dc4758d6bc
Move resolve* from NativeResolutionEnqueuer to NativeDataResolver
...
This is a step towards making NativeResolutionEnqueuer model independent.
R=efortuna@google.com
Review-Url: https://codereview.chromium.org/2778803002 .
2017-03-28 11:08:47 +02:00
5aa715b7a8
Use Entity in native_data.
...
NativeBasicDataImpl and NativeDataImpl are now reusable for J/K-elements.
R=efortuna@google.com
Review-Url: https://codereview.chromium.org/2775283002 .
2017-03-28 11:02:49 +02:00
40072e2cce
Mostly use Entity in InterceptorData
...
This is a step towards making InterceptorDataImpl reusable for J/K-elements.
R=efortuna@google.com
Review-Url: https://codereview.chromium.org/2781483002 .
2017-03-28 10:53:06 +02:00
e7a72961c8
Make codegen and optimizations depend more directly on data objects.
...
This is a move towards passing all information through data objects
computed by previous phases rather than pulling it directly from
Backend or Compiler.
This is needed to support a shift from model K to model J between
resolution and codegen.
R=efortuna@google.com
Review-Url: https://codereview.chromium.org/2777163002 .
2017-03-28 10:38:29 +02:00
a2286d6eb1
Derive CommonElements directly from ElementEnvironment.
...
This prepares for a merge of CommonElements(Impl) and BackendHelpers.
R=efortuna@google.com
Review-Url: https://codereview.chromium.org/2775223002 .
2017-03-28 10:32:23 +02:00
02573b3caa
Remove Fasta's copy of accessors.dart.
...
R=asgerf@google.com , jensj@google.com
Review-Url: https://codereview.chromium.org/2777883002 .
2017-03-27 14:04:43 +02:00
a9811024cf
Split NativeBasicDataImpl
...
- into NativeBasicDataBuilderImpl and NativeBasicDataImpl
R=sigmund@google.com
Review-Url: https://codereview.chromium.org/2756103003 .
2017-03-27 11:58:57 +02:00
65b06efb58
Split MirrorsAnalysis
...
- into MirrorsResolutionAnalysis(Impl), MirrorsCodegenAnalysis(Impl)
R=sigmund@google.com
Review-Url: https://codereview.chromium.org/2748383003 .
2017-03-27 11:33:25 +02:00
184c2a5778
Lazily initialize library elements.
...
BUG=
R=sigmund@google.com
Review-Url: https://codereview.chromium.org/2771303002 .
2017-03-24 13:43:55 -07:00
4156ca6a2b
Eliminate multi-callback structure for LibraryLoader.
...
BUG=
R=sigmund@google.com
Review-Url: https://codereview.chromium.org/2760923004 .
2017-03-24 12:20:46 -07:00
ee591a87d2
Fix warnings in dart2js and analyzer caused by Fasta.
...
Review-Url: https://codereview.chromium.org/2766323005 .
2017-03-23 15:39:03 +01:00
585829f755
Fix #29130 : only consider classes allocated with generative constructors as potential closures
...
BUG=
R=sra@google.com
Review-Url: https://codereview.chromium.org/2767103003 .
2017-03-22 16:03:43 -07:00
d6e324f056
Rerun GVN if load elimination creates GVN candidates.
...
This does not fire often but is occasionally useful. e.g
t3 = t2._collection$_length;
if (t3 === 0)
timer = null;
else {
if (t3 === 0)
H.throwExpression(new P.StateError("No such element"));
timer = t2._collection$_first;
}
-->
if (t2._collection$_length === 0)
timer = null;
else
timer = t2._collection$_first;
After load-elimination creates t3 = t2._collections$_length, t3 == 0 becomes a GVN candidate. Then the inner "x" in "if (x) ... else { if (x) ... }" is replaced by "false".
The original code was something like "timer = t2.isEmpty ? null : t2.first".
BUG=
R=sigmund@google.com
Review-Url: https://codereview.chromium.org/2441463003 .
2017-03-22 12:17:18 -07:00
002d5c7150
Update messages.yaml with asynchronous messages.
...
R=karlklose@google.com
Review-Url: https://codereview.chromium.org/2754793004 .
2017-03-22 12:27:38 +01:00
819b226598
Add begin events for const and new expressions.
...
R=paulberry@google.com
Review-Url: https://codereview.chromium.org/2767543003 .
2017-03-22 12:06:23 +01:00
cf5d9e4c39
Missing import.
