http://codereview.chromium.org/3068001/diff/40001/41007
File src/arm/codegen-arm.cc (right):

http://codereview.chromium.org/3068001/diff/40001/41007#newcode794
src/arm/codegen-arm.cc:794: true_target->Branch(ne);
These branches are not needed.  Just fall through to the cc_reg_ = ne
which instructs the code generator that the answer is in the status
flags.

http://codereview.chromium.org/3068001/diff/40001/41007#newcode7778
src/arm/codegen-arm.cc:7778: __ b(&true_result, gt);
You can just do Ret(gt) since you know that r0 is non-zero.

http://codereview.chromium.org/3068001/diff/40001/41007#newcode7784
src/arm/codegen-arm.cc:7784: __ b(&true_result, gt);
And here.

http://codereview.chromium.org/3068001/diff/40001/41007#newcode7787
src/arm/codegen-arm.cc:7787: __ ldr(scratch0, FieldMemOperand(tos_,
String::kLengthOffset));
You can just load the length into r0 and return here.

http://codereview.chromium.org/3068001/diff/40001/41002
File src/objects.cc (right):

http://codereview.chromium.org/3068001/diff/40001/41002#newcode115
src/objects.cc:115: Object* Object::ToBooleanSlowPart() {
This is only called for heap numbers and only on non-VFP3 chips.  If we
care about non-VFP3 performance (we don't much) we should just do the
work in the stub.  If we don't care we could just use the existing
HeapNumber::ToBoolean below.

http://codereview.chromium.org/3068001/show

--
v8-dev mailing list
[email protected]
http://groups.google.com/group/v8-dev

Reply via email to