ARM: Add support for DoMathAbs with double inputs....

src/arm/lithium-codegen-arm.cc

 // Copyright 2011 the V8 project authors. All rights reserved.
 // Redistribution and use in source and binary forms, with or without
 // modification, are permitted provided that the following conditions are
 // met:
 //
 //     * Redistributions of source code must retain the above copyright
 //       notice, this list of conditions and the following disclaimer.
 //     * Redistributions in binary form must reproduce the above
 //       copyright notice, this list of conditions and the following
 //       disclaimer in the documentation and/or other materials provided
@@ skipping 1165 matching lines @@
   Register left = ToRegister(instr->InputAt(0));
   Register right = ToRegister(instr->InputAt(1));
 
   __ PushSafepointRegistersAndDoubles();
   GenericBinaryOpStub stub(op, OVERWRITE_LEFT, left, right);
   __ CallStub(&stub);
   RecordSafepointWithRegistersAndDoubles(instr->pointer_map(),
                                          0,
                                          Safepoint::kNoDeoptimizationIndex);
   // Overwrite the stored value of r0 with the result of the stub.
-  __ str(r0, MemOperand(sp, DwVfpRegister::kNumAllocatableRegisters *
-                        kDoubleSize));
+  __ StoreToSafepointRegistersAndDoublesSlot(r0);
   __ PopSafepointRegistersAndDoubles();
 }
 
 
 void LCodeGen::DoMulI(LMulI* instr) {
   Register scratch = scratch0();
   Register left = ToRegister(instr->InputAt(0));
   Register right = EmitLoadRegister(instr->InputAt(1), scratch);
 
   if (instr->hydrogen()->CheckFlag(HValue::kBailoutOnMinusZero) &&
@@ skipping 1302 matching lines @@
 
 
 void LCodeGen::DoCallConstantFunction(LCallConstantFunction* instr) {
   ASSERT(ToRegister(instr->result()).is(r0));
   __ mov(r1, Operand(instr->function()));
   CallKnownFunction(instr->function(), instr->arity(), instr);
 }
 
 
 void LCodeGen::DoDeferredMathAbsTaggedHeapNumber(LUnaryMathOperation* instr) {
+  ASSERT(instr->InputAt(0)->Equals(instr->result()));
   Register input = ToRegister(instr->InputAt(0));
   Register scratch = scratch0();
 
   // Deoptimize if not a heap number.
   __ ldr(scratch, FieldMemOperand(input, HeapObject::kMapOffset));
   __ LoadRoot(ip, Heap::kHeapNumberMapRootIndex);
   __ cmp(scratch, Operand(ip));
   DeoptimizeIf(ne, instr->environment());
 
   Label done;
-
-  Label negative;
-  __ ldr(scratch, FieldMemOperand(input, HeapNumber::kExponentOffset));
+  Register exponent = scratch0();
+  scratch = no_reg;
+  __ ldr(exponent, FieldMemOperand(input, HeapNumber::kExponentOffset));
   // Check the sign of the argument. If the argument is positive, just
   // return it. We do not need to patch the stack since |input| and
-  // |result| are the same register and |input| will be restored
+  // |result| are the same register and |input| would be restored
   // unchanged by popping safepoint registers.
-  __ tst(scratch, Operand(HeapNumber::kSignMask));
-  __ b(ne, &negative);
-  __ jmp(&done);
+  __ tst(exponent, Operand(HeapNumber::kSignMask));
+  __ b(eq, &done);
 
-  __ bind(&negative);
+  // Input is negative. Reverse its sign.
   // Preserve the value of all registers.
   __ PushSafepointRegisters();
 
-  Register tmp = input.is(r0) ? r1 : r0;
+  // Registers were saved at the safepoint, so we can use
+  // many scratch registers.
+  Register tmp1 = input.is(r0) ? r1 : r0;

[Mads Ager (chromium), 2011/01/27 08:27:56]

Kevin had a nicer suggestion for these. We intend to use r1,r2,r3,r4 for
tmp1,tmp2,tmp3,tmp4. Input can only be one of them and if it is we will use r0
instead.

Register tmp1 = input.is(r1) ? r0 : r1;
Register tmp2 = input.is(r2) ? r0 : r2;
Register tmp3 = input.is(r3) ? r0 : r3;

[Alexandre, 2011/01/28 13:57:59]

I like that! Done.
On 2011/01/27 08:27:56, Mads Ager wrote:

   Register tmp2 = input.is(r2) ? r3 : r2;
   Register tmp3 = input.is(r4) ? r5 : r4;
+  Register tmp4 = input.is(r6) ? r7 : r6;
+
+  // exponent: floating point exponent value.
 
