Add support for double-to-int & double-to-long in optimizing.

- Add support for the double-to-int and double-to-long Dex
  instructions in the optimizing compiler.
- Add S1 to the list of ARM FPU parameter registers so that
  a double value can be passed as parameter during a call
  to the runtime through D0.
- Have art::x86_64::X86_64Assembler::cvttsd2si work with
  64-bit operands.
- Generate x86, x86-64 and ARM (but not ARM64) code for
  double to int and double to long HTypeConversion nodes.
- Add related tests to test/422-type-conversion.

Change-Id: Ic93b9ec6630c26e940f7966a3346ad3fd5a2ab3a
diff --git a/compiler/optimizing/builder.cc b/compiler/optimizing/builder.cc
index 76efef0..fe32da0 100644
--- a/compiler/optimizing/builder.cc
+++ b/compiler/optimizing/builder.cc
@@ -1260,6 +1260,16 @@
       break;
     }
 
+    case Instruction::DOUBLE_TO_INT: {
+      Conversion_12x(instruction, Primitive::kPrimDouble, Primitive::kPrimInt, dex_pc);
+      break;
+    }
+
+    case Instruction::DOUBLE_TO_LONG: {
+      Conversion_12x(instruction, Primitive::kPrimDouble, Primitive::kPrimLong, dex_pc);
+      break;
+    }
+
     case Instruction::DOUBLE_TO_FLOAT: {
       Conversion_12x(instruction, Primitive::kPrimDouble, Primitive::kPrimFloat, dex_pc);
       break;
diff --git a/compiler/optimizing/code_generator_arm.cc b/compiler/optimizing/code_generator_arm.cc
index 5076c85..36af393 100644
--- a/compiler/optimizing/code_generator_arm.cc
+++ b/compiler/optimizing/code_generator_arm.cc
@@ -44,7 +44,7 @@
 static constexpr Register kRuntimeParameterCoreRegisters[] = { R0, R1, R2, R3 };
 static constexpr size_t kRuntimeParameterCoreRegistersLength =
     arraysize(kRuntimeParameterCoreRegisters);
-static constexpr SRegister kRuntimeParameterFpuRegisters[] = { S0 };
+static constexpr SRegister kRuntimeParameterFpuRegisters[] = { S0, S1 };
 static constexpr size_t kRuntimeParameterFpuRegistersLength =
     arraysize(kRuntimeParameterFpuRegisters);
 
@@ -1365,9 +1365,11 @@
   Primitive::Type input_type = conversion->GetInputType();
   DCHECK_NE(result_type, input_type);
 
-  // Float-to-long conversions invoke the runtime.
+  // The float-to-long and double-to-long type conversions rely on a
+  // call to the runtime.
   LocationSummary::CallKind call_kind =
-      (input_type == Primitive::kPrimFloat && result_type == Primitive::kPrimLong)
+      ((input_type == Primitive::kPrimFloat || input_type == Primitive::kPrimDouble)
+       && result_type == Primitive::kPrimLong)
       ? LocationSummary::kCall
       : LocationSummary::kNoCall;
   LocationSummary* locations =
@@ -1422,8 +1424,10 @@
           break;
 
         case Primitive::kPrimDouble:
-          LOG(FATAL) << "Type conversion from " << input_type
-                     << " to " << result_type << " not yet implemented";
+          // Processing a Dex `double-to-int' instruction.
+          locations->SetInAt(0, Location::RequiresFpuRegister());
+          locations->SetOut(Location::RequiresRegister());
+          locations->AddTemp(Location::RequiresFpuRegister());
           break;
 
         default:
@@ -1452,10 +1456,15 @@
           break;
         }
 
-        case Primitive::kPrimDouble:
-          LOG(FATAL) << "Type conversion from " << input_type << " to "
-                     << result_type << " not yet implemented";
+        case Primitive::kPrimDouble: {
+          // Processing a Dex `double-to-long' instruction.
+          InvokeRuntimeCallingConvention calling_convention;
+          locations->SetInAt(0, Location::FpuRegisterPairLocation(
+              calling_convention.GetFpuRegisterAt(0),
+              calling_convention.GetFpuRegisterAt(1)));
+          locations->SetOut(Location::RegisterPairLocation(R0, R1));
           break;
+        }
 
         default:
           LOG(FATAL) << "Unexpected type conversion from " << input_type
@@ -1614,10 +1623,15 @@
           break;
         }
 
