Changes for running tests on arm64 hardware.

Changes for running tests on arm64 hardware.

R=regis@google.com

Committed: https://code.google.com/p/dart/source/detail?r=37046

[regis, 2014-06-05 17:55:35]

LGTM with one question

https://codereview.chromium.org/311413002/diff/1/runtime/vm/runtime_entry_arm...
File runtime/vm/runtime_entry_arm64.cc (right):

https://codereview.chromium.org/311413002/diff/1/runtime/vm/runtime_entry_arm...
runtime/vm/runtime_entry_arm64.cc:49: __ mov(R26, SP);
R25 and R26 are callee-saved and will thus be preserved by the leaf call. But
how do you know they are available and not currently used by the calling Dart
code?

[zra, 2014-06-05 18:35:53]

https://codereview.chromium.org/311413002/diff/1/runtime/vm/runtime_entry_arm...
File runtime/vm/runtime_entry_arm64.cc (right):

https://codereview.chromium.org/311413002/diff/1/runtime/vm/runtime_entry_arm...
runtime/vm/runtime_entry_arm64.cc:49: __ mov(R26, SP);
On 2014/06/05 17:55:34, regis wrote:
> R25 and R26 are callee-saved and will thus be preserved by the leaf call. But
> how do you know they are available and not currently used by the calling Dart
> code?

Before IL instructions with a kCall LocationSummary, the compiler spills the
live registers. We don't have any leaf runtime calls outside of these IL
instructions.

[regis, 2014-06-05 18:42:30]

On 2014/06/05 18:35:53, zra wrote:
>
https://codereview.chromium.org/311413002/diff/1/runtime/vm/runtime_entry_arm...
> File runtime/vm/runtime_entry_arm64.cc (right):
> 
>
https://codereview.chromium.org/311413002/diff/1/runtime/vm/runtime_entry_arm...
> runtime/vm/runtime_entry_arm64.cc:49: __ mov(R26, SP);
> On 2014/06/05 17:55:34, regis wrote:
> > R25 and R26 are callee-saved and will thus be preserved by the leaf call.
But
> > how do you know they are available and not currently used by the calling
Dart
> > code?
> 
> Before IL instructions with a kCall LocationSummary, the compiler spills the
> live registers. We don't have any leaf runtime calls outside of these IL
> instructions.

That's right. I remember now.
At some time, we could try to only spill the caller-saved registers before a
leaf runtime call. This would require changes for sure. Maybe not worth it.

Thanks

[zra, 2014-06-05 18:44:01]

Committed patchset #1 manually as r37046 (presubmit successful).