   Label allocated, slow;
-  __ LoadRoot(scratch, Heap::kHeapNumberMapRootIndex);
-  __ AllocateHeapNumber(tmp, tmp2, tmp3, scratch, &slow);
+  __ LoadRoot(tmp4, Heap::kHeapNumberMapRootIndex);
+  __ AllocateHeapNumber(tmp1, tmp2, tmp3, tmp4, &slow);
   __ b(&allocated);
 
   // Slow case: Call the runtime system to do the number allocation.
   __ bind(&slow);
 
   __ CallRuntimeSaveDoubles(Runtime::kAllocateHeapNumber);
   RecordSafepointWithRegisters(
       instr->pointer_map(), 0, Safepoint::kNoDeoptimizationIndex);
   // Set the pointer to the new heap number in tmp.
-  if (!tmp.is(r0)) __ mov(tmp, Operand(r0));
-
+  if (!tmp1.is(r0)) __ mov(tmp1, Operand(r0));
   // Restore input_reg after call to runtime.
-  MemOperand input_register_slot = masm()->SafepointRegisterSlot(input);
-  __ ldr(input, input_register_slot);
+  __ LoadFromSafepointRegisterSlot(input);
+  __ ldr(exponent, FieldMemOperand(input, HeapNumber::kExponentOffset));
 
   __ bind(&allocated);
-  __ ldr(tmp2, FieldMemOperand(input, HeapNumber::kExponentOffset));
-  __ bic(tmp2, tmp2, Operand(HeapNumber::kSignMask));
-  __ str(tmp2, FieldMemOperand(tmp, HeapNumber::kExponentOffset));
+  // exponent: floating point exponent value.
+  // tmp1: allocated heap number.
+  __ bic(exponent, exponent, Operand(HeapNumber::kSignMask));
+  __ str(exponent, FieldMemOperand(tmp1, HeapNumber::kExponentOffset));
   __ ldr(tmp2, FieldMemOperand(input, HeapNumber::kMantissaOffset));
-  __ str(tmp2, FieldMemOperand(tmp, HeapNumber::kMantissaOffset));
+  __ str(tmp2, FieldMemOperand(tmp1, HeapNumber::kMantissaOffset));
 
-  __ str(tmp, input_register_slot);
+  __ str(tmp1, masm()->SafepointRegisterSlot(input));
   __ PopSafepointRegisters();
 
   __ bind(&done);
 }
 
 
 void LCodeGen::EmitIntegerMathAbs(LUnaryMathOperation* instr) {
-  Label is_positive;
-  uint32_t kSignMask = 0x80000000u;
   Register input = ToRegister(instr->InputAt(0));
-  __ tst(input, Operand(kSignMask));
-  __ b(eq, &is_positive);
-  __ rsb(input, input, Operand(0), SetCC);
+  __ cmp(input, Operand(0));
+  // We can make rsb conditional because the previous cmp instruction
+  // will clear the V (overflow) flag and rsb won't set this flag
+  // if input is positive.
+  __ rsb(input, input, Operand(0), SetCC, mi);
   // Deoptimize on overflow.
   DeoptimizeIf(vs, instr->environment());
-  __ bind(&is_positive);
 }
 
 
 void LCodeGen::DoMathAbs(LUnaryMathOperation* instr) {
   // Class for deferred case.
   class DeferredMathAbsTaggedHeapNumber: public LDeferredCode {
    public:
     DeferredMathAbsTaggedHeapNumber(LCodeGen* codegen,
                                     LUnaryMathOperation* instr)
         : LDeferredCode(codegen), instr_(instr) { }
     virtual void Generate() {
       codegen()->DoDeferredMathAbsTaggedHeapNumber(instr_);
     }
    private:
     LUnaryMathOperation* instr_;
   };
 
   ASSERT(instr->InputAt(0)->Equals(instr->result()));
   Representation r = instr->hydrogen()->value()->representation();
   if (r.IsDouble()) {
     DwVfpRegister input = ToDoubleRegister(instr->InputAt(0));
-    // __ vabs(input, input);
-    Abort("Double DoMathAbs unimplemented");
+    __ vabs(input, input);
   } else if (r.IsInteger32()) {
     EmitIntegerMathAbs(instr);
   } else {
     // Representation is tagged.
     DeferredMathAbsTaggedHeapNumber* deferred =
         new DeferredMathAbsTaggedHeapNumber(this, instr);
     Register input = ToRegister(instr->InputAt(0));
     // Smi check.
     __ JumpIfNotSmi(input, deferred->entry());
     // If smi, handle it directly.
@@ skipping 1056 matching lines @@
 
 
 void LCodeGen::DoOsrEntry(LOsrEntry* instr) {
   Abort("DoOsrEntry unimplemented.");
 }
 
 
 #undef __
 
 } }  // namespace v8::internal