AVX-512: decoder for AVX-512, made by Alexey Bader.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@198013 91177308-0d34-0410-b5e6-96231b3b80d8

7 files changed, 400 insertions, 126 deletions

diff --git a/lib/Target/X86/AsmParser/X86AsmParser.cpp b/lib/Target/X86/AsmParser/X86AsmParser.cpp
index ef363b8a61..0f8e9f7b4a 100644
--- a/lib/Target/X86/AsmParser/X86AsmParser.cpp
+++ b/lib/Target/X86/AsmParser/X86AsmParser.cpp
@@ -2067,6 +2067,7 @@ ParseInstruction(ParseInstructionInfo &Info, StringRef Name, SMLoc NameLoc,
           return true;
         }
       }
+      // TODO: add parsing of broadcasts {1to8}, {1to16}
       // Parse "zeroing non-masked" semantic {z}
       if (getLexer().is(AsmToken::LCurly)) {
         Operands.push_back(X86Operand::CreateToken("{z}", consumeToken()));
diff --git a/lib/Target/X86/Disassembler/X86Disassembler.cpp b/lib/Target/X86/Disassembler/X86Disassembler.cpp
index 903e36cfe6..459ea932c0 100644
--- a/lib/Target/X86/Disassembler/X86Disassembler.cpp
+++ b/lib/Target/X86/Disassembler/X86Disassembler.cpp
@@ -418,13 +418,22 @@ static bool translateRMMemory(MCInst &mcInst, InternalInstruction &insn,
     bool IndexIs256 = (Opcode == X86::VGATHERQPDYrm ||
                        Opcode == X86::VGATHERDPSYrm ||
                        Opcode == X86::VGATHERQPSYrm ||
+                       Opcode == X86::VGATHERDPDZrm ||
+                       Opcode == X86::VPGATHERDQZrm ||
                        Opcode == X86::VPGATHERQQYrm ||
                        Opcode == X86::VPGATHERDDYrm ||
                        Opcode == X86::VPGATHERQDYrm);
-    if (IndexIs128 || IndexIs256) {
+    bool IndexIs512 = (Opcode == X86::VGATHERQPDZrm ||
+                       Opcode == X86::VGATHERDPSZrm ||
+                       Opcode == X86::VGATHERQPSZrm ||
+                       Opcode == X86::VPGATHERQQZrm ||
+                       Opcode == X86::VPGATHERDDZrm ||
+                       Opcode == X86::VPGATHERQDZrm);
+    if (IndexIs128 || IndexIs256 || IndexIs512) {
       unsigned IndexOffset = insn.sibIndex -
                          (insn.addressSize == 8 ? SIB_INDEX_RAX:SIB_INDEX_EAX);
-      SIBIndex IndexBase = IndexIs256 ? SIB_INDEX_YMM0 : SIB_INDEX_XMM0;
+      SIBIndex IndexBase = IndexIs512 ? SIB_INDEX_ZMM0 :
+                           IndexIs256 ? SIB_INDEX_YMM0 : SIB_INDEX_XMM0;
       insn.sibIndex = (SIBIndex)(IndexBase + 
                            (insn.sibIndex == SIB_INDEX_NONE ? 4 : IndexOffset));
     }
@@ -565,6 +574,9 @@ static bool translateRM(MCInst &mcInst, const OperandSpecifier &operand,
   case TYPE_XMM128:
   case TYPE_XMM256:
   case TYPE_XMM512:
+  case TYPE_VK1:
+  case TYPE_VK8:
+  case TYPE_VK16:
   case TYPE_DEBUGREG:
   case TYPE_CONTROLREG:
     return translateRMRegister(mcInst, insn);
@@ -596,7 +608,7 @@ static bool translateRM(MCInst &mcInst, const OperandSpecifier &operand,
 ///
 /// @param mcInst       - The MCInst to append to.
 /// @param stackPos     - The stack position to translate.
-/// @return             - 0 on success; nonzero otherwise.
+/// @return             - false on success; true otherwise.
 static bool translateFPRegister(MCInst &mcInst,
                                uint8_t stackPos) {
   if (stackPos >= 8) {
@@ -609,6 +621,23 @@ static bool translateFPRegister(MCInst &mcInst,
   return false;
 }
 
+/// translateMaskRegister - Translates a 3-bit mask register number to
+///   LLVM form, and appends it to an MCInst.
+///
+/// @param mcInst       - The MCInst to append to.
+/// @param maskRegNum   - Number of mask register from 0 to 7.
+/// @return             - false on success; true otherwise.
+static bool translateMaskRegister(MCInst &mcInst,
+                                uint8_t maskRegNum) {
+  if (maskRegNum >= 8) {
+    debug("Invalid mask register number");
+    return true;
+  }
+
+  mcInst.addOperand(MCOperand::CreateReg(X86::K0 + maskRegNum));
+  return false;
+}
+
 /// translateOperand - Translates an operand stored in an internal instruction 
 ///   to LLVM's format and appends it to an MCInst.
 ///
@@ -626,6 +655,8 @@ static bool translateOperand(MCInst &mcInst, const OperandSpecifier &operand,
   case ENCODING_REG:
     translateRegister(mcInst, insn.reg);
     return false;
+  case ENCODING_WRITEMASK:
+    return translateMaskRegister(mcInst, insn.writemask);
   case ENCODING_RM:
     return translateRM(mcInst, operand, insn, Dis);
   case ENCODING_CB:
diff --git a/lib/Target/X86/Disassembler/X86DisassemblerDecoder.c b/lib/Target/X86/Disassembler/X86DisassemblerDecoder.c
index c81a85755f..4b97678d0c 100644
--- a/lib/Target/X86/Disassembler/X86DisassemblerDecoder.c
+++ b/lib/Target/X86/Disassembler/X86DisassemblerDecoder.c
@@ -40,7 +40,7 @@
  * @return          - The InstructionContext to use when looking up an
  *                    an instruction with these attributes.
  */
-static InstructionContext contextForAttrs(uint8_t attrMask) {
+static InstructionContext contextForAttrs(uint16_t attrMask) {
   return CONTEXTS_SYM[attrMask];
 }
 
@@ -57,7 +57,7 @@ static InstructionContext contextForAttrs(uint8_t attrMask) {
  */
 static int modRMRequired(OpcodeType type,
                          InstructionContext insnContext,
-                         uint8_t opcode) {
+                         uint16_t opcode) {
   const struct ContextDecision* decision = 0;
 
   switch (type) {
@@ -444,9 +444,60 @@ static int readPrefixes(struct InternalInstruction* insn) {
       dbgprintf(insn, "Found prefix 0x%hhx", byte);
   }
 
-  insn->vexXopType = TYPE_NO_VEX_XOP;
+  insn->vectorExtensionType = TYPE_NO_VEX_XOP;
 