-        case Primitive::kPrimDouble:
-          LOG(FATAL) << "Type conversion from " << input_type
-                     << " to " << result_type << " not yet implemented";
+        case Primitive::kPrimDouble: {
+          // Processing a Dex `double-to-int' instruction.
+          SRegister temp_s = locations->GetTemp(0).AsFpuRegisterPairLow<SRegister>();
+          DRegister temp_d = FromLowSToD(temp_s);
+          __ vmovd(temp_d, FromLowSToD(in.AsFpuRegisterPairLow<SRegister>()));
+          __ vcvtid(temp_s, temp_d);
+          __ vmovrs(out.AsRegister<Register>(), temp_s);
           break;
+        }
 
         default:
           LOG(FATAL) << "Unexpected type conversion from " << input_type
@@ -1643,15 +1657,16 @@
 
         case Primitive::kPrimFloat:
           // Processing a Dex `float-to-long' instruction.
-          // This call does not actually record PC information.
           codegen_->InvokeRuntime(QUICK_ENTRY_POINT(pF2l),
                                   conversion,
                                   conversion->GetDexPc());
           break;
 
         case Primitive::kPrimDouble:
-          LOG(FATAL) << "Type conversion from " << input_type << " to "
-                     << result_type << " not yet implemented";
+          // Processing a Dex `double-to-long' instruction.
+          codegen_->InvokeRuntime(QUICK_ENTRY_POINT(pD2l),
+                                  conversion,
+                                  conversion->GetDexPc());
           break;
 
         default:
diff --git a/compiler/optimizing/code_generator_x86.cc b/compiler/optimizing/code_generator_x86.cc
index 2aa121d..2fd712f 100644
--- a/compiler/optimizing/code_generator_x86.cc
+++ b/compiler/optimizing/code_generator_x86.cc
@@ -1330,9 +1330,11 @@
   Primitive::Type input_type = conversion->GetInputType();
   DCHECK_NE(result_type, input_type);
 
-  // Float-to-long conversions invoke the runtime.
+  // The float-to-long and double-to-long type conversions rely on a
+  // call to the runtime.
   LocationSummary::CallKind call_kind =
-      (input_type == Primitive::kPrimFloat && result_type == Primitive::kPrimLong)
+      ((input_type == Primitive::kPrimFloat || input_type == Primitive::kPrimDouble)
+       && result_type == Primitive::kPrimLong)
       ? LocationSummary::kCall
       : LocationSummary::kNoCall;
   LocationSummary* locations =
@@ -1387,8 +1389,10 @@
           break;
 
         case Primitive::kPrimDouble:
-          LOG(FATAL) << "Type conversion from " << input_type
-                     << " to " << result_type << " not yet implemented";
+          // Processing a Dex `double-to-int' instruction.
+          locations->SetInAt(0, Location::RequiresFpuRegister());
+          locations->SetOut(Location::RequiresRegister());
+          locations->AddTemp(Location::RequiresFpuRegister());
           break;
 
         default:
@@ -1411,15 +1415,27 @@
         case Primitive::kPrimFloat: {
           // Processing a Dex `float-to-long' instruction.
           InvokeRuntimeCallingConvention calling_convention;
-          locations->SetInAt(0, Location::RegisterLocation(calling_convention.GetRegisterAt(0)));
+          // Note that on x86 floating-point parameters are passed
+          // through core registers (here, EAX).
+          locations->SetInAt(0, Location::RegisterLocation(
+              calling_convention.GetRegisterAt(0)));
           // The runtime helper puts the result in EAX, EDX.
           locations->SetOut(Location::RegisterPairLocation(EAX, EDX));
           break;
         }
 
-        case Primitive::kPrimDouble:
-          LOG(FATAL) << "Type conversion from " << input_type << " to "
-                     << result_type << " not yet implemented";
+        case Primitive::kPrimDouble: {
+          // Processing a Dex `double-to-long' instruction.
+          InvokeRuntimeCallingConvention calling_convention;
+          // Note that on x86 floating-point parameters are passed
+          // through core registers (here, EAX and ECX).
+          locations->SetInAt(0, Location::RegisterPairLocation(
+              calling_convention.GetRegisterAt(0),
+              calling_convention.GetRegisterAt(1)));
+          // The runtime helper puts the result in EAX, EDX.
+          locations->SetOut(Location::RegisterPairLocation(EAX, EDX));
+          break;
+        }
           break;
 
         default:
@@ -1607,10 +1623,30 @@
           break;
         }
 
-        case Primitive::kPrimDouble:
-          LOG(FATAL) << "Type conversion from " << input_type
-                     << " to " << result_type << " not yet implemented";
+        case Primitive::kPrimDouble: {
+          // Processing a Dex `double-to-int' instruction.
+          XmmRegister input = in.AsFpuRegister<XmmRegister>();
+          Register output = out.AsRegister<Register>();
+          XmmRegister temp = locations->GetTemp(0).AsFpuRegister<XmmRegister>();
+          Label done, nan;
+
+          __ movl(output, Immediate(kPrimIntMax));
+          // temp = int-to-double(output)
+          __ cvtsi2sd(temp, output);
+          // if input >= temp goto done
+          __ comisd(input, temp);
+          __ j(kAboveEqual, &done);
+          // if input == NaN goto nan
+          __ j(kUnordered, &nan);
+          // output = double-to-int-truncate(input)
+          __ cvttsd2si(output, input);
+          __ jmp(&done);
+          __ Bind(&nan);
+          //  output = 0
+          __ xorl(output, output);
+          __ Bind(&done);
           break;
+        }
 
         default:
           LOG(FATAL) << "Unexpected type conversion from " << input_type
@@ -1634,13 +1670,13 @@
         case Primitive::kPrimFloat:
           // Processing a Dex `float-to-long' instruction.
           __ fs()->call(Address::Absolute(QUICK_ENTRYPOINT_OFFSET(kX86WordSize, pF2l)));
-          // This call does not actually record PC information.
           codegen_->RecordPcInfo(conversion, conversion->GetDexPc());
           break;
 
