CHECK: :[[#@LINE-1]]:17: error: invalid operand for instruction
 
 # Invalid mnemonics
 aaa a10, a12
 # CHECK: :[[#@LINE-1]]:1: error: unrecognized instruction mnemonic
 
 # Invalid operand types
 and sp, a2, 10
 # CHECK: :[[#@LINE-1]]:13: error: invalid operand for instruction
 addi sp, a1, a2
 # CHECK: :[[#@LINE-1]]:14: error: expected immediate in range [-128, 127]
 
-# Invalid register names
+# Check invalid register names for different formats
 # Instruction format RRR
 or r2, sp, a3
 # CHECK: :[[#@LINE-1]]:4: error: invalid operand for instruction
 and a1, r10, a3
 # CHECK: :[[#@LINE-1]]:9: error: invalid operand for instruction
 sub a1, sp, a100
 # CHECK: :[[#@LINE-1]]:13: error: invalid operand for instruction
 
 # Instruction format RRI8
 addi a101, sp, 10
 # CHECK: :[[#@LINE-1]]:6: error: invalid operand for instruction
 addi a1, r10, 10
 # CHECK: :[[#@LINE-1]]:10: error: invalid operand for instruction
 
-# Invalid operands order
+# Instruction format RSR
+wsr.uregister a2
+# CHECK: :[[#@LINE-1]]:1: error: invalid register name
+wsr a2, uregister
+# CHECK: :[[#@LINE-1]]:9: error: invalid operand for instruction
+
+# Check invalid operands order for different formats
 # Instruction format RRI8
 addi a1, 10, a2
 # CHECK: :[[#@LINE-1]]:10: error: invalid operand for instruction
+
+# Instruction format RSR
+wsr sar, a2
+# CHECK: :[[#@LINE-1]]:5: error: invalid operand for instruction

llvm/test/MC/Xtensa/Core/memory.s

+# RUN: llvm-mc %s -triple=xtensa -show-encoding \
+# RUN:     | FileCheck -check-prefixes=CHECK,CHECK-INST %s
+
+
+.align  4
+LBL0:
+
+# Instruction format RRI8
+# CHECK-INST: l8ui a2, a1, 3
+# CHECK: encoding: [0x22,0x01,0x03]
+l8ui a2, sp, 3
+
+# Instruction format RRI8
+# CHECK-INST: l16si a3, a1, 4
+# CHECK: encoding: [0x32,0x91,0x02]
+l16si a3, sp, 4
+
+# Instruction format RRI8
+# CHECK-INST: l16ui a4, a1, 6
+# CHECK: encoding: [0x42,0x11,0x03]
+l16ui a4, sp, 6
+
+# Instruction format RRI8
+# CHECK-INST: l32i a5, a1, 8
+# CHECK: encoding: [0x52,0x21,0x02]
+l32i a5, sp, 8
+
+# Instruction format RRI8
+# CHECK-INST: s8i a2, a1, 3
+# CHECK: encoding: [0x22,0x41,0x03]
+s8i a2, sp, 3
+
+# Instruction format RRI8
+# CHECK-INST: s16i a3, a1, 4
+# CHECK: encoding: [0x32,0x51,0x02]
+s16i a3, sp, 4
+
+# Instruction format RRI8
+# CHECK-INST: s32i a5, a1, 8
+# CHECK: encoding: [0x52,0x61,0x02]
+s32i a5, sp, 8

llvm/test/MC/Xtensa/Core/move.s

 # CHECK: encoding: [0x40,0x23,0x83]
 moveqz a2, a3, a4
 
 # Instruction format RRR
 # CHECK-INST: movgez a3, a11, a12
 # CHECK: encoding: [0xc0,0x3b,0xb3]
 movgez a3, a11, a12
 
+# Instruction format RRI8
+# CHECK-INST: movi a1, -2048
+# CHECK: encoding: [0x12,0xa8,0x00]
+movi a1, -2048
+
 # Instruction format RRR
 # CHECK-INST: movltz a7, a8, a9
 # CHECK: encoding: [0x90,0x78,0xa3]
 movltz a7, a8, a9
 