-  if (byte == 0xc4) {
+  if (byte == 0x62) {
+    uint8_t byte1, byte2;
+
+    if (consumeByte(insn, &byte1)) {
+      dbgprintf(insn, "Couldn't read second byte of EVEX prefix");
+      return -1;
+    }
+
+    if (lookAtByte(insn, &byte2)) {
+      dbgprintf(insn, "Couldn't read third byte of EVEX prefix");
+      return -1;
+    }
+
+    if ((insn->mode == MODE_64BIT || (byte1 & 0xc0) == 0xc0) &&
+       ((~byte1 & 0xc) == 0xc) && ((byte2 & 0x4) == 0x4)) {
+      insn->vectorExtensionType = TYPE_EVEX;
+    }
+    else {
+      unconsumeByte(insn); /* unconsume byte1 */
+      unconsumeByte(insn); /* unconsume byte  */
+      insn->necessaryPrefixLocation = insn->readerCursor - 2;
+    }
+
+    if (insn->vectorExtensionType == TYPE_EVEX) {
+      insn->vectorExtensionPrefix[0] = byte;
+      insn->vectorExtensionPrefix[1] = byte1;
+      if (consumeByte(insn, &insn->vectorExtensionPrefix[2])) {
+        dbgprintf(insn, "Couldn't read third byte of EVEX prefix");
+        return -1;
+      }
+      if (consumeByte(insn, &insn->vectorExtensionPrefix[3])) {
+        dbgprintf(insn, "Couldn't read fourth byte of EVEX prefix");
+        return -1;
+      }
+
+      /* We simulate the REX prefix for simplicity's sake */
+      if (insn->mode == MODE_64BIT) {
+        insn->rexPrefix = 0x40
+                        | (wFromEVEX3of4(insn->vectorExtensionPrefix[2]) << 3)
+                        | (rFromEVEX2of4(insn->vectorExtensionPrefix[1]) << 2)
+                        | (xFromEVEX2of4(insn->vectorExtensionPrefix[1]) << 1)
+                        | (bFromEVEX2of4(insn->vectorExtensionPrefix[1]) << 0);
+      }
+
+      hasOpSize = (VEX_PREFIX_66 == ppFromEVEX3of4(insn->vectorExtensionPrefix[2]));
+
+      dbgprintf(insn, "Found EVEX prefix 0x%hhx 0x%hhx 0x%hhx 0x%hhx",
+              insn->vectorExtensionPrefix[0], insn->vectorExtensionPrefix[1],
+              insn->vectorExtensionPrefix[2], insn->vectorExtensionPrefix[3]);
+    }
+  }
+  else if (byte == 0xc4) {
     uint8_t byte1;
 
     if (lookAtByte(insn, &byte1)) {
@@ -455,7 +506,7 @@ static int readPrefixes(struct InternalInstruction* insn) {
     }
 
     if (insn->mode == MODE_64BIT || (byte1 & 0xc0) == 0xc0) {
-      insn->vexXopType = TYPE_VEX_3B;
+      insn->vectorExtensionType = TYPE_VEX_3B;
       insn->necessaryPrefixLocation = insn->readerCursor - 1;
     }
     else {
@@ -463,22 +514,22 @@ static int readPrefixes(struct InternalInstruction* insn) {
       insn->necessaryPrefixLocation = insn->readerCursor - 1;
     }
 
-    if (insn->vexXopType == TYPE_VEX_3B) {
-      insn->vexXopPrefix[0] = byte;
-      consumeByte(insn, &insn->vexXopPrefix[1]);
-      consumeByte(insn, &insn->vexXopPrefix[2]);
+    if (insn->vectorExtensionType == TYPE_VEX_3B) {
+      insn->vectorExtensionPrefix[0] = byte;
+      consumeByte(insn, &insn->vectorExtensionPrefix[1]);
+      consumeByte(insn, &insn->vectorExtensionPrefix[2]);
 
       /* We simulate the REX prefix for simplicity's sake */
 
       if (insn->mode == MODE_64BIT) {
         insn->rexPrefix = 0x40
-                        | (wFromVEX3of3(insn->vexXopPrefix[2]) << 3)
-                        | (rFromVEX2of3(insn->vexXopPrefix[1]) << 2)
-                        | (xFromVEX2of3(insn->vexXopPrefix[1]) << 1)
-                        | (bFromVEX2of3(insn->vexXopPrefix[1]) << 0);
+                        | (wFromVEX3of3(insn->vectorExtensionPrefix[2]) << 3)
+                        | (rFromVEX2of3(insn->vectorExtensionPrefix[1]) << 2)
+                        | (xFromVEX2of3(insn->vectorExtensionPrefix[1]) << 1)
+                        | (bFromVEX2of3(insn->vectorExtensionPrefix[1]) << 0);
       }
 
-      switch (ppFromVEX3of3(insn->vexXopPrefix[2]))
+      switch (ppFromVEX3of3(insn->vectorExtensionPrefix[2]))
       {
       default:
         break;
@@ -488,8 +539,8 @@ static int readPrefixes(struct InternalInstruction* insn) {
       }
 
       dbgprintf(insn, "Found VEX prefix 0x%hhx 0x%hhx 0x%hhx",
-                insn->vexXopPrefix[0], insn->vexXopPrefix[1],
-                insn->vexXopPrefix[2]);
+                insn->vectorExtensionPrefix[0], insn->vectorExtensionPrefix[1],
+                insn->vectorExtensionPrefix[2]);
     }
   }
   else if (byte == 0xc5) {
@@ -501,22 +552,22 @@ static int readPrefixes(struct InternalInstruction* insn) {
     }
 
     if (insn->mode == MODE_64BIT || (byte1 & 0xc0) == 0xc0) {
-      insn->vexXopType = TYPE_VEX_2B;
+      insn->vectorExtensionType = TYPE_VEX_2B;
     }
     else {
       unconsumeByte(insn);
     }
 
-    if (insn->vexXopType == TYPE_VEX_2B) {
-      insn->vexXopPrefix[0] = byte;
-      consumeByte(insn, &insn->vexXopPrefix[1]);
+    if (insn->vectorExtensionType == TYPE_VEX_2B) {
+      insn->vectorExtensionPrefix[0] = byte;
+      consumeByte(insn, &insn->vectorExtensionPrefix[1]);
 
       if (insn->mode == MODE_64BIT) {
         insn->rexPrefix = 0x40
-                        | (rFromVEX2of2(insn->vexXopPrefix[1]) << 2);
+                        | (rFromVEX2of2(insn->vectorExtensionPrefix[1]) << 2);
       }
 
-      switch (ppFromVEX2of2(insn->vexXopPrefix[1]))
+      switch (ppFromVEX2of2(insn->vectorExtensionPrefix[1]))
       {
       default:
         break;
@@ -525,7 +576,9 @@ static int readPrefixes(struct InternalInstruction* insn) {
         break;
       }
 
