MIPS: arch-independent changes to support mips.

MIPS: arch-independent changes to support mips.

This change supports all non-crankshaft features except serialization.

This must be built after the changes in http://codereview.chromium.org/6966031
are landed.

BUG=
TEST=

Committed: http://code.google.com/p/v8/source/detail?r=8023

[Søren Thygesen Gjesse, 2011-05-23 09:20:58]

Please consider the change where on overloaded VisitPointer is required.

http://codereview.chromium.org/7039058/diff/1/src/mips/assembler-mips-inl.h
File src/mips/assembler-mips-inl.h (right):

http://codereview.chromium.org/7039058/diff/1/src/mips/assembler-mips-inl.h#n...
src/mips/assembler-mips-inl.h:131: reconstructed_obj_ptr_ =
4 space indentation

http://codereview.chromium.org/7039058/diff/3001/src/mips/assembler-mips-inl....
src/mips/assembler-mips-inl.h:234: // that do not need it.
I think we should try to avoid adding this overloaded VisitPointer. How about
something like this:

Object* orig = *target_object_address();
visitor->VisitPointer(target_object_address();
if (orig != *target_object_address()) {
  set_target_object(*target_object_address());
}

This will avoid having special casing when visiting some types of relocation
info. Making the change this way avoids changes to mark-compact.cc and
objects.h.

[Paul Lind, 2011-05-24 06:30:22]

http://codereview.chromium.org/7039058/diff/1/src/mips/assembler-mips-inl.h
File src/mips/assembler-mips-inl.h (right):

http://codereview.chromium.org/7039058/diff/1/src/mips/assembler-mips-inl.h#n...
src/mips/assembler-mips-inl.h:131: reconstructed_obj_ptr_ =
On 2011/05/23 09:20:58, Søren Gjesse wrote:
> 4 space indentation

Done.

http://codereview.chromium.org/7039058/diff/3001/src/mips/assembler-mips-inl....
src/mips/assembler-mips-inl.h:234: // that do not need it.
On 2011/05/23 09:20:58, Søren Gjesse wrote:
> I think we should try to avoid adding this overloaded VisitPointer. How about
> something like this:
> 
> Object* orig = *target_object_address();
> visitor->VisitPointer(target_object_address();
> if (orig != *target_object_address()) {
>   set_target_object(*target_object_address());
> }
> 
> This will avoid having special casing when visiting some types of relocation
> info. Making the change this way avoids changes to mark-compact.cc and
> objects.h.

Nice idea, much cleaner than my approach. Implemented slightly differently,
since we can't call target_object_address() after VisitPointer, or we re-read
old value. Removed the now-unnecessary updates to mark-compact.cc & objects.h.
Thanks.

[Søren Thygesen Gjesse, 2011-05-24 07:54:36]

LGTM!

The full compiler now runs all tests!

Thanks for working on this!

I will commit it.

http://codereview.chromium.org/7039058/diff/10001/src/mips/assembler-mips-inl.h
File src/mips/assembler-mips-inl.h (right):

http://codereview.chromium.org/7039058/diff/10001/src/mips/assembler-mips-inl...
src/mips/assembler-mips-inl.h:234: visitor->VisitPointer(p);
Instead of the additional local var p you could use reconstructed_obj_ptr_
instead, but maybe that is relying to much on the internal details of how
target_object_address() works.