Switch the fixed-length disassembler to be table-driven.

Refactor the TableGen'erated fixed length disassemblmer to use a
table-driven state machine rather than a massive set of nested
switch() statements.

As a result, the ARM Disassembler (ARMDisassembler.cpp) builds much more
quickly and generates a smaller end result. For a Release+Asserts build on
a 16GB 3.4GHz i7 iMac w/ SSD:

Time to compile at -O2 (averaged w/ hot caches):
  Previous: 35.5s
  New:       8.9s

TEXT size:
  Previous: 447,251
  New:      297,661

Builds in 25% of the time previously required and generates code 66% of
the size.

Execution time of the disassembler is only slightly slower (7% disassembling
10 million ARM instructions, 19.6s vs 21.0s). The new implementation has
not yet been tuned, however, so the performance should almost certainly
be recoverable should it become a concern.

llvm-svn: 161888

include/llvm/MC/MCFixedLenDisassembler.h

@@ -0,0 +1,32 @@
+//===-- llvm/MC/MCFixedLenDisassembler.h - Decoder driver -------*- C++ -*-===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+// Fixed length disassembler decoder state machine driver.
+//===----------------------------------------------------------------------===//
+#ifndef MCFIXEDLENDISASSEMBLER_H
+#define MCFIXEDLENDISASSEMBLER_H
+
+namespace llvm {
+
+namespace MCD {
+// Disassembler state machine opcodes.
+enum DecoderOps {
+  OPC_ExtractField = 1, // OPC_ExtractField(uint8_t Start, uint8_t Len)
+  OPC_FilterValue,      // OPC_FilterValue(uleb128 Val, uint16_t NumToSkip)
+  OPC_CheckField,       // OPC_CheckField(uint8_t Start, uint8_t Len,
+                        //                uleb128 Val, uint16_t NumToSkip)
+  OPC_CheckPredicate,   // OPC_CheckPredicate(uleb128 PIdx, uint16_t NumToSkip)
+  OPC_Decode,           // OPC_Decode(uleb128 Opcode, uleb128 DIdx)
+  OPC_SoftFail,         // OPC_SoftFail(uleb128 PMask, uleb128 NMask)
+  OPC_Fail              // OPC_Fail()
+};
+
+} // namespace MCDecode
+} // namespace llvm
+
+#endif

include/llvm/Support/LEB128.h

@@ -19,7 +19,7 @@
 
 namespace llvm {
 
-/// Utility function to encode a SLEB128 value.
+/// Utility function to encode a SLEB128 value to an output stream.
 static inline void encodeSLEB128(int64_t Value, raw_ostream &OS) {
   bool More;
   do {
@@ -34,7 +34,7 @@ static inline void encodeSLEB128(int64_t Value, raw_ostream &OS) {
   } while (More);
 }
 
-/// Utility function to encode a ULEB128 value.
+/// Utility function to encode a ULEB128 value to an output stream.
 static inline void encodeULEB128(uint64_t Value, raw_ostream &OS,
                                  unsigned Padding = 0) {
   do {
@@ -53,6 +53,43 @@ static inline void encodeULEB128(uint64_t Value, raw_ostream &OS,
   }
 }
 
+/// Utility function to encode a ULEB128 value to a buffer. Returns
+/// the length in bytes of the encoded value.
+static inline unsigned encodeULEB128(uint64_t Value, uint8_t *p,
+                                     unsigned Padding = 0) {
+  uint8_t *orig_p = p;
+  do {
+    uint8_t Byte = Value & 0x7f;
+    Value >>= 7;
+    if (Value != 0 || Padding != 0)
+      Byte |= 0x80; // Mark this byte that that more bytes will follow.
+    *p++ = Byte;
+  } while (Value != 0);
+
+  // Pad with 0x80 and emit a null byte at the end.
+  if (Padding != 0) {
+    for (; Padding != 1; --Padding)
+      *p++ = '\x80';
+    *p++ = '\x00';
+  }
+  return (unsigned)(p - orig_p);
+}
+
+
+/// Utility function to decode a ULEB128 value.
+static inline uint64_t decodeULEB128(const uint8_t *p, unsigned *n = 0) {
+  const uint8_t *orig_p = p;
+  uint64_t Value = 0;
+  unsigned Shift = 0;
+  do {
+    Value += (*p & 0x7f) << Shift;
+    Shift += 7;
+  } while (*p++ >= 128);
+  if (n)
+    *n = (unsigned)(p - orig_p);
+  return Value;
+}
+
 }  // namespace llvm
 