-      dbgprintf(insn, "Found VEX prefix 0x%hhx 0x%hhx", insn->vexXopPrefix[0], insn->vexXopPrefix[1]);
+      dbgprintf(insn, "Found VEX prefix 0x%hhx 0x%hhx",
+                insn->vectorExtensionPrefix[0],
+                insn->vectorExtensionPrefix[1]);
     }
   }
   else if (byte == 0x8f) {
@@ -537,7 +590,7 @@ static int readPrefixes(struct InternalInstruction* insn) {
     }
 
     if ((byte1 & 0x38) != 0x0) { /* 0 in these 3 bits is a POP instruction. */
-      insn->vexXopType = TYPE_XOP;
+      insn->vectorExtensionType = TYPE_XOP;
       insn->necessaryPrefixLocation = insn->readerCursor - 1;
     }
     else {
@@ -545,22 +598,22 @@ static int readPrefixes(struct InternalInstruction* insn) {
       insn->necessaryPrefixLocation = insn->readerCursor - 1;
     }
 
-    if (insn->vexXopType == TYPE_XOP) {
-      insn->vexXopPrefix[0] = byte;
-      consumeByte(insn, &insn->vexXopPrefix[1]);
-      consumeByte(insn, &insn->vexXopPrefix[2]);
+    if (insn->vectorExtensionType == TYPE_XOP) {
+      insn->vectorExtensionPrefix[0] = byte;
+      consumeByte(insn, &insn->vectorExtensionPrefix[1]);
+      consumeByte(insn, &insn->vectorExtensionPrefix[2]);
 
       /* We simulate the REX prefix for simplicity's sake */
 
       if (insn->mode == MODE_64BIT) {
         insn->rexPrefix = 0x40
-                        | (wFromXOP3of3(insn->vexXopPrefix[2]) << 3)
-                        | (rFromXOP2of3(insn->vexXopPrefix[1]) << 2)
-                        | (xFromXOP2of3(insn->vexXopPrefix[1]) << 1)
-                        | (bFromXOP2of3(insn->vexXopPrefix[1]) << 0);
+                        | (wFromXOP3of3(insn->vectorExtensionPrefix[2]) << 3)
+                        | (rFromXOP2of3(insn->vectorExtensionPrefix[1]) << 2)
+                        | (xFromXOP2of3(insn->vectorExtensionPrefix[1]) << 1)
+                        | (bFromXOP2of3(insn->vectorExtensionPrefix[1]) << 0);
       }
 
-      switch (ppFromXOP3of3(insn->vexXopPrefix[2]))
+      switch (ppFromXOP3of3(insn->vectorExtensionPrefix[2]))
       {
       default:
         break;
@@ -570,8 +623,8 @@ static int readPrefixes(struct InternalInstruction* insn) {
       }
 
       dbgprintf(insn, "Found XOP prefix 0x%hhx 0x%hhx 0x%hhx",
-                insn->vexXopPrefix[0], insn->vexXopPrefix[1],
-                insn->vexXopPrefix[2]);
+                insn->vectorExtensionPrefix[0], insn->vectorExtensionPrefix[1],
+                insn->vectorExtensionPrefix[2]);
     }
   }
   else {
@@ -646,13 +699,29 @@ static int readOpcode(struct InternalInstruction* insn) {
 
   insn->opcodeType = ONEBYTE;
 
-  if (insn->vexXopType == TYPE_VEX_3B)
+  if (insn->vectorExtensionType == TYPE_EVEX)
   {
-    switch (mmmmmFromVEX2of3(insn->vexXopPrefix[1]))
-    {
+    switch (mmFromEVEX2of4(insn->vectorExtensionPrefix[1])) {
+    default:
+      dbgprintf(insn, "Unhandled mm field for instruction (0x%hhx)",
+                mmFromEVEX2of4(insn->vectorExtensionPrefix[1]));
+      return -1;
+    case VEX_LOB_0F:
+      insn->opcodeType = TWOBYTE;
+      return consumeByte(insn, &insn->opcode);
+    case VEX_LOB_0F38:
+      insn->opcodeType = THREEBYTE_38;
+      return consumeByte(insn, &insn->opcode);
+    case VEX_LOB_0F3A:
+      insn->opcodeType = THREEBYTE_3A;
+      return consumeByte(insn, &insn->opcode);
+    }
+  }
+  else if (insn->vectorExtensionType == TYPE_VEX_3B) {
+    switch (mmmmmFromVEX2of3(insn->vectorExtensionPrefix[1])) {
     default:
       dbgprintf(insn, "Unhandled m-mmmm field for instruction (0x%hhx)",
-                mmmmmFromVEX2of3(insn->vexXopPrefix[1]));
+                mmmmmFromVEX2of3(insn->vectorExtensionPrefix[1]));
       return -1;
     case VEX_LOB_0F:
       insn->opcodeType = TWOBYTE;
@@ -665,18 +734,15 @@ static int readOpcode(struct InternalInstruction* insn) {
       return consumeByte(insn, &insn->opcode);
     }
   }
-  else if (insn->vexXopType == TYPE_VEX_2B)
-  {
+  else if (insn->vectorExtensionType == TYPE_VEX_2B) {
     insn->opcodeType = TWOBYTE;
     return consumeByte(insn, &insn->opcode);
   }
-  else if (insn->vexXopType == TYPE_XOP)
-  {
-    switch (mmmmmFromXOP2of3(insn->vexXopPrefix[1]))
-    {
+  else if (insn->vectorExtensionType == TYPE_XOP) {
+    switch (mmmmmFromXOP2of3(insn->vectorExtensionPrefix[1])) {
     default:
       dbgprintf(insn, "Unhandled m-mmmm field for instruction (0x%hhx)",
-                mmmmmFromVEX2of3(insn->vexXopPrefix[1]));
+                mmmmmFromVEX2of3(insn->vectorExtensionPrefix[1]));
       return -1;
     case XOP_MAP_SELECT_8:
       insn->opcodeType = XOP8_MAP;
@@ -760,10 +826,10 @@ static int readModRM(struct InternalInstruction* insn);
  */
 static int getIDWithAttrMask(uint16_t* instructionID,
                              struct InternalInstruction* insn,
-                             uint8_t attrMask) {
+                             uint16_t attrMask) {
   BOOL hasModRMExtension;
 
-  uint8_t instructionClass;
+  uint16_t instructionClass;
 
   instructionClass = contextForAttrs(attrMask);
 
