Modify growth policy for table of canonical types in each class.

Modify growth policy for table of canonical types in each class.
Modify growth policy for cache of type arguments instantiations.
Add language test for generic closure.

R=iposva@google.com

Committed: https://code.google.com/p/dart/source/detail?r=32940

[regis, 2014-02-21 18:57:59]

I will remove the obsolete code for closure allocation in a following cl.

Ryan, I cc'ed you so that we can discuss mirrors and type printing (see new test
without mirrors for now).

Thanks,
Regis

[Ivan Posva, 2014-02-21 22:26:56]

LGTMwC

-Ivan

https://codereview.chromium.org/175403004/diff/1/runtime/vm/object.cc
File runtime/vm/object.cc (right):

https://codereview.chromium.org/175403004/diff/1/runtime/vm/object.cc#newcode...
runtime/vm/object.cc:4266: length = (length == 1) ? 3 : ((length - 1) * 2 + 1);
How about not allowing unbounded exponential growth?

length = (length > 64) ?
  (length + 64) :
  ((length == 1) ? 3 : ((length - 1) * 2 + 1));

https://codereview.chromium.org/175403004/diff/1/runtime/vm/object.cc#newcode...
runtime/vm/object.cc:12921: const intptr_t new_length = (length == 0) ? 1 :
(length * 2);
ditto

[regis, 2014-02-21 23:08:47]

Thanks!

https://codereview.chromium.org/175403004/diff/1/runtime/vm/object.cc
File runtime/vm/object.cc (right):

https://codereview.chromium.org/175403004/diff/1/runtime/vm/object.cc#newcode...
runtime/vm/object.cc:4266: length = (length == 1) ? 3 : ((length - 1) * 2 + 1);
On 2014/02/21 22:26:56, Ivan Posva wrote:
> How about not allowing unbounded exponential growth?
> 
> length = (length > 64) ?
>   (length + 64) :
>   ((length == 1) ? 3 : ((length - 1) * 2 + 1));

Done.

https://codereview.chromium.org/175403004/diff/1/runtime/vm/object.cc#newcode...
runtime/vm/object.cc:12921: const intptr_t new_length = (length == 0) ? 1 :
(length * 2);
On 2014/02/21 22:26:56, Ivan Posva wrote:
> ditto

Done.

[regis, 2014-02-21 23:09:08]

Committed patchset #2 manually as r32940 (presubmit successful).

[Florian Schneider, 2014-02-24 09:29:00]

dbc

https://codereview.chromium.org/175403004/diff/70001/runtime/vm/stub_code_ia3...
File runtime/vm/stub_code_ia32.cc (right):

https://codereview.chromium.org/175403004/diff/70001/runtime/vm/stub_code_ia3...
runtime/vm/stub_code_ia32.cc:1226: // but only implicit instance closures.
Simplify.
I noticed that as well after I changed the way regular closure are allocated to
allow allocation sinking: Use a AllocateObject + 2 * StoreInstanceField to
initialize the closure object.

I'm thinking about doing the same for instance closures so they also work with
allocation sinking. As a side-effect, this stub can be removed completely. What
do you think?