Merge "Implement the x86 initialize-type...from-code entry points." into ics-mr1-plus-art
diff --git a/src/compiler/codegen/x86/Assemble.cc b/src/compiler/codegen/x86/Assemble.cc
index fc9878e..d9cdabf 100644
--- a/src/compiler/codegen/x86/Assemble.cc
+++ b/src/compiler/codegen/x86/Assemble.cc
@@ -130,16 +130,16 @@
   0x81, 0x7 /* RegMem32/imm32 */, 0x83, 0x7 /* RegMem32/imm8 */),
 #undef ENCODING_MAP
 
-  { kX86Imul16RRI,   kRegRegImm,             IS_TERTIARY_OP | SETS_CCODES, { 0x66, 0, 0x69, 0, 0, 0, 0, 2 }, "Imul16RRI", "" },
-  { kX86Imul16RMI,   kRegMemImm,   IS_LOAD | IS_QUAD_OP     | SETS_CCODES, { 0x66, 0, 0x69, 0, 0, 0, 0, 2 }, "Imul16RMI", "" },
-  { kX86Imul16RAI,   kRegArrayImm, IS_LOAD | IS_SEXTUPLE_OP | SETS_CCODES, { 0x66, 0, 0x69, 0, 0, 0, 0, 2 }, "Imul16RAI", "" },
+  { kX86Imul16RRI,   kRegRegImm,             IS_TERTIARY_OP | SETS_CCODES, { 0x66, 0, 0x69, 0, 0, 0, 0, 2 }, "Imul16RRI", "!0r,!1r,!2d" },
+  { kX86Imul16RMI,   kRegMemImm,   IS_LOAD | IS_QUAD_OP     | SETS_CCODES, { 0x66, 0, 0x69, 0, 0, 0, 0, 2 }, "Imul16RMI", "!0r,[!1r+!2d],!3d" },
+  { kX86Imul16RAI,   kRegArrayImm, IS_LOAD | IS_SEXTUPLE_OP | SETS_CCODES, { 0x66, 0, 0x69, 0, 0, 0, 0, 2 }, "Imul16RAI", "!0r,[!1r+!2r<<!3d+!4d],!5d" },
 
-  { kX86Imul32RRI,   kRegRegImm,             IS_TERTIARY_OP | SETS_CCODES, { 0, 0, 0x69, 0, 0, 0, 0, 2 }, "Imul32RRI", "" },
-  { kX86Imul32RMI,   kRegMemImm,   IS_LOAD | IS_QUAD_OP     | SETS_CCODES, { 0, 0, 0x69, 0, 0, 0, 0, 2 }, "Imul32RMI", "" },
-  { kX86Imul32RAI,   kRegArrayImm, IS_LOAD | IS_SEXTUPLE_OP | SETS_CCODES, { 0, 0, 0x69, 0, 0, 0, 0, 2 }, "Imul32RAI", "" },
-  { kX86Imul32RRI8,  kRegRegImm,             IS_TERTIARY_OP | SETS_CCODES, { 0, 0, 0x6B, 0, 0, 0, 0, 1 }, "Imul32RRI8", "" },
-  { kX86Imul32RMI8,  kRegMemImm,   IS_LOAD | IS_QUAD_OP     | SETS_CCODES, { 0, 0, 0x6B, 0, 0, 0, 0, 1 }, "Imul32RMI8", "" },
-  { kX86Imul32RAI8,  kRegArrayImm, IS_LOAD | IS_SEXTUPLE_OP | SETS_CCODES, { 0, 0, 0x6B, 0, 0, 0, 0, 1 }, "Imul32RAI8", "" },
+  { kX86Imul32RRI,   kRegRegImm,             IS_TERTIARY_OP | SETS_CCODES, { 0, 0, 0x69, 0, 0, 0, 0, 2 }, "Imul32RRI", "!0r,!1r,!2d" },
+  { kX86Imul32RMI,   kRegMemImm,   IS_LOAD | IS_QUAD_OP     | SETS_CCODES, { 0, 0, 0x69, 0, 0, 0, 0, 2 }, "Imul32RMI", "!0r,[!1r+!2d],!3d" },
+  { kX86Imul32RAI,   kRegArrayImm, IS_LOAD | IS_SEXTUPLE_OP | SETS_CCODES, { 0, 0, 0x69, 0, 0, 0, 0, 2 }, "Imul32RAI", "!0r,[!1r+!2r<<!3d+!4d],!5d" },
+  { kX86Imul32RRI8,  kRegRegImm,             IS_TERTIARY_OP | SETS_CCODES, { 0, 0, 0x6B, 0, 0, 0, 0, 1 }, "Imul32RRI8", "!0r,!1r,!2d" },
+  { kX86Imul32RMI8,  kRegMemImm,   IS_LOAD | IS_QUAD_OP     | SETS_CCODES, { 0, 0, 0x6B, 0, 0, 0, 0, 1 }, "Imul32RMI8", "!0r,[!1r+!2d],!3d" },
+  { kX86Imul32RAI8,  kRegArrayImm, IS_LOAD | IS_SEXTUPLE_OP | SETS_CCODES, { 0, 0, 0x6B, 0, 0, 0, 0, 1 }, "Imul32RAI8", "!0r,[!1r+!2r<<!3d+!4d],!5d" },
 
