Introduce an ElementTypeMask to recognize simple constraints like "the type of parameter foo is the…

Introduce an ElementTypeMask to recognize simple constraints like "the type of parameter foo is the type of parameter foo".

R=kasperl@google.com

Committed: https://code.google.com/p/dart/source/detail?r=24186

[ngeoffray, 2013-06-18 13:06:30]

This recovers the performance regression due to removing the backend inferrer.

[kasperl, 2013-06-19 08:45:01]

LGTM. We should think about whether or not we want to keep inventing new type
masks or if always wrapping the type mask in the inferrer would be cleaner. If
the inferrer had another concept like:

   class InferredType {
     TypeMask mask;
     Element element;
   }

and consistently used those (with support for containers, etc.) that would give
a better decoupling.

https://codereview.chromium.org/17017003/diff/4001/sdk/lib/_internal/compiler...
File sdk/lib/_internal/compiler/implementation/types/element_type_mask.dart
(right):

https://codereview.chromium.org/17017003/diff/4001/sdk/lib/_internal/compiler...
sdk/lib/_internal/compiler/implementation/types/element_type_mask.dart:13: final
Function fetchForwardTo;
I think this deserves a comment. Explain why you have this callback function --
where is it used, etc.

https://codereview.chromium.org/17017003/diff/4001/sdk/lib/_internal/compiler...
File sdk/lib/_internal/compiler/implementation/types/inferrer_visitor.dart
(right):

https://codereview.chromium.org/17017003/diff/4001/sdk/lib/_internal/compiler...
sdk/lib/_internal/compiler/implementation/types/inferrer_visitor.dart:45: //
TODO(ngeoffray): Remove this quick check once the TODO below is
Isn't this always a good optimization? It avoid computing unnecessary unions.

https://codereview.chromium.org/17017003/diff/4001/sdk/lib/_internal/compiler...
File sdk/lib/_internal/compiler/implementation/types/simple_types_inferrer.dart
(right):

https://codereview.chromium.org/17017003/diff/4001/sdk/lib/_internal/compiler...
sdk/lib/_internal/compiler/implementation/types/simple_types_inferrer.dart:958:
TypeMask computeLUBFor(TypeMask firstType,
Maybe change this to computeLubFor to make it easier on the eyes.

https://codereview.chromium.org/17017003/diff/4001/sdk/lib/_internal/compiler...
File sdk/lib/_internal/compiler/implementation/types/type_mask.dart (right):

https://codereview.chromium.org/17017003/diff/4001/sdk/lib/_internal/compiler...
sdk/lib/_internal/compiler/implementation/types/type_mask.dart:53: bool get
isForwarding;
Do you need to add these to the TypeMask implementation in
concrete_type_inferrer.dart?

https://codereview.chromium.org/17017003/diff/4001/sdk/lib/_internal/compiler...
File sdk/lib/_internal/compiler/implementation/types/union_type_mask.dart
(right):

https://codereview.chromium.org/17017003/diff/4001/sdk/lib/_internal/compiler...
sdk/lib/_internal/compiler/implementation/types/union_type_mask.dart:27: while
(mask.isForwarding) {
Maybe move this code into a helper?

[ngeoffray, 2013-06-19 13:42:55]

Thanks Kasper.

https://codereview.chromium.org/17017003/diff/4001/sdk/lib/_internal/compiler...
File sdk/lib/_internal/compiler/implementation/types/element_type_mask.dart
(right):

https://codereview.chromium.org/17017003/diff/4001/sdk/lib/_internal/compiler...
sdk/lib/_internal/compiler/implementation/types/element_type_mask.dart:13: final
Function fetchForwardTo;
On 2013/06/19 08:45:01, kasperl wrote:
> I think this deserves a comment. Explain why you have this callback function
--
> where is it used, etc.

Done.

https://codereview.chromium.org/17017003/diff/4001/sdk/lib/_internal/compiler...
File sdk/lib/_internal/compiler/implementation/types/inferrer_visitor.dart
(right):

https://codereview.chromium.org/17017003/diff/4001/sdk/lib/_internal/compiler...
sdk/lib/_internal/compiler/implementation/types/inferrer_visitor.dart:45: //
TODO(ngeoffray): Remove this quick check once the TODO below is
On 2013/06/19 08:45:01, kasperl wrote:
> Isn't this always a good optimization? It avoid computing unnecessary unions.

Yes. OK. Removed the TODO.

https://codereview.chromium.org/17017003/diff/4001/sdk/lib/_internal/compiler...
File sdk/lib/_internal/compiler/implementation/types/simple_types_inferrer.dart
(right):

https://codereview.chromium.org/17017003/diff/4001/sdk/lib/_internal/compiler...
sdk/lib/_internal/compiler/implementation/types/simple_types_inferrer.dart:958:
TypeMask computeLUBFor(TypeMask firstType,
On 2013/06/19 08:45:01, kasperl wrote:
> Maybe change this to computeLubFor to make it easier on the eyes.

Done.

https://codereview.chromium.org/17017003/diff/4001/sdk/lib/_internal/compiler...
File sdk/lib/_internal/compiler/implementation/types/type_mask.dart (right):

https://codereview.chromium.org/17017003/diff/4001/sdk/lib/_internal/compiler...
sdk/lib/_internal/compiler/implementation/types/type_mask.dart:53: bool get
isForwarding;
On 2013/06/19 08:45:01, kasperl wrote:
> Do you need to add these to the TypeMask implementation in
> concrete_type_inferrer.dart?

Yes, done.

https://codereview.chromium.org/17017003/diff/4001/sdk/lib/_internal/compiler...
File sdk/lib/_internal/compiler/implementation/types/union_type_mask.dart
(right):

https://codereview.chromium.org/17017003/diff/4001/sdk/lib/_internal/compiler...
sdk/lib/_internal/compiler/implementation/types/union_type_mask.dart:27: while
(mask.isForwarding) {
On 2013/06/19 08:45:01, kasperl wrote:
> Maybe move this code into a helper?

Done.

[ngeoffray, 2013-06-19 13:59:54]

Committed patchset #3 manually as r24186 (presubmit successful).

[ahe, 2013-06-24 07:47:56]

This CL introduced a performance regression in dart2js compile-time compiling
itself.  With this CL, dart2js is 1.65 times slower at compiling itself.

I'll investigate how much of this is due to lack of hashCode implementations.