azalea/kotlin-fork
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
110,149 Commits 1 Branch 0 Tags
master
Commit Graph

5988 Commits

Author SHA1 Message Date
Dmitry Petrov 46d2e52543 JVM Minor: move uninitialized value check to stack recovery 2021-07-20 19:51:04 +03:00
Dmitry Petrov e72d24960d JVM we don't merge frames in stack normalization 2021-07-20 19:51:03 +03:00
Dmitry Petrov 91afa3335c JVM represent uninitialized values explicitly in stack normalization 2021-07-20 19:51:02 +03:00
Dmitry Petrov 35b4a9d4cc JVM treat all variable loads as "initialized" in stack normalization 2021-07-20 19:51:00 +03:00
Dmitry Petrov c8b705fd2d JVM update license for HackedFixStackMethodAnalyzerBase.kt 2021-07-20 19:50:59 +03:00
Dmitry Petrov 8763235053 JVM specially hacked method analyzer for FixStack 2021-07-20 19:50:58 +03:00
Dmitry Petrov 3ecd612ce7 JVM FixStackValue & FixStackInterpreter 2021-07-20 19:50:57 +03:00
Dmitriy Novozhilov 85c7f386eb Add Kotlin 1.7 language version 2021-07-20 13:33:41 +03:00
Ilmir Usmanov e870a200c4 Merge consequent LVT records 
that is, if LVT record starts where previous one ends, merge them.
 #KT-47749 Fixed
 2021-07-19 22:07:16 +00:00
Ilmir Usmanov 5ae01c8b2a Do not duplicate $result in LVT 
#KT-47749
 2021-07-19 22:07:16 +00:00
Ilmir Usmanov b4d356c5bd Split LVT record for known nulls 
Since they are not spilled, the logic for splitting LVT records, that
is applied for spilled variables, was not applied for known nulls.
Fix that by applying the logic to them.
 #KT-47749
 2021-07-19 22:07:15 +00:00
Mads Ager 83dddd73b0 Always add a local variable for its live ranges. 
There used to be code that extended a previous range instead.
However, that does not work as that extension could have the
local cover code where it does not exists. Since we no longer
extend the range of locals, we should always introduce a new
one even if there was another one for a previous range.
 2021-07-19 22:07:14 +00:00
Dmitry Petrov e276dec4de JVM don't merge local values in FixStackAnalyzer 
We care only about stacks there.
This yields about 10-15% in a pathological case such as KT-41510.
 2021-07-19 19:24:57 +03:00
Dmitry Petrov b98322c1b4 JVM don't optimize methods with too many TCBs 2021-07-15 20:18:18 +03:00
Dmitry Petrov d65d66c03a JVM reserve 2 slots for node predecessors/successors in CFG 2021-07-15 20:18:15 +03:00
Dmitry Petrov 6c2d93bb8b JVM traverse nodes in reverse order in 'findSafelyReachableReturns' 2021-07-15 20:18:13 +03:00
Mikhail Glukhikh 113d2653aa Drop deprecated -Xexperimental flag from compiler and tests 2021-07-14 21:18:23 +03:00
pyos 07cb3a5ff8 JVM: do not reify methods of objects in lambdas 
All type parameters used in them are not from the inline function
anyway.
 2021-07-14 10:11:05 +02:00
pyos 717cf2066a JVM: remove more redundant properties from LambdaInfo 2021-07-14 10:11:05 +02:00
pyos c1c56ca388 JVM: refactor extraction of default inline lambdas a bit more 2021-07-14 10:11:05 +02:00
pyos 91cf1a1a4d JVM: remove a redundant DefaultLambda field 
and add a comment explaining one branch of the inliner... Better than
nothing, I guess?
 2021-07-14 10:11:05 +02:00
pyos 100d2d629c JVM: inline default lambda coercions from bridge invoke 
The type of the default lambda may be a subtype of the parameter type,
so we can't really generate our own coercions at all as we don't know
the precise Kotlin type of the `invoke` method.
 2021-07-14 10:11:05 +02:00
pyos d0b9c4ae6b JVM: generalize getMethodNode 
to allow matching by arbitrary predicates.
 2021-07-14 10:11:05 +02:00
pyos 664bb055c8 JVM_IR: support codegen of property delegate methods 
The purpose of a property-related method is now determined by its
origin; codegen no longer assumes all methods with Property metadata
source are `getX$annotations`.
 2021-07-12 22:38:43 +02:00
Dmitry Petrov 804db3ce91 JVM KT-47613 custom control flow analyzer for CFG builder 2021-07-11 20:16:09 +03:00
Dmitry Petrov fe71435104 JVM KT-47613 traverse nodes backwards in backward analysis 2021-07-11 20:16:09 +03:00
Dmitry Petrov d99d25e51e JVM use SPBs in fix stack analyzer 2021-07-10 19:25:25 +03:00
Dmitry Petrov 38f45d2969 JVM update licence for InstructionLivenessAnalyzer.kt 2021-07-10 19:25:24 +03:00
Dmitry Petrov 68560c60f3 JVM use FastMethodAnalyzer when possible 2021-07-10 19:25:22 +03:00
Dmitry Petrov c54e680021 JVM FastMethodAnalyzer: prune some exception edges 2021-07-10 19:25:21 +03:00
Dmitry Petrov 07b89c6b4b JVM FastMethodAnalyzer 2021-07-10 19:25:20 +03:00
Alexander Udalov ddfa94e7e9 Support Nothing type in typeOf 
The proper support will come in KT-15518, but that would be a breaking
change even for stable Kotlin without kotlin-reflect. Before that issue
is fixed, represent Nothing in types with the Void class, and use a flag
in the no-reflect implementation to remember that it's not actually the
Void class itself.