   { kX86Mov8MR, kMemReg,    IS_STORE | IS_TERTIARY_OP, { 0, 0,             0x88, 0, 0, 0, 0, 0 }, "Mov8MR", "[!0r+!1d],!2r" },
   { kX86Mov8AR, kArrayReg,  IS_STORE | IS_QUIN_OP,     { 0, 0,             0x88, 0, 0, 0, 0, 0 }, "Mov8AR", "[!0r+!1r<<!2d+!3d],!4r" },
@@ -707,6 +707,63 @@
   DCHECK_EQ(0, entry->skeleton.immediate_bytes);
 }
 
+static void emitRegRegImm(CompilationUnit* cUnit, const X86EncodingMap* entry,
+                          uint8_t reg1, uint8_t reg2, int32_t imm) {
+  if (entry->skeleton.prefix1 != 0) {
+    cUnit->codeBuffer.push_back(entry->skeleton.prefix1);
+    if (entry->skeleton.prefix2 != 0) {
+      cUnit->codeBuffer.push_back(entry->skeleton.prefix2);
+    }
+  } else {
+    DCHECK_EQ(0, entry->skeleton.prefix2);
+  }
+  cUnit->codeBuffer.push_back(entry->skeleton.opcode);
+  if (entry->skeleton.opcode == 0x0F) {
+    cUnit->codeBuffer.push_back(entry->skeleton.extra_opcode1);
+    if (entry->skeleton.extra_opcode1 == 0x38 || entry->skeleton.extra_opcode2 == 0x3A) {
+      cUnit->codeBuffer.push_back(entry->skeleton.extra_opcode2);
+    } else {
+      DCHECK_EQ(0, entry->skeleton.extra_opcode2);
+    }
+  } else {
+    DCHECK_EQ(0, entry->skeleton.extra_opcode1);
+    DCHECK_EQ(0, entry->skeleton.extra_opcode2);
+  }
+  if (FPREG(reg1)) {
+    reg1 = reg1 & FP_REG_MASK;
+  }
+  if (FPREG(reg2)) {
+    reg2 = reg2 & FP_REG_MASK;
+  }
+  DCHECK_LT(reg1, 8);
+  DCHECK_LT(reg2, 8);
+  uint8_t modrm = (3 << 6) | (reg1 << 3) | reg2;
+  cUnit->codeBuffer.push_back(modrm);
+  DCHECK_EQ(0, entry->skeleton.modrm_opcode);
+  DCHECK_EQ(0, entry->skeleton.ax_opcode);
+  switch (entry->skeleton.immediate_bytes) {
+    case 1:
+      DCHECK(IS_SIMM8(imm));
+      cUnit->codeBuffer.push_back(imm & 0xFF);
+      break;
+    case 2:
+      DCHECK(IS_SIMM16(imm));
+      cUnit->codeBuffer.push_back(imm & 0xFF);
+      cUnit->codeBuffer.push_back((imm >> 8) & 0xFF);
+      break;
+    case 4:
+      cUnit->codeBuffer.push_back(imm & 0xFF);
+      cUnit->codeBuffer.push_back((imm >> 8) & 0xFF);
+      cUnit->codeBuffer.push_back((imm >> 16) & 0xFF);
+      cUnit->codeBuffer.push_back((imm >> 24) & 0xFF);
+      break;
+    default:
+      LOG(FATAL) << "Unexpected immediate bytes (" << entry->skeleton.immediate_bytes
+                 << ") for instruction: " << entry->name;
+      break;
+  }
+}
+
 static void emitRegImm(CompilationUnit* cUnit, const X86EncodingMap* entry,
                        uint8_t reg, int imm) {
   if (entry->skeleton.prefix1 != 0) {
@@ -1063,7 +1120,7 @@
 }
 