@@ -826,7 +892,7 @@ static BOOL is16BitEquivalent(const char* orig, const char* equiv) {
  *                nonzero otherwise.
  */
 static int getID(struct InternalInstruction* insn, const void *miiArg) {
-  uint8_t attrMask;
+  uint16_t attrMask;
   uint16_t instructionID;
 
   dbgprintf(insn, "getID()");
@@ -836,11 +902,35 @@ static int getID(struct InternalInstruction* insn, const void *miiArg) {
   if (insn->mode == MODE_64BIT)
     attrMask |= ATTR_64BIT;
 
-  if (insn->vexXopType != TYPE_NO_VEX_XOP) {
-    attrMask |= ATTR_VEX;
+  if (insn->vectorExtensionType != TYPE_NO_VEX_XOP) {
+    attrMask |= (insn->vectorExtensionType == TYPE_EVEX) ? ATTR_EVEX : ATTR_VEX;
+
+    if (insn->vectorExtensionType == TYPE_EVEX) {
+      switch (ppFromEVEX3of4(insn->vectorExtensionPrefix[2])) {
+      case VEX_PREFIX_66:
+        attrMask |= ATTR_OPSIZE;
+        break;
+      case VEX_PREFIX_F3:
+        attrMask |= ATTR_XS;
+        break;
+      case VEX_PREFIX_F2:
+        attrMask |= ATTR_XD;
+        break;
+      }
 
-    if (insn->vexXopType == TYPE_VEX_3B) {
-      switch (ppFromVEX3of3(insn->vexXopPrefix[2])) {
+      if (zFromEVEX4of4(insn->vectorExtensionPrefix[3]))
+        attrMask |= ATTR_EVEXKZ;
+      if (bFromEVEX4of4(insn->vectorExtensionPrefix[3]))
+        attrMask |= ATTR_EVEXB;
+      if (aaaFromEVEX4of4(insn->vectorExtensionPrefix[3]))
+        attrMask |= ATTR_EVEXK;
+      if (lFromEVEX4of4(insn->vectorExtensionPrefix[3]))
+        attrMask |= ATTR_EVEXL;
+      if (l2FromEVEX4of4(insn->vectorExtensionPrefix[3]))
+        attrMask |= ATTR_EVEXL2;
+    }
+    else if (insn->vectorExtensionType == TYPE_VEX_3B) {
+      switch (ppFromVEX3of3(insn->vectorExtensionPrefix[2])) {
       case VEX_PREFIX_66:
         attrMask |= ATTR_OPSIZE;
         break;
@@ -852,11 +942,11 @@ static int getID(struct InternalInstruction* insn, const void *miiArg) {
         break;
       }
 
-      if (lFromVEX3of3(insn->vexXopPrefix[2]))
+      if (lFromVEX3of3(insn->vectorExtensionPrefix[2]))
         attrMask |= ATTR_VEXL;
     }
-    else if (insn->vexXopType == TYPE_VEX_2B) {
-      switch (ppFromVEX2of2(insn->vexXopPrefix[1])) {
+    else if (insn->vectorExtensionType == TYPE_VEX_2B) {
+      switch (ppFromVEX2of2(insn->vectorExtensionPrefix[1])) {
       case VEX_PREFIX_66:
         attrMask |= ATTR_OPSIZE;
         break;
@@ -868,11 +958,11 @@ static int getID(struct InternalInstruction* insn, const void *miiArg) {
         break;
       }
 
-      if (lFromVEX2of2(insn->vexXopPrefix[1]))
+      if (lFromVEX2of2(insn->vectorExtensionPrefix[1]))
         attrMask |= ATTR_VEXL;
     }
-    else if (insn->vexXopType == TYPE_XOP) {
-      switch (ppFromXOP3of3(insn->vexXopPrefix[2])) {
+    else if (insn->vectorExtensionType == TYPE_XOP) {
+      switch (ppFromXOP3of3(insn->vectorExtensionPrefix[2])) {
       case VEX_PREFIX_66:
         attrMask |= ATTR_OPSIZE;
         break;
@@ -884,7 +974,7 @@ static int getID(struct InternalInstruction* insn, const void *miiArg) {
         break;
       }
 
-      if (lFromXOP3of3(insn->vexXopPrefix[2]))
+      if (lFromXOP3of3(insn->vectorExtensionPrefix[2]))
         attrMask |= ATTR_VEXL;
     }
     else {
@@ -1033,6 +1123,8 @@ static int readSIB(struct InternalInstruction* insn) {
     return -1;
 
   index = indexFromSIB(insn->sib) | (xFromREX(insn->rexPrefix) << 3);
+  if (insn->vectorExtensionType == TYPE_EVEX)
+    index |= v2FromEVEX4of4(insn->vectorExtensionPrefix[3]) << 4;
 
   switch (index) {
   case 0x4:
@@ -1183,6 +1275,10 @@ static int readModRM(struct InternalInstruction* insn) {
 
   reg |= rFromREX(insn->rexPrefix) << 3;
   rm  |= bFromREX(insn->rexPrefix) << 3;
+  if (insn->vectorExtensionType == TYPE_EVEX) {
+    reg |= r2FromEVEX2of4(insn->vectorExtensionPrefix[1]) << 4;
+    rm  |=  xFromEVEX2of4(insn->vectorExtensionPrefix[1]) << 4;
+  }
 
   insn->reg = (Reg)(insn->regBase + reg);
 
@@ -1229,6 +1325,7 @@ static int readModRM(struct InternalInstruction* insn) {
     case 0x0:
       insn->eaDisplacement = EA_DISP_NONE; /* readSIB may override this */
       switch (rm) {
+      case 0x14:
       case 0x4:
       case 0xc:   /* in case REXW.b is set */
         insn->eaBase = (insn->addressSize == 4 ?
@@ -1252,6 +1349,7 @@ static int readModRM(struct InternalInstruction* insn) {
     case 0x2:
       insn->eaDisplacement = (mod == 0x1 ? EA_DISP_8 : EA_DISP_32);
       switch (rm) {
+      case 0x14:
       case 0x4:
       case 0xc:   /* in case REXW.b is set */
         insn->eaBase = EA_BASE_sib;
@@ -1312,6 +1410,10 @@ static int readModRM(struct InternalInstruction* insn) {
     case TYPE_XMM32:                                      \
     case TYPE_XMM:                                        \
       return prefix##_XMM0 + index;                       \
+    case TYPE_VK1:                                        \
+    case TYPE_VK8:                                        \
+    case TYPE_VK16:                                       \
+      return prefix##_K0 + index;                         \
     case TYPE_MM64:                                       \
     case TYPE_MM32:                                       \
     case TYPE_MM:                                         \
@@ -1550,12 +1652,14 @@ static int readImmediate(struct InternalInstruction* insn, uint8_t size) {
 static int readVVVV(struct InternalInstruction* insn) {
   dbgprintf(insn, "readVVVV()");
 