...
Review-Url: https://codereview.chromium.org/2765183002 .
2017-03-22 10:01:57 +01:00
983a47293c
Fix IdentifierContext for primary expressions. Also pass IdentifierContext to handleThisExpression and handleSuperExpression.
...
R=paulberry@google.com
Review-Url: https://codereview.chromium.org/2768523002 .
2017-03-22 09:39:29 +01:00
a5e9674010
Fix compiler formatter errors.
...
BUG=
R=sra@google.com
Review-Url: https://codereview.chromium.org/2767723003 .
2017-03-21 20:47:49 -07:00
ff26d99868
Type tracing bug fix for issue 28919.
...
Storing a traced value into a non-traced list should escape the traced value.
Also: changes to graph dumper to include all call nodes.
BUG= https://github.com/dart-lang/sdk/issues/28919
R=efortuna@google.com , sigmund@google.com
Review-Url: https://codereview.chromium.org/2762493002 .
2017-03-21 10:55:34 -07:00
a1a34a1c84
[kernel] Debugging of switch statement
...
BUG=
R=ahe@google.com , kmillikin@google.com
Review-Url: https://codereview.chromium.org/2750013002 .
2017-03-21 11:36:48 +01:00
4d208e2ffb
[fasta] Add and correct even more offsets
...
E.g. kernel debugging of async functions should now work.
BUG=
R=ahe@google.com
Review-Url: https://codereview.chromium.org/2746013003 .
2017-03-21 10:57:30 +01:00
02b5710a1d
Add return value.
...
Review-Url: https://codereview.chromium.org/2753153003 .
2017-03-17 19:59:01 +01:00
8979b040ec
Improve parsing of async and generator methods.
...
R=johnniwinther@google.com
Review-Url: https://codereview.chromium.org/2759663002 .
2017-03-17 18:03:42 +01:00
323d5b0b7f
Track async state in parser.
...
R=johnniwinther@google.com
Review-Url: https://codereview.chromium.org/2759623003 .
2017-03-17 14:43:07 +01:00
5934a2c443
Change comments that are interpreted by analyzer.
...
R=johnniwinther@google.com
Review-Url: https://codereview.chromium.org/2753253003 .
2017-03-17 12:29:26 +01:00
d29c4e2256
Compute ids for IR nodes and check equivalence with the AST.
...
This creates a basis for checking type inference based on kernel IR against
the unittests for the original AST based implementation.
R=efortuna@google.com
Review-Url: https://codereview.chromium.org/2750353003 .
2017-03-17 09:50:01 +01:00
0fa285f880
Pulling the element model out of global type inference.
...
BUG=
R=johnniwinther@google.com , sigmund@google.com
Review-Url: https://codereview.chromium.org/2746293006 .
2017-03-16 18:03:46 -07:00
c3330b1407
Complain about built-in identifiers.
...
R=johnniwinther@google.com , sigmund@google.com
Review-Url: https://codereview.chromium.org/2750863002 .
2017-03-16 11:01:27 +01:00
4321db0c8b
Split MirrorsData
...
- into MirrorsDataBuilder and MirrorsData both implemented by MirrorsDataImpl
R=sigmund@google.com
Review-Url: https://codereview.chromium.org/2743243003 .
2017-03-16 10:12:21 +01:00
33717e010a
Run dartfmt on dart2js codebase.
...
BUG=
R=sra@google.com
Review-Url: https://codereview.chromium.org/2746333008 .
2017-03-15 20:38:18 -07:00
e933720d88
dart2js: replace calls to codeUnitAt when index has favourable type
...
If the index is known to be a non-negative integer, only a simple
bounds check is required, so call a simpler version of codeUnitAt.
In a large app, about 60% of calls meet this condition.
Benchmarks do not show speedups, but that is because they are too
simple so global type inference can usually tell that the index is not
null, and often that it is always a non-negative integer, so the full version of codeUnitAt is optimized to look like the simple version. This does not happen in large apps.
BUG=
R=efortuna@google.com
Review-Url: https://codereview.chromium.org/2746033006 .
2017-03-15 19:16:23 -07:00
4a8ed3c260
Lower shift-left when count is masked into valid range.
...
This pattern, e.g. (1 << (a & 15)), occurs in Uri parsing.
R=efortuna@google.com
Review-Url: https://codereview.chromium.org/2756543003 .
2017-03-15 18:52:28 -07:00
ef79634132
Re-land "Add support for metadata on type variables to Fasta parser." (with fixes).