 # Instruction format RRR
 # CHECK-INST: movnez a10, a11, a12
 # CHECK: encoding: [0xc0,0xab,0x93]
 movnez a10, a11, a12

llvm/test/MC/Xtensa/Core/processor-control.s

 # CHECK: encoding: [0x00,0x20,0x00]
 isync
 
 # Instruction format RRR
 # CHECK-INST: nop
 # CHECK: encoding: [0xf0,0x20,0x00]
 nop
 
+# Instruction format RSR
+# CHECK-INST: rsr a8, sar
+# CHECK: encoding: [0x80,0x03,0x03]
+rsr a8, sar
+
+# CHECK-INST: rsr a8, sar
+# CHECK: encoding: [0x80,0x03,0x03]
+rsr.sar a8
+
+# CHECK-INST: rsr a8, sar
+# CHECK: encoding: [0x80,0x03,0x03]
+rsr a8, 3
+
 # Instruction format RRR
 # CHECK-INST: rsync
 # CHECK: encoding: [0x10,0x20,0x00]
 rsync
+
+# Instruction format RSR
+# CHECK-INST: wsr a8, sar
+# CHECK: encoding: [0x80,0x03,0x13]
+wsr a8, sar
+
+# CHECK-INST: wsr a8, sar
+# CHECK: encoding: [0x80,0x03,0x13]
+wsr.sar a8
+
+# CHECK-INST: wsr a8, sar
+# CHECK: encoding: [0x80,0x03,0x13]
+wsr a8, 3
+
+# Instruction format RRR
+# CHECK-INST: xsr a8, sar
+# CHECK: encoding: [0x80,0x03,0x61]
+xsr a8, sar
+
+# CHECK-INST: xsr a8, sar
+# CHECK: encoding: [0x80,0x03,0x61]
+xsr.sar a8
+
+# CHECK-INST: xsr a8, sar
+# CHECK: encoding: [0x80,0x03,0x61]
+xsr a8, 3

llvm/test/MC/Xtensa/Core/shift.s

+# RUN: llvm-mc %s -triple=xtensa -show-encoding \
+# RUN:     | FileCheck -check-prefixes=CHECK,CHECK-INST %s
+
+
+.align  4
+LBL0:
+
+# Instruction format RRR
+# CHECK-INST: extui a1, a2, 7, 8
+# CHECK: encoding: [0x20,0x17,0x74]
+extui a1, a2, 7, 8
+
+# Instruction format RRR
+# CHECK-INST: sll a10, a11
+# CHECK: encoding: [0x00,0xab,0xa1]
+sll a10, a11
+
+# Instruction format RRR
+# CHECK-INST: slli a5, a1, 15
+# CHECK: encoding: [0x10,0x51,0x11]
+slli a5, a1, 15
+
+# Instruction format RRR
+# CHECK-INST: sra a12, a3
+# CHECK: encoding: [0x30,0xc0,0xb1]
+sra a12, a3
+
+# Instruction format RRR
+# CHECK-INST: srai a8, a5, 0
+# CHECK: encoding: [0x50,0x80,0x21]
+srai a8, a5, 0
+
+# Instruction format RRR
+# CHECK-INST: src a3, a4, a5
+# CHECK: encoding: [0x50,0x34,0x81]
+src a3, a4, a5
+
+# Instruction format RRR
+# CHECK-INST: srl a6, a7
+# CHECK: encoding: [0x70,0x60,0x91]
+srl a6, a7
+
+# Instruction format RRR
+# CHECK-INST: srli a3, a4, 8
+# CHECK: encoding: [0x40,0x38,0x41]
+srli a3, a4, 8
+
+# Instruction format RRR
+# CHECK-INST: ssa8l a14
+# CHECK: encoding: [0x00,0x2e,0x40]
+ssa8l a14
+
+# Instruction format RRR
+# CHECK-INST: ssai 31
+# CHECK: encoding: [0x10,0x4f,0x40]
+ssai 31
+
+# Instruction format RRR
+# CHECK-INST: ssl a0
+# CHECK: encoding: [0x00,0x10,0x40]
+ssl a0
+
+# Instruction format RRR
+# CHECK-INST: ssr a2
+# CHECK: encoding: [0x00,0x02,0x40]
+ssr a2