-  if (insn->vexXopType == TYPE_VEX_3B)
-    insn->vvvv = vvvvFromVEX3of3(insn->vexXopPrefix[2]);
-  else if (insn->vexXopType == TYPE_VEX_2B)
-    insn->vvvv = vvvvFromVEX2of2(insn->vexXopPrefix[1]);
-  else if (insn->vexXopType == TYPE_XOP)
-    insn->vvvv = vvvvFromXOP3of3(insn->vexXopPrefix[2]);
+  if (insn->vectorExtensionType == TYPE_EVEX)
+    insn->vvvv = vvvvFromEVEX3of4(insn->vectorExtensionPrefix[2]);
+  else if (insn->vectorExtensionType == TYPE_VEX_3B)
+    insn->vvvv = vvvvFromVEX3of3(insn->vectorExtensionPrefix[2]);
+  else if (insn->vectorExtensionType == TYPE_VEX_2B)
+    insn->vvvv = vvvvFromVEX2of2(insn->vectorExtensionPrefix[1]);
+  else if (insn->vectorExtensionType == TYPE_XOP)
+    insn->vvvv = vvvvFromXOP3of3(insn->vectorExtensionPrefix[2]);
   else
     return -1;
 
@@ -1566,6 +1670,23 @@ static int readVVVV(struct InternalInstruction* insn) {
 }
 
 /*
+ * readMaskRegister - Reads an mask register from the opcode field of an
+ *   instruction.
+ *
+ * @param insn    - The instruction whose opcode field is to be read.
+ * @return        - 0 on success; nonzero otherwise.
+ */
+static int readMaskRegister(struct InternalInstruction* insn) {
+  dbgprintf(insn, "readMaskRegister()");
+
+  if (insn->vectorExtensionType != TYPE_EVEX)
+    return -1;
+
+  insn->writemask = aaaFromEVEX4of4(insn->vectorExtensionPrefix[3]);
+  return 0;
+}
+
+/*
  * readOperands - Consults the specifier for an instruction and consumes all
  *   operands for that instruction, interpreting them as it goes.
  *
@@ -1675,6 +1796,10 @@ static int readOperands(struct InternalInstruction* insn) {
       if (fixupReg(insn, &x86OperandSets[insn->spec->operands][index]))
         return -1;
       break;
+    case ENCODING_WRITEMASK:
+      if (readMaskRegister(insn))
+        return -1;
+      break;
     case ENCODING_DUP:
       break;
     default:
diff --git a/lib/Target/X86/Disassembler/X86DisassemblerDecoder.h b/lib/Target/X86/Disassembler/X86DisassemblerDecoder.h
index 6d03d5ca5f..294c47688f 100644
--- a/lib/Target/X86/Disassembler/X86DisassemblerDecoder.h
+++ b/lib/Target/X86/Disassembler/X86DisassemblerDecoder.h
@@ -45,6 +45,21 @@ extern "C" {
 #define xFromREX(rex)        (((rex) & 0x2) >> 1)
 #define bFromREX(rex)        ((rex) & 0x1)
 
+#define rFromEVEX2of4(evex)     (((~(evex)) & 0x80) >> 7)
+#define xFromEVEX2of4(evex)     (((~(evex)) & 0x40) >> 6)
+#define bFromEVEX2of4(evex)     (((~(evex)) & 0x20) >> 5)
+#define r2FromEVEX2of4(evex)    (((~(evex)) & 0x10) >> 4)
+#define mmFromEVEX2of4(evex)    ((evex) & 0x3)
+#define wFromEVEX3of4(evex)     (((evex) & 0x80) >> 7)
+#define vvvvFromEVEX3of4(evex)  (((~(evex)) & 0x78) >> 3)
+#define ppFromEVEX3of4(evex)    ((evex) & 0x3)
+#define zFromEVEX4of4(evex)     (((evex) & 0x80) >> 7)
+#define l2FromEVEX4of4(evex)    (((evex) & 0x40) >> 6)
+#define lFromEVEX4of4(evex)     (((evex) & 0x20) >> 5)
+#define bFromEVEX4of4(evex)     (((evex) & 0x10) >> 4)
+#define v2FromEVEX4of4(evex)    (((~evex) & 0x8) >> 3)
+#define aaaFromEVEX4of4(evex)   ((evex) & 0x7)
+
 #define rFromVEX2of3(vex)       (((~(vex)) & 0x80) >> 7)
 #define xFromVEX2of3(vex)       (((~(vex)) & 0x40) >> 6)
 #define bFromVEX2of3(vex)       (((~(vex)) & 0x20) >> 5)
@@ -314,6 +329,16 @@ extern "C" {
   ENTRY(ZMM30)    \
   ENTRY(ZMM31)
 
+#define REGS_MASKS \
+  ENTRY(K0)        \
+  ENTRY(K1)        \
+  ENTRY(K2)        \
+  ENTRY(K3)        \
+  ENTRY(K4)        \
+  ENTRY(K5)        \
+  ENTRY(K6)        \
+  ENTRY(K7)
+
 #define REGS_SEGMENT \
   ENTRY(ES)          \
   ENTRY(CS)          \
@@ -361,6 +386,7 @@ extern "C" {
   REGS_XMM            \
   REGS_YMM            \
   REGS_ZMM            \
+  REGS_MASKS          \
   REGS_SEGMENT        \
   REGS_DEBUG          \
   REGS_CONTROL        \
@@ -463,7 +489,7 @@ typedef enum {
 } XOPMapSelect;
 
 /*
- * VEXPrefixCode - Possible values for the VEX.pp field
+ * VEXPrefixCode - Possible values for the VEX.pp/EVEX.pp field
  */
 
 typedef enum {
@@ -474,11 +500,12 @@ typedef enum {
 } VEXPrefixCode;
 
 typedef enum {
-  TYPE_NO_VEX_XOP = 0x0,
-  TYPE_VEX_2B = 0x1,
-  TYPE_VEX_3B = 0x2,
-  TYPE_XOP = 0x3
-} VEXXOPType;
+  TYPE_NO_VEX_XOP   = 0x0,
+  TYPE_VEX_2B       = 0x1,
+  TYPE_VEX_3B       = 0x2,
+  TYPE_EVEX         = 0x3,
+  TYPE_XOP          = 0x4
+} VectorExtensionType;
 
 typedef uint8_t BOOL;
 
@@ -536,10 +563,10 @@ struct InternalInstruction {
   uint8_t prefixPresent[0x100];
   /* contains the location (for use with the reader) of the prefix byte */
   uint64_t prefixLocations[0x100];
-  /* The value of the VEX/XOP prefix, if present */
-  uint8_t vexXopPrefix[3];
-  /* The length of the VEX prefix (0 if not present) */
-  VEXXOPType vexXopType;
+  /* The value of the vector extention prefix(EVEX/VEX/XOP), if present */
+  uint8_t vectorExtensionPrefix[4];
+  /* The type of the vector extension prefix */
+  VectorExtensionType vectorExtensionType;
   /* The value of the REX prefix, if present */
   uint8_t rexPrefix;
   /* The location where a mandatory prefix would have to be (i.e., right before
@@ -585,6 +612,9 @@ struct InternalInstruction {
      instructions */
   Reg                           vvvv;
 