...
Last time I landed this I left a crucial `pop()` out of BodyBuilder,
resulting in test failures.
R=sigmund@google.com
Review-Url: https://codereview.chromium.org/2750503012 .
2017-03-15 12:42:23 -07:00
ae85bcf954
Parse for-loop statements with Fasta.
...
R=ahe@google.com , paulberry@google.com
BUG=
Review-Url: https://codereview.chromium.org/2746333005 .
2017-03-15 11:40:32 -07:00
6c76f592a6
Move TreeElementMixin to avoid out-of-library private overrides
...
This is one of two solutions for Issue 28809
https://codereview.chromium.org/2722923002
https://codereview.chromium.org/2722043002
We should pick one.
R=johnniwinther@google.com , sigmund@google.com
Review-Url: https://codereview.chromium.org/2722923002 .
2017-03-15 09:56:28 -07:00
cb8b56523c
Pending renames and cleanups after commits
...
Review-Url: https://codereview.chromium.org/2751983002 .
2017-03-15 15:33:42 +01:00
99fa028982
Handle 'part of' declarations with URI correctly in Fasta.
...
R=paulberry@google.com
Review-Url: https://codereview.chromium.org/2747173002 .
2017-03-15 15:11:36 +01:00
4414836936
Register type constants through ConstantUse
...
R=sigmund@google.com
Review-Url: https://codereview.chromium.org/2750503003 .
2017-03-15 14:52:09 +01:00
0924b17267
Create LookupMapAnalysis late.
...
R=sigmund@google.com
Review-Url: https://codereview.chromium.org/2745133002 .
2017-03-15 14:46:02 +01:00
5ec55890fa
Add ConstantUse to WorldImpact
...
R=sigmund@google.com
Review-Url: https://codereview.chromium.org/2742283002 .
2017-03-15 14:34:40 +01:00
db7262d4ae
Extract CodegenImpactTransformer from ImpactTransformer
...
+ remove backend use from JavaScriptImpactTransformer
+ almost remove use of backend in CodegenImpactTransformer
R=sigmund@google.com
Review-Url: https://codereview.chromium.org/2728393005 .
2017-03-15 14:11:16 +01:00
69130277ef
Split NativeDataImpl
...
- into NativeClassDataImpl and NativeDataImpl
R=sigmund@google.com
Review-Url: https://codereview.chromium.org/2738513007 .
2017-03-15 13:54:40 +01:00
4b9297caad
Create ResolutionEnqueuer after library loading.
...
R=sigmund@google.com
Review-Url: https://codereview.chromium.org/2735763002 .
2017-03-15 13:46:48 +01:00
b08686ba67
Prepare for late creation of ResolutionEnqueuer
...
- late as in 'after library loading'
R=sigmund@google.com
Review-Url: https://codereview.chromium.org/2729383002 .
2017-03-15 13:33:19 +01:00
8e1e08ce4a
Add ClassEntity.library, MemberEntity.library and FunctionEntity.isExternal
...
- and mostly use entities in native_data.dart
R=sigmund@google.com
Review-Url: https://codereview.chromium.org/2731173002 .
2017-03-15 13:18:31 +01:00
fbbf1e90ec
Split js interop registration into library/class/member elements
...
R=sigmund@google.com
Review-Url: https://codereview.chromium.org/2731163002 .
2017-03-15 12:58:55 +01:00
69b47ea010
Add NativeDataResolver
...
- to further align the NativeData computation towards separating builders from data.
R=sigmund@google.com
Review-Url: https://codereview.chromium.org/2732793002 .
2017-03-15 11:54:48 +01:00
f7ec06f210
Add NativeClassData/NativeClassDataBuilder
...
- for information computed during loading (pre-resolution)
R=sigmund@google.com
Review-Url: https://codereview.chromium.org/2727253007 .
2017-03-15 11:24:13 +01:00
28f7099f17
Remove backend reference from ResolutionEnqueuerListener.
...
R=sigmund@google.com
Review-Url: https://codereview.chromium.org/2729943004 .
2017-03-15 11:17:26 +01:00
452c380db1
Late creation of CodegenEnqueuerListener and CodegenWorldBuilder
...
- now with no reference to JavaScriptBackend!
R=sigmund@google.com
Review-Url: https://codereview.chromium.org/2727373003 .
2017-03-15 11:08:43 +01:00
Johnni Winther