 #KT-39166 Fixed
 2021-07-09 14:40:05 +02:00
Alexander Udalov 02774fae0c Report error on non-reified type parameter with recursive bound in typeOf 
Instead of throwing an exception.

 #KT-40173
 2021-07-09 14:31:52 +02:00
Alexander Udalov 438ce57183 Report error on typeOf<suspend ...>() 
Otherwise an invalid type is constructed which causes kotlin-reflect to
crash, and stdlib implementation to render the type incorrectly. The
reason is that suspend functional types are not properly supported in
reflection. Once they are supported, this error can be removed.

 #KT-47562
 2021-07-09 14:31:52 +02:00
Dmitry Petrov 24fcadb869 JVM don't run CCE on methods without optimizable conditional jumps 2021-07-08 22:11:59 +03:00
pyos 50797dba8d JVM: do not use crossinline flag when inlining assertions 
Crossinline lambdas *can* be inlined into objects, but don't *have* to;
the correct place should be determined from the context, not from the
parameter.
 2021-07-07 16:48:01 +02:00
pyos e64cda61d5 JVM_IR: do not generate $assertionsDisabled twice during inlining 2021-07-07 16:48:01 +02:00
Alexander Udalov 0cb905a4b1 Support mutable collection types in typeOf 
flexibleTypes_1_6.kt is fixed for JVM IR in a subsequent commit.

 #KT-35877 Fixed
 2021-07-01 19:33:55 +02:00
Alexander Udalov 6e975b3498 Support flexible types internally in typeOf 
#KT-45066 Fixed
 2021-07-01 19:33:55 +02:00
Mikhael Bogdanov 5f62b72c82 Properly process big types 2021-06-26 06:10:16 +02:00
Mikhael Bogdanov 0fd1f549a9 Properly process special symbols during indy-with-constants concatenation 
#KT-47320 Fixed
 2021-06-26 06:10:16 +02:00
Dmitry Petrov c77884f067 Refactor SAM type handling, replace non-approximated arguments with * 2021-06-22 21:13:56 +03:00
Victor Petukhov 4aeabb6b0f Use upper bound aware type approximator for intersection types inside sam types in contravariant positions to build proper types in terms of subtyping 2021-06-22 21:13:55 +03:00
pyos ee4d9a89ea JVM_IR: read inline fun bytecode to detect unused property references 
I.e. optimize `$$delegatedProperties` in the same way as the old
backend. This is more reliable than trying to match bytecode.
 2021-06-21 16:19:23 +03:00
pyos aa47191de4 JVM: hide ExpressionCodegen in finally block generation during inlining 2021-06-16 12:13:32 +02:00
pyos 108bd01698 JVM: refine the stack spilling around inline calls 
Not all suspend functions need it - only those with suspension points.
 2021-06-16 12:13:32 +02:00
pyos b136acb185 JVM: move callSiteFile to InlineCallSiteInfo 2021-06-16 12:13:32 +02:00
pyos 6c1a5e1211 JVM: make inline function argument processing a bit shorter 2021-06-16 12:13:32 +02:00
pyos 392e4fba42 JVM: fix inlining of default lambdas of signature (Result) -> Any 
They have no `invoke` bridge, and the overridden invoke expectes a boxed
`Result` as an argument.
 2021-06-16 12:13:32 +02:00
pyos a0a14d9e25 JVM: remove descriptors from SourceCompilerForInline 
Printing the call site source on errors isn't really useful anyway - the
old backend points to a specific PsiElement, and JVM_IR wraps the
exception in FunctionCodegen (and printing the source never worked with
JVM_IR anyway).
 2021-06-16 12:13:32 +02:00