+  /* The writemask for AVX-512 instructions which is contained in EVEX.aaa */
+  Reg                           writemask;
+
   /* The ModR/M byte, which contains most register operands and some portion of
      all memory operands */
   BOOL                          consumedModRM;
diff --git a/lib/Target/X86/Disassembler/X86DisassemblerDecoderCommon.h b/lib/Target/X86/Disassembler/X86DisassemblerDecoderCommon.h
index 44d61294d4..7afe088c6b 100644
--- a/lib/Target/X86/Disassembler/X86DisassemblerDecoderCommon.h
+++ b/lib/Target/X86/Disassembler/X86DisassemblerDecoderCommon.h
@@ -53,16 +53,22 @@
  * processed correctly.  Most of these indicate the presence of particular
  * prefixes, but ATTR_64BIT is simply an attribute of the decoding context.
  */
-#define ATTRIBUTE_BITS          \
-  ENUM_ENTRY(ATTR_NONE,   0x00) \
-  ENUM_ENTRY(ATTR_64BIT,  0x01) \
-  ENUM_ENTRY(ATTR_XS,     0x02) \
-  ENUM_ENTRY(ATTR_XD,     0x04) \
-  ENUM_ENTRY(ATTR_REXW,   0x08) \
-  ENUM_ENTRY(ATTR_OPSIZE, 0x10) \
-  ENUM_ENTRY(ATTR_ADSIZE, 0x20) \
-  ENUM_ENTRY(ATTR_VEX,    0x40) \
-  ENUM_ENTRY(ATTR_VEXL,   0x80)
+#define ATTRIBUTE_BITS                  \
+  ENUM_ENTRY(ATTR_NONE,   0x00)         \
+  ENUM_ENTRY(ATTR_64BIT,  (0x1 << 0))   \
+  ENUM_ENTRY(ATTR_XS,     (0x1 << 1))   \
+  ENUM_ENTRY(ATTR_XD,     (0x1 << 2))   \
+  ENUM_ENTRY(ATTR_REXW,   (0x1 << 3))   \
+  ENUM_ENTRY(ATTR_OPSIZE, (0x1 << 4))   \
+  ENUM_ENTRY(ATTR_ADSIZE, (0x1 << 5))   \
+  ENUM_ENTRY(ATTR_VEX,    (0x1 << 6))   \
+  ENUM_ENTRY(ATTR_VEXL,   (0x1 << 7))   \
+  ENUM_ENTRY(ATTR_EVEX,   (0x1 << 8))   \
+  ENUM_ENTRY(ATTR_EVEXL,  (0x1 << 9))   \
+  ENUM_ENTRY(ATTR_EVEXL2, (0x1 << 10))  \
+  ENUM_ENTRY(ATTR_EVEXK,  (0x1 << 11))  \
+  ENUM_ENTRY(ATTR_EVEXKZ, (0x1 << 12))  \
+  ENUM_ENTRY(ATTR_EVEXB,  (0x1 << 13))
 
 #define ENUM_ENTRY(n, v) n = v,
 enum attributeBits {
@@ -73,7 +79,7 @@ enum attributeBits {
 
 /*
  * Combinations of the above attributes that are relevant to instruction
- * decode.  Although other combinations are possible, they can be reduced to
+ * decode. Although other combinations are possible, they can be reduced to
  * these without affecting the ultimately decoded instruction.
  */
 