 void emitUnimplemented(CompilationUnit* cUnit, const X86EncodingMap* entry, LIR* lir) {
-  UNIMPLEMENTED(WARNING) << "encoding for: " << entry->name;
+  UNIMPLEMENTED(WARNING) << "encoding kind for " << entry->name << " " << buildInsnString(entry->fmt, lir, 0);
   for (int i = 0; i < oatGetInsnSize(lir); ++i) {
     cUnit->codeBuffer.push_back(0xCC);  // push breakpoint instruction - int 3
   }
@@ -1239,6 +1296,9 @@
       case kRegReg:  // lir operands - 0: reg1, 1: reg2
         emitRegReg(cUnit, entry, lir->operands[0], lir->operands[1]);
         break;
+      case kRegRegImm:
+        emitRegRegImm(cUnit, entry, lir->operands[0], lir->operands[1], lir->operands[2]);
+        break;
       case kRegImm:  // lir operands - 0: reg, 1: immediate
         emitRegImm(cUnit, entry, lir->operands[0], lir->operands[1]);
         break;
diff --git a/src/compiler/codegen/x86/FP/X86FP.cc b/src/compiler/codegen/x86/FP/X86FP.cc
index 29aaeeb..fe0d149 100644
--- a/src/compiler/codegen/x86/FP/X86FP.cc
+++ b/src/compiler/codegen/x86/FP/X86FP.cc
@@ -16,8 +16,8 @@
 
 namespace art {
 
-bool genArithOpFloat(CompilationUnit *cUnit, MIR *mir, RegLocation rlDest,
-                     RegLocation rlSrc1, RegLocation rlSrc2) {
+static bool genArithOpFloat(CompilationUnit *cUnit, MIR *mir, RegLocation rlDest,
+                            RegLocation rlSrc1, RegLocation rlSrc2) {
   X86OpCode op = kX86Nop;
   RegLocation rlResult;
 
@@ -149,7 +149,7 @@
       // These are easy to implement inline except when the src is > MAX_INT/LONG the result
       // needs to be changed from 0x80000000 to 0x7FFFFFF (requires an in memory float/double
       // literal constant to compare against).
-      UNIMPLEMENTED(WARNING) << "inline [df]2i";
+      UNIMPLEMENTED(WARNING) << "inline [df]2i " << PrettyMethod(cUnit->method_idx, *cUnit->dex_file);
     case Instruction::LONG_TO_DOUBLE:
     case Instruction::FLOAT_TO_LONG:
     case Instruction::LONG_TO_FLOAT:
diff --git a/src/compiler/codegen/x86/X86/Gen.cc b/src/compiler/codegen/x86/X86/Gen.cc
index f2dbc11..ffc5952 100644
--- a/src/compiler/codegen/x86/X86/Gen.cc
+++ b/src/compiler/codegen/x86/X86/Gen.cc
@@ -168,7 +168,7 @@
 
 void genNegFloat(CompilationUnit *cUnit, RegLocation rlDest, RegLocation rlSrc)
 {
-    UNIMPLEMENTED(WARNING) << "genNegFloat";
+    UNIMPLEMENTED(WARNING) << "genNegFloat " << PrettyMethod(cUnit->method_idx, *cUnit->dex_file);
     newLIR0(cUnit, kX86Bkpt);
 #if 0
     RegLocation rlResult;
@@ -182,7 +182,7 @@
 