 #endif  // LLVM_SYSTEM_LEB128_H

lib/Target/Mips/Disassembler/MipsDisassembler.cpp

@@ -16,6 +16,7 @@
 #include "MipsRegisterInfo.h"
 #include "llvm/MC/EDInstInfo.h"
 #include "llvm/MC/MCDisassembler.h"
+#include "llvm/MC/MCFixedLenDisassembler.h"
 #include "llvm/Support/MemoryObject.h"
 #include "llvm/Support/TargetRegistry.h"
 #include "llvm/MC/MCSubtargetInfo.h"
@@ -274,7 +275,8 @@ MipsDisassembler::getInstruction(MCInst &instr,
     return MCDisassembler::Fail;
 
   // Calling the auto-generated decoder function.
-  Result = decodeMipsInstruction32(instr, Insn, Address, this, STI);
+  Result = decodeInstruction(DecoderTableMips32, instr, Insn, Address,
+                             this, STI);
   if (Result != MCDisassembler::Fail) {
     Size = 4;
     return Result;
@@ -298,13 +300,15 @@ Mips64Disassembler::getInstruction(MCInst &instr,
     return MCDisassembler::Fail;
 
   // Calling the auto-generated decoder function.
-  Result = decodeMips64Instruction32(instr, Insn, Address, this, STI);
+  Result = decodeInstruction(DecoderTableMips6432, instr, Insn, Address,
+                             this, STI);
   if (Result != MCDisassembler::Fail) {
     Size = 4;
     return Result;
   }
   // If we fail to decode in Mips64 decoder space we can try in Mips32
-  Result = decodeMipsInstruction32(instr, Insn, Address, this, STI);
+  Result = decodeInstruction(DecoderTableMips32, instr, Insn, Address,
+                             this, STI);
   if (Result != MCDisassembler::Fail) {
     Size = 4;
     return Result;
@@ -379,8 +383,8 @@ static DecodeStatus DecodeMem(MCInst &Inst,
                               uint64_t Address,
                               const void *Decoder) {
   int Offset = SignExtend32<16>(Insn & 0xffff);
-  unsigned Reg = fieldFromInstruction32(Insn, 16, 5);
-  unsigned Base = fieldFromInstruction32(Insn, 21, 5);
+  unsigned Reg = fieldFromInstruction(Insn, 16, 5);
+  unsigned Base = fieldFromInstruction(Insn, 21, 5);
 
   Reg = getReg(Decoder, Mips::CPURegsRegClassID, Reg);
   Base = getReg(Decoder, Mips::CPURegsRegClassID, Base);
@@ -401,8 +405,8 @@ static DecodeStatus DecodeFMem(MCInst &Inst,
                                uint64_t Address,
                                const void *Decoder) {
   int Offset = SignExtend32<16>(Insn & 0xffff);
-  unsigned Reg = fieldFromInstruction32(Insn, 16, 5);
-  unsigned Base = fieldFromInstruction32(Insn, 21, 5);
+  unsigned Reg = fieldFromInstruction(Insn, 16, 5);
+  unsigned Base = fieldFromInstruction(Insn, 21, 5);
 
   Reg = getReg(Decoder, Mips::FGR64RegClassID, Reg);
   Base = getReg(Decoder, Mips::CPURegsRegClassID, Base);
@@ -484,7 +488,7 @@ static DecodeStatus DecodeJumpTarget(MCInst &Inst,
                                      uint64_t Address,
                                      const void *Decoder) {
 
-  unsigned JumpOffset = fieldFromInstruction32(Insn, 0, 26) << 2;
+  unsigned JumpOffset = fieldFromInstruction(Insn, 0, 26) << 2;
   Inst.addOperand(MCOperand::CreateImm(JumpOffset));
   return MCDisassembler::Success;
 }