@@ -198,38 +204,38 @@ enum attributeBits {
   ENUM_ENTRY(IC_EVEX_L2_W_XS_B,     4,  "requires EVEX_B, L2, W and XS prefix")    \
   ENUM_ENTRY(IC_EVEX_L2_W_XD_B,     4,  "requires EVEX_B, L2, W and XD prefix")    \
   ENUM_ENTRY(IC_EVEX_L2_W_OPSIZE_B, 4,  "requires EVEX_B, L2, W and OpSize")       \
-  ENUM_ENTRY(IC_EVEX_K_B,             1,  "requires EVEX_B and EVEX_K prefix")             \
-  ENUM_ENTRY(IC_EVEX_XS_K_B,          2,  "requires EVEX_B, EVEX_K and the XS prefix")     \
-  ENUM_ENTRY(IC_EVEX_XD_K_B,          2,  "requires EVEX_B, EVEX_K and the XD prefix")     \
-  ENUM_ENTRY(IC_EVEX_OPSIZE_K_B,      2,  "requires EVEX_B, EVEX_K and the OpSize prefix") \
-  ENUM_ENTRY(IC_EVEX_W_K_B,           3,  "requires EVEX_B, EVEX_K and the W prefix")      \
-  ENUM_ENTRY(IC_EVEX_W_XS_K_B,        4,  "requires EVEX_B, EVEX_K, W, and XS prefix")     \
-  ENUM_ENTRY(IC_EVEX_W_XD_K_B,        4,  "requires EVEX_B, EVEX_K, W, and XD prefix")     \
-  ENUM_ENTRY(IC_EVEX_W_OPSIZE_K_B,    4,  "requires EVEX_B, EVEX_K, W, and OpSize")        \
-  ENUM_ENTRY(IC_EVEX_L_K_B,           3,  "requires EVEX_B, EVEX_K and the L prefix")       \
-  ENUM_ENTRY(IC_EVEX_L_XS_K_B,        4,  "requires EVEX_B, EVEX_K and the L and XS prefix")\
-  ENUM_ENTRY(IC_EVEX_L_XD_K_B,        4,  "requires EVEX_B, EVEX_K and the L and XD prefix")\
-  ENUM_ENTRY(IC_EVEX_L_OPSIZE_K_B,    4,  "requires EVEX_B, EVEX_K, L, and OpSize")         \
-  ENUM_ENTRY(IC_EVEX_L_W_K_B,         3,  "requires EVEX_B, EVEX_K, L and W")               \
-  ENUM_ENTRY(IC_EVEX_L_W_XS_K_B,      4,  "requires EVEX_B, EVEX_K, L, W and XS prefix")    \
-  ENUM_ENTRY(IC_EVEX_L_W_XD_K_B,      4,  "requires EVEX_B, EVEX_K, L, W and XD prefix")    \
-  ENUM_ENTRY(IC_EVEX_L_W_OPSIZE_K_B,  4,  "requires EVEX_B, EVEX_K, L, W and OpSize")       \
-  ENUM_ENTRY(IC_EVEX_L2_K_B,          3,  "requires EVEX_B, EVEX_K and the L2 prefix")       \
-  ENUM_ENTRY(IC_EVEX_L2_XS_K_B,       4,  "requires EVEX_B, EVEX_K and the L2 and XS prefix")\
-  ENUM_ENTRY(IC_EVEX_L2_XD_K_B,       4,  "requires EVEX_B, EVEX_K and the L2 and XD prefix")\
-  ENUM_ENTRY(IC_EVEX_L2_OPSIZE_K_B,   4,  "requires EVEX_B, EVEX_K, L2, and OpSize")         \
-  ENUM_ENTRY(IC_EVEX_L2_W_K_B,        3,  "requires EVEX_B, EVEX_K, L2 and W")               \
-  ENUM_ENTRY(IC_EVEX_L2_W_XS_K_B,     4,  "requires EVEX_B, EVEX_K, L2, W and XS prefix")    \
-  ENUM_ENTRY(IC_EVEX_L2_W_XD_K_B,     4,  "requires EVEX_B, EVEX_K, L2, W and XD prefix")    \
-  ENUM_ENTRY(IC_EVEX_L2_W_OPSIZE_K_B, 4,  "requires EVEX_B, EVEX_K, L2, W and OpSize")       \
-  ENUM_ENTRY(IC_EVEX_KZ_B,             1,  "requires EVEX_B and EVEX_KZ prefix")             \
-  ENUM_ENTRY(IC_EVEX_XS_KZ_B,          2,  "requires EVEX_B, EVEX_KZ and the XS prefix")     \
-  ENUM_ENTRY(IC_EVEX_XD_KZ_B,          2,  "requires EVEX_B, EVEX_KZ and the XD prefix")     \
-  ENUM_ENTRY(IC_EVEX_OPSIZE_KZ_B,      2,  "requires EVEX_B, EVEX_KZ and the OpSize prefix") \
-  ENUM_ENTRY(IC_EVEX_W_KZ_B,           3,  "requires EVEX_B, EVEX_KZ and the W prefix")      \
-  ENUM_ENTRY(IC_EVEX_W_XS_KZ_B,        4,  "requires EVEX_B, EVEX_KZ, W, and XS prefix")     \
-  ENUM_ENTRY(IC_EVEX_W_XD_KZ_B,        4,  "requires EVEX_B, EVEX_KZ, W, and XD prefix")     \
-  ENUM_ENTRY(IC_EVEX_W_OPSIZE_KZ_B,    4,  "requires EVEX_B, EVEX_KZ, W, and OpSize")        \
+  ENUM_ENTRY(IC_EVEX_K_B,           1,  "requires EVEX_B and EVEX_K prefix")             \
+  ENUM_ENTRY(IC_EVEX_XS_K_B,        2,  "requires EVEX_B, EVEX_K and the XS prefix")     \
+  ENUM_ENTRY(IC_EVEX_XD_K_B,        2,  "requires EVEX_B, EVEX_K and the XD prefix")     \
+  ENUM_ENTRY(IC_EVEX_OPSIZE_K_B,    2,  "requires EVEX_B, EVEX_K and the OpSize prefix") \
+  ENUM_ENTRY(IC_EVEX_W_K_B,         3,  "requires EVEX_B, EVEX_K and the W prefix")      \
+  ENUM_ENTRY(IC_EVEX_W_XS_K_B,      4,  "requires EVEX_B, EVEX_K, W, and XS prefix")     \
+  ENUM_ENTRY(IC_EVEX_W_XD_K_B,      4,  "requires EVEX_B, EVEX_K, W, and XD prefix")     \
+  ENUM_ENTRY(IC_EVEX_W_OPSIZE_K_B,  4,  "requires EVEX_B, EVEX_K, W, and OpSize")        \
+  ENUM_ENTRY(IC_EVEX_L_K_B,         3,  "requires EVEX_B, EVEX_K and the L prefix")       \
+  ENUM_ENTRY(IC_EVEX_L_XS_K_B,      4,  "requires EVEX_B, EVEX_K and the L and XS prefix")\
+  ENUM_ENTRY(IC_EVEX_L_XD_K_B,      4,  "requires EVEX_B, EVEX_K and the L and XD prefix")\
+  ENUM_ENTRY(IC_EVEX_L_OPSIZE_K_B,  4,  "requires EVEX_B, EVEX_K, L, and OpSize")         \
+  ENUM_ENTRY(IC_EVEX_L_W_K_B,       3,  "requires EVEX_B, EVEX_K, L and W")               \
+  ENUM_ENTRY(IC_EVEX_L_W_XS_K_B,    4,  "requires EVEX_B, EVEX_K, L, W and XS prefix")    \
+  ENUM_ENTRY(IC_EVEX_L_W_XD_K_B,    4,  "requires EVEX_B, EVEX_K, L, W and XD prefix")    \
+  ENUM_ENTRY(IC_EVEX_L_W_OPSIZE_K_B,4,  "requires EVEX_B, EVEX_K, L, W and OpSize")       \
+  ENUM_ENTRY(IC_EVEX_L2_K_B,        3,  "requires EVEX_B, EVEX_K and the L2 prefix")       \
+  ENUM_ENTRY(IC_EVEX_L2_XS_K_B,     4,  "requires EVEX_B, EVEX_K and the L2 and XS prefix")\
+  ENUM_ENTRY(IC_EVEX_L2_XD_K_B,     4,  "requires EVEX_B, EVEX_K and the L2 and XD prefix")\
+  ENUM_ENTRY(IC_EVEX_L2_OPSIZE_K_B, 4,  "requires EVEX_B, EVEX_K, L2, and OpSize")         \
+  ENUM_ENTRY(IC_EVEX_L2_W_K_B,      3,  "requires EVEX_B, EVEX_K, L2 and W")               \
+  ENUM_ENTRY(IC_EVEX_L2_W_XS_K_B,   4,  "requires EVEX_B, EVEX_K, L2, W and XS prefix")    \
+  ENUM_ENTRY(IC_EVEX_L2_W_XD_K_B,   4,  "requires EVEX_B, EVEX_K, L2, W and XD prefix")    \
+  ENUM_ENTRY(IC_EVEX_L2_W_OPSIZE_K_B,4,  "requires EVEX_B, EVEX_K, L2, W and OpSize")       \
+  ENUM_ENTRY(IC_EVEX_KZ_B,           1,  "requires EVEX_B and EVEX_KZ prefix")             \
+  ENUM_ENTRY(IC_EVEX_XS_KZ_B,        2,  "requires EVEX_B, EVEX_KZ and the XS prefix")     \
+  ENUM_ENTRY(IC_EVEX_XD_KZ_B,        2,  "requires EVEX_B, EVEX_KZ and the XD prefix")     \
+  ENUM_ENTRY(IC_EVEX_OPSIZE_KZ_B,    2,  "requires EVEX_B, EVEX_KZ and the OpSize prefix") \
+  ENUM_ENTRY(IC_EVEX_W_KZ_B,         3,  "requires EVEX_B, EVEX_KZ and the W prefix")      \
+  ENUM_ENTRY(IC_EVEX_W_XS_KZ_B,      4,  "requires EVEX_B, EVEX_KZ, W, and XS prefix")     \
+  ENUM_ENTRY(IC_EVEX_W_XD_KZ_B,      4,  "requires EVEX_B, EVEX_KZ, W, and XD prefix")     \
+  ENUM_ENTRY(IC_EVEX_W_OPSIZE_KZ_B,  4,  "requires EVEX_B, EVEX_KZ, W, and OpSize")        \
   ENUM_ENTRY(IC_EVEX_L_KZ_B,           3,  "requires EVEX_B, EVEX_KZ and the L prefix")       \
   ENUM_ENTRY(IC_EVEX_L_XS_KZ_B,        4,  "requires EVEX_B, EVEX_KZ and the L and XS prefix")\
   ENUM_ENTRY(IC_EVEX_L_XD_KZ_B,        4,  "requires EVEX_B, EVEX_KZ and the L and XD prefix")\
diff --git a/test/MC/Disassembler/X86/avx-512.txt b/test/MC/Disassembler/X86/avx-512.txt
new file mode 100644
index 0000000000..e5ad2a9e44
--- /dev/null
+++ b/test/MC/Disassembler/X86/avx-512.txt
@@ -0,0 +1,59 @@
+# RUN: llvm-mc --disassemble %s -triple=x86_64-apple-darwin9 -mcpu=knl | FileCheck %s
+
+# CHECK: vpbroadcastd    %xmm18, %zmm28 {%k7} {z}
+0x62 0x22 0x7d 0xcf 0x58 0xe2
+
+# CHECK: vbroadcastss    (%rsp), %zmm28
+0x62 0x62 0x7d 0x48 0x18 0x24 0x24
+
+# CHECK: vblendmpd       (%rsi), %zmm2, %zmm8 {%k7}
+0x62 0x72 0xed 0x4f 0x65 0x06
+
+# CHECK: vpermpd (%rsi,%r10,4), %zmm2, %zmm8
+0x62 0x32 0xed 0x48 0x16 0x04 0x96
+
+# CHECK: vpbroadcastmw2d %k2, %zmm8
+0x62 0xd2 0x7e 0x48 0x3a 0xd0
+
+# CHECK: vpbroadcastq    (%r9,%rax), %zmm28
+0x62 0x42 0xfd 0x48 0x59 0x24 0x01
+
+# CHECK: vbroadcastss    %xmm0, %zmm1
+0x62 0xf2 0x7d 0x48 0x18 0xc8
+
+# CHECK: vextracti32x4   $4, %zmm0, (%r10)
+0x62 0xd3 0x7d 0x48 0x39 0x02 0x04
+
+# CHECK: vextracti32x4   $4, %zmm0, %xmm1
+0x62 0xf3 0x7d 0x48 0x39 0xc1 0x04
+
+# CHECK: vinserti32x4    $1, %xmm21, %zmm5, %zmm17
+0x62 0xa3 0x55 0x48 0x38 0xcd 0x01
+
+# CHECK: vmovaps %zmm21, %zmm5 {%k3}
+0x62 0xb1 0x7c 0x4b 0x28 0xed
+
+# CHECK: vgatherdps      (%rsi,%zmm0,4), %zmm1 {%k2}
+0x62 0xf2 0x7d 0x4a 0x92 0x0c 0x86
+
+# CHECK: vgatherdpd      (%rsi,%ymm0,4), %zmm1 {%k2}
+0x62 0xf2 0xfd 0x4a 0x92 0x0c 0x86
+
+#####################################################
+#                MASK INSTRUCTIONS                  #
+#####################################################
+
+# CHECK: kshiftlw        $3, %k1, %k2
+0xc4 0xe3 0xf9 0x32 0xd1 0x03
+
+# CHECK: kmovw   (%rdi), %k1
+0xc5 0xf8 0x90 0x0f
+
+# CHECK: kmovw   %k1, %eax
+0xc5 0xf8 0x93 0xc1
+
+# CHECK: kandw   %k1, %k2, %k3
+0xc5 0xec 0x41 0xd9
+
+# CHECK: kmovw   %k5, %k1
+0xc5 0xf8 0x90 0xcd
diff --git a/utils/TableGen/X86DisassemblerTables.cpp b/utils/TableGen/X86DisassemblerTables.cpp
index bdb4793185..7d2c9c7fe8 100644
--- a/utils/TableGen/X86DisassemblerTables.cpp
+++ b/utils/TableGen/X86DisassemblerTables.cpp
@@ -636,14 +636,36 @@ void DisassemblerTables::emitInstructionInfo(raw_ostream &o,
 }
 
