Code patching for ARM64.

Code patching for ARM64.

R=regis@google.com

Committed: https://code.google.com/p/dart/source/detail?r=35156

[zra, 2014-04-17 21:27:27]

https://codereview.chromium.org/241573002/diff/20001/runtime/vm/assembler_arm...
File runtime/vm/assembler_arm64.h (right):

https://codereview.chromium.org/241573002/diff/20001/runtime/vm/assembler_arm...
runtime/vm/assembler_arm64.h:638: sub(pp, pp, Operand(kHeapObjectTag));
When in the PP register, the object pool pointer is untagged. When I push it on
the stack with PushPP I tag it again. PopPP then untags when restoring from the
stack. This will make loading from the object pool only one instruction for the
first 4096 entries. Otherwise, becuase the offset wouldn't be aligned, it would
always be at least two instructions.

https://codereview.chromium.org/241573002/diff/20001/runtime/vm/constants_arm...
File runtime/vm/constants_arm64.h (right):

https://codereview.chromium.org/241573002/diff/20001/runtime/vm/constants_arm...
runtime/vm/constants_arm64.h:505: kImm19Mask = 0x7ffff << kImm19Shift,
Changing to match kImm16Mask.

[regis, 2014-04-17 21:42:28]

LGTM

https://codereview.chromium.org/241573002/diff/20001/runtime/vm/assembler_arm...
File runtime/vm/assembler_arm64.h (right):

https://codereview.chromium.org/241573002/diff/20001/runtime/vm/assembler_arm...
runtime/vm/assembler_arm64.h:638: sub(pp, pp, Operand(kHeapObjectTag));
On 2014/04/17 21:27:27, zra wrote:
> When in the PP register, the object pool pointer is untagged. When I push it
on
> the stack with PushPP I tag it again. PopPP then untags when restoring from
the
> stack. This will make loading from the object pool only one instruction for
the
> first 4096 entries. Otherwise, becuase the offset wouldn't be aligned, it
would
> always be at least two instructions.

Sounds good. How about you merge most of this comment into the comment on line
637?

[zra, 2014-04-17 21:56:32]

https://codereview.chromium.org/241573002/diff/20001/runtime/vm/assembler_arm...
File runtime/vm/assembler_arm64.h (right):

https://codereview.chromium.org/241573002/diff/20001/runtime/vm/assembler_arm...
runtime/vm/assembler_arm64.h:638: sub(pp, pp, Operand(kHeapObjectTag));
On 2014/04/17 21:42:28, regis wrote:
> On 2014/04/17 21:27:27, zra wrote:
> > When in the PP register, the object pool pointer is untagged. When I push it
> on
> > the stack with PushPP I tag it again. PopPP then untags when restoring from
> the
> > stack. This will make loading from the object pool only one instruction for
> the
> > first 4096 entries. Otherwise, becuase the offset wouldn't be aligned, it
> would
> > always be at least two instructions.
> 
> Sounds good. How about you merge most of this comment into the comment on line
> 637?

Done.

[zra, 2014-04-17 22:00:18]

Committed patchset #3 manually as r35156 (presubmit successful).