 void genNegDouble(CompilationUnit *cUnit, RegLocation rlDest, RegLocation rlSrc)
 {
-    UNIMPLEMENTED(WARNING) << "genNegDouble";
+    UNIMPLEMENTED(WARNING) << "genNegDouble" << PrettyMethod(cUnit->method_idx, *cUnit->dex_file);
     newLIR0(cUnit, kX86Bkpt);
 #if 0
     RegLocation rlResult;
diff --git a/src/disassembler_x86.cc b/src/disassembler_x86.cc
index eb52d5a..d2c860b 100644
--- a/src/disassembler_x86.cc
+++ b/src/disassembler_x86.cc
@@ -383,13 +383,14 @@
     reg_in_opcode = true;
     break;
   case 0xC0: case 0xC1:
+  case 0xD0: case 0xD1:
     static const char* shift_opcodes[] =
         {"rol", "ror", "rcl", "rcr", "shl", "shr", "unknown-shift", "sar"};
     modrm_opcodes = shift_opcodes;
     has_modrm = true;
     reg_is_opcode = true;
     store = true;
-    immediate_bytes = 1;
+    immediate_bytes = ((*instr & 0xf0) == 0xc0) ? 1 : 0;
     byte_operand = *instr == 0xC0;
     break;
   case 0xC3: opcode << "ret"; break;
diff --git a/src/oat/runtime/arm/runtime_support_arm.S b/src/oat/runtime/arm/runtime_support_arm.S
index e1323d6..2ef162e 100644
--- a/src/oat/runtime/arm/runtime_support_arm.S
+++ b/src/oat/runtime/arm/runtime_support_arm.S
@@ -369,7 +369,7 @@
     SETUP_REF_ONLY_CALLEE_SAVE_FRAME           @ save callee saves in case of GC
     mov    r2, r9                              @ pass Thread::Current
     mov    r3, sp                              @ pass SP
-    @ artInitializeTypeFromCode(uint32_t type_idx, Method* referrer, Thread*, SP)
+    @ artInitializeTypeAndVerifyAccessFromCode(uint32_t type_idx, Method* referrer, Thread*, SP)
     bl     artInitializeTypeAndVerifyAccessFromCode
     RESTORE_REF_ONLY_CALLEE_SAVE_FRAME
     cmp    r0, #0                              @ success if result is non-null
diff --git a/src/oat/runtime/support_math.cc b/src/oat/runtime/support_math.cc
index 7c6e0b1..cbffc2b 100644
--- a/src/oat/runtime/support_math.cc
+++ b/src/oat/runtime/support_math.cc
@@ -62,6 +62,10 @@
   return -1;
 }
 
+int64_t Lmul(int64_t a, int64_t b) {
+  return a * b;
+}
+
 extern "C" int64_t artLdivFromCode(int64_t a, int64_t b) {
   return a / b;
 }
diff --git a/src/oat/runtime/x86/oat_support_entrypoints_x86.cc b/src/oat/runtime/x86/oat_support_entrypoints_x86.cc
index 6720e3e..7ded2d1 100644
--- a/src/oat/runtime/x86/oat_support_entrypoints_x86.cc
+++ b/src/oat/runtime/x86/oat_support_entrypoints_x86.cc
@@ -71,6 +71,7 @@
 extern int32_t CmplDouble(double a, double b);
 extern int32_t CmpgFloat(float a, float b);
 extern int32_t CmplFloat(float a, float b);
+extern int64_t Lmul(int64_t a, int64_t b);
 extern "C" int32_t art_idiv_from_code(int32_t, int32_t);
 extern "C" int32_t art_idivmod_from_code(int32_t, int32_t);
 extern "C" int64_t art_ldiv_from_code(int64_t, int64_t);
@@ -163,31 +164,31 @@
   points->pCmpgFloat = CmpgFloat;
   points->pCmplDouble = CmplDouble;
   points->pCmplFloat = CmplFloat;
-  //points->pDadd = NULL;
-  //points->pDdiv = NULL;
-  //points->pDmul = NULL;
-  //points->pDsub = NULL;
+  //points->pDadd = NULL; // Not needed on x86.
+  //points->pDdiv = NULL; // Not needed on x86.
+  //points->pDmul = NULL; // Not needed on x86.
+  //points->pDsub = NULL; // Not needed on x86.
   //points->pF2d = NULL;
   //points->pFmod = NULL;
   //points->pI2d = NULL;
   //points->pL2d = NULL;
   //points->pD2f = NULL;
-  //points->pFadd = NULL;
-  //points->pFdiv = NULL;
+  //points->pFadd = NULL; // Not needed on x86.
+  //points->pFdiv = NULL; // Not needed on x86.
   //points->pFmodf = NULL;
-  //points->pFmul = NULL;
-  //points->pFsub = NULL;
+  //points->pFmul = NULL; // Not needed on x86.
+  //points->pFsub = NULL; // Not needed on x86.
   //points->pI2f = NULL;
   //points->pL2f = NULL;
-  //points->pD2iz = NULL;
-  //points->pF2iz = NULL;
+  points->pD2iz = D2I;
+  points->pF2iz = F2I;
   points->pIdiv = art_idiv_from_code;
   points->pIdivmod = art_idivmod_from_code;
   points->pD2l = D2L;
   points->pF2l = F2L;
   points->pLdiv = art_ldiv_from_code;
   points->pLdivmod = art_ldivmod_from_code;
-  //points->pLmul = NULL;
+  points->pLmul = Lmul;
   points->pShlLong = art_lshl_from_code;
   points->pShrLong = art_lshr_from_code;
   points->pUshrLong = art_lushr_from_code;