 void DisassemblerTables::emitContextTable(raw_ostream &o, unsigned &i) const {
+  const unsigned int tableSize = 16384;
   o.indent(i * 2) << "static const uint8_t " CONTEXTS_STR
-                     "[256] = {\n";
+                     "[" << tableSize << "] = {\n";
   i++;
 
-  for (unsigned index = 0; index < 256; ++index) {
+  for (unsigned index = 0; index < tableSize; ++index) {
     o.indent(i * 2);
 
-    if ((index & ATTR_VEXL) && (index & ATTR_REXW) && (index & ATTR_OPSIZE))
+    if (index & ATTR_EVEX) {
+      o << "IC_EVEX";
+      if (index & ATTR_EVEXL2)
+        o << "_L2";
+      else if (index & ATTR_EVEXL)
+        o << "_L";
+      if (index & ATTR_REXW)
+        o << "_W";
+      if (index & ATTR_OPSIZE)
+        o << "_OPSIZE";
+      else if (index & ATTR_XD)
+        o << "_XD";
+      else if (index & ATTR_XS)
+        o << "_XS";
+      if (index & ATTR_EVEXKZ)
+        o << "_KZ";
+      else if (index & ATTR_EVEXK)
+        o << "_K";
+      if (index & ATTR_EVEXB)
+        o << "_B";
+    }
+    else if ((index & ATTR_VEXL) && (index & ATTR_REXW) && (index & ATTR_OPSIZE))
       o << "IC_VEX_L_W_OPSIZE";
     else if ((index & ATTR_VEXL) && (index & ATTR_REXW) && (index & ATTR_XD))
       o << "IC_VEX_L_W_XD";
@@ -713,7 +735,7 @@ void DisassemblerTables::emitContextTable(raw_ostream &o, unsigned &i) const {
     else
       o << "IC";
 
-    if (index < 255)
+    if (index < tableSize - 1)
       o << ",";
     else
       o << " ";