Handle both global and local variables potentially shadowed by...

Handle both global and local variables potentially shadowed by
eval-introduced variables in full-codegen.

Make sure that x64 assembler records source positions for calls.

Committed: http://code.google.com/p/v8/source/detail?r=5441

[Søren Thygesen Gjesse, 2010-09-10 07:44:24]

LGTM

http://codereview.chromium.org/3357022/diff/9001/7002
File src/arm/full-codegen-arm.cc (right):

http://codereview.chromium.org/3357022/diff/9001/7002#newcode900
src/arm/full-codegen-arm.cc:900: // Walk the rest of the chain using a single
register without
delete "using a single register"

http://codereview.chromium.org/3357022/diff/9001/7002#newcode901
src/arm/full-codegen-arm.cc:901: // clobbering esi.
esi -> cp

http://codereview.chromium.org/3357022/diff/9001/7002#newcode933
src/arm/full-codegen-arm.cc:933: if (potential_slot->var()->mode() ==
Variable::CONST) {
Line 1071 below could use the same pattern (a conditional LoadRoot instead a
jump around it).

http://codereview.chromium.org/3357022/diff/9001/7002#newcode1819
src/arm/full-codegen-arm.cc:1819: __ CallRuntime(Runtime::kLoadContextSlot, 2);
Maybe change these two pushes to Push(r0, r1). It cannot generate an stm
instruction in this case, but anyway.

http://codereview.chromium.org/3357022/diff/9001/7002#newcode1834
src/arm/full-codegen-arm.cc:1834: __ ldr(r1, FieldMemOperand(r1,
GlobalObject::kGlobalReceiverOffset));
Maybe change to Push(r0, r1), or even better have
EmitDynamicLoadFromSlotFastCase return result in r1 and load global receiver in
r0 to use Push(r1, r0) which will generate an stm instruction.

http://codereview.chromium.org/3357022/diff/9001/7006
File src/x64/full-codegen-x64.cc (right):

http://codereview.chromium.org/3357022/diff/9001/7006#newcode964
src/x64/full-codegen-x64.cc:964: // clobbering esi.
esi -> rsi

[Mads Ager (chromium), 2010-09-10 10:51:42]

http://codereview.chromium.org/3357022/diff/9001/7002
File src/arm/full-codegen-arm.cc (right):

http://codereview.chromium.org/3357022/diff/9001/7002#newcode900
src/arm/full-codegen-arm.cc:900: // Walk the rest of the chain using a single
register without
On 2010/09/10 07:44:24, Søren Gjesse wrote:
> delete "using a single register"

Done.

http://codereview.chromium.org/3357022/diff/9001/7002#newcode901
src/arm/full-codegen-arm.cc:901: // clobbering esi.
On 2010/09/10 07:44:24, Søren Gjesse wrote:
> esi -> cp

Done.

http://codereview.chromium.org/3357022/diff/9001/7002#newcode933
src/arm/full-codegen-arm.cc:933: if (potential_slot->var()->mode() ==
Variable::CONST) {
On 2010/09/10 07:44:24, Søren Gjesse wrote:
> Line 1071 below could use the same pattern (a conditional LoadRoot instead a
> jump around it).

Good catch. Done. :)

http://codereview.chromium.org/3357022/diff/9001/7002#newcode1819
src/arm/full-codegen-arm.cc:1819: __ CallRuntime(Runtime::kLoadContextSlot, 2);
On 2010/09/10 07:44:24, Søren Gjesse wrote:
> Maybe change these two pushes to Push(r0, r1). It cannot generate an stm
> instruction in this case, but anyway.

Done.

http://codereview.chromium.org/3357022/diff/9001/7002#newcode1834
src/arm/full-codegen-arm.cc:1834: __ ldr(r1, FieldMemOperand(r1,
GlobalObject::kGlobalReceiverOffset));
On 2010/09/10 07:44:24, Søren Gjesse wrote:
> Maybe change to Push(r0, r1), or even better have
> EmitDynamicLoadFromSlotFastCase return result in r1 and load global receiver
in
> r0 to use Push(r1, r0) which will generate an stm instruction.

That would be neat. I made it Push(r0, r1) for now. It makes it simpler in other
cases that it returns the result in the same register as the runtime context
slot lookup.

http://codereview.chromium.org/3357022/diff/9001/7006
File src/x64/full-codegen-x64.cc (right):

http://codereview.chromium.org/3357022/diff/9001/7006#newcode964
src/x64/full-codegen-x64.cc:964: // clobbering esi.
On 2010/09/10 07:44:24, Søren Gjesse wrote:
> esi -> rsi

Done.