A64: Move veneer emission checking in the Assembler.

A64: Move veneer emission checking in the Assembler.

The previous heuristic would break as a significant amount of code could be
generated without checking for veneer emission.
The veneer emission is now done in the Assembler, in a very similar way to
constant pool emission.

BUG=v8:3177
LOG=N
R=ulan@chromium.org

Committed: https://code.google.com/p/v8/source/detail?r=19661

[Alexandre Rames, 2014-02-26 18:39:00]

Since we regularly see failures when the veneer check margin is overflowed,
update the way we check for veneers.

[ulan, 2014-02-28 09:59:24]

LGTM, thanks for implementing this!

https://codereview.chromium.org/181873002/diff/1/src/a64/assembler-a64.h
File src/a64/assembler-a64.h (right):

https://codereview.chromium.org/181873002/diff/1/src/a64/assembler-a64.h#newc...
src/a64/assembler-a64.h:1763: void CheckVeneerPool(bool force_emit,
Did I miss something or CheckVeneerPool is never called with force_emit=true?

[Alexandre Rames, 2014-02-28 10:13:13]

https://codereview.chromium.org/181873002/diff/1/src/a64/assembler-a64.h
File src/a64/assembler-a64.h (right):

https://codereview.chromium.org/181873002/diff/1/src/a64/assembler-a64.h#newc...
src/a64/assembler-a64.h:1763: void CheckVeneerPool(bool force_emit,
On 2014/02/28 09:59:24, ulan wrote:
> Did I miss something or CheckVeneerPool is never called with force_emit=true?

You did not.
At least for now, forcing the generation should never be required as linked
Labels must be bound before the end of code generation.
I kept it like this to mirror CheckConstPool and thinking it may be useful in
the future.
Maybe I should just remove the parameter?

[ulan, 2014-02-28 10:20:53]

> Maybe I should just remove the parameter?
I don't have strong opinion on this, both options are fine.

[Alexandre Rames, 2014-02-28 10:50:46]

On 2014/02/28 10:20:53, ulan wrote:
> > Maybe I should just remove the parameter?
> I don't have strong opinion on this, both options are fine.

I uploaded a new patch set with the parameter removed.
My final thoughts: removing the parameter makes clear that forcing the emission
should not be needed currently. This can easily be added back if we need to.

[ulan, 2014-03-04 15:29:44]

> I uploaded a new patch set with the parameter removed.

New patch set looks good. Since the CL will turn a64 bots green and doesn't
affect other architectures, feel free to land this CL when ready with
--bypass-hooks.

[Alexandre Rames, 2014-03-04 15:54:22]

Committed patchset #2 manually as r19661 (tree was closed).