         case Primitive::kPrimDouble:
-          LOG(FATAL) << "Type conversion from " << input_type << " to "
-                     << result_type << " not yet implemented";
+          // Processing a Dex `double-to-long' instruction.
+          __ fs()->call(Address::Absolute(QUICK_ENTRYPOINT_OFFSET(kX86WordSize, pD2l)));
+          codegen_->RecordPcInfo(conversion, conversion->GetDexPc());
           break;
 
         default:
diff --git a/compiler/optimizing/code_generator_x86_64.cc b/compiler/optimizing/code_generator_x86_64.cc
index 5761fb1..39a9766 100644
--- a/compiler/optimizing/code_generator_x86_64.cc
+++ b/compiler/optimizing/code_generator_x86_64.cc
@@ -1370,8 +1370,10 @@
           break;
 
         case Primitive::kPrimDouble:
-          LOG(FATAL) << "Type conversion from " << input_type
-                     << " to " << result_type << " not yet implemented";
+          // Processing a Dex `double-to-int' instruction.
+          locations->SetInAt(0, Location::RequiresFpuRegister());
+          locations->SetOut(Location::RequiresRegister());
+          locations->AddTemp(Location::RequiresFpuRegister());
           break;
 
         default:
@@ -1401,8 +1403,10 @@
           break;
 
         case Primitive::kPrimDouble:
-          LOG(FATAL) << "Type conversion from " << input_type << " to "
-                     << result_type << " not yet implemented";
+          // Processing a Dex `double-to-long' instruction.
+          locations->SetInAt(0, Location::RequiresFpuRegister());
+          locations->SetOut(Location::RequiresRegister());
+          locations->AddTemp(Location::RequiresFpuRegister());
           break;
 
         default:
@@ -1589,10 +1593,30 @@
           break;
         }
 
-        case Primitive::kPrimDouble:
-          LOG(FATAL) << "Type conversion from " << input_type
-                     << " to " << result_type << " not yet implemented";
+        case Primitive::kPrimDouble: {
+          // Processing a Dex `double-to-int' instruction.
+          XmmRegister input = in.AsFpuRegister<XmmRegister>();
+          CpuRegister output = out.AsRegister<CpuRegister>();
+          XmmRegister temp = locations->GetTemp(0).AsFpuRegister<XmmRegister>();
+          Label done, nan;
+
+          __ movl(output, Immediate(kPrimIntMax));
+          // temp = int-to-double(output)
+          __ cvtsi2sd(temp, output);
+          // if input >= temp goto done
+          __ comisd(input, temp);
+          __ j(kAboveEqual, &done);
+          // if input == NaN goto nan
+          __ j(kUnordered, &nan);
+          // output = double-to-int-truncate(input)
+          __ cvttsd2si(output, input);
+          __ jmp(&done);
+          __ Bind(&nan);
+          //  output = 0
+          __ xorl(output, output);
+          __ Bind(&done);
           break;
+        }
 
         default:
           LOG(FATAL) << "Unexpected type conversion from " << input_type
@@ -1620,14 +1644,14 @@
           Label done, nan;
 
           __ movq(output, Immediate(kPrimLongMax));
-          // temp = int-to-float(output)
+          // temp = long-to-float(output)
           __ cvtsi2ss(temp, output, true);
           // if input >= temp goto done
           __ comiss(input, temp);
           __ j(kAboveEqual, &done);
           // if input == NaN goto nan
           __ j(kUnordered, &nan);
-          // output = float-to-int-truncate(input)
+          // output = float-to-long-truncate(input)
           __ cvttss2si(output, input, true);
           __ jmp(&done);
           __ Bind(&nan);
@@ -1637,10 +1661,30 @@
           break;
         }
 
-        case Primitive::kPrimDouble:
-          LOG(FATAL) << "Type conversion from " << input_type << " to "
-                     << result_type << " not yet implemented";
+        case Primitive::kPrimDouble: {
+          // Processing a Dex `double-to-long' instruction.
+          XmmRegister input = in.AsFpuRegister<XmmRegister>();
+          CpuRegister output = out.AsRegister<CpuRegister>();
+          XmmRegister temp = locations->GetTemp(0).AsFpuRegister<XmmRegister>();
+          Label done, nan;
+
+          __ movq(output, Immediate(kPrimLongMax));
+          // temp = long-to-double(output)
+          __ cvtsi2sd(temp, output, true);
+          // if input >= temp goto done
+          __ comisd(input, temp);
+          __ j(kAboveEqual, &done);
+          // if input == NaN goto nan
+          __ j(kUnordered, &nan);
+          // output = double-to-long-truncate(input)
+          __ cvttsd2si(output, input, true);
+          __ jmp(&done);
+          __ Bind(&nan);
+          //  output = 0
+          __ xorq(output, output);
+          __ Bind(&done);
           break;
+        }
 
         default:
           LOG(FATAL) << "Unexpected type conversion from " << input_type