The catch entry block has all locals in fixed locations
(Rj) where j = kNumberOfRegisters - i for parameter i.
This means we reserve a range of DBC registers at the top-end of the frame.
Those registers are blocked for general allocation to avoid any overlap
with the rest of the registers that are allocated from the bottom.
Each optimized frame with a try-catch will be kNumberOfRegisters wide.
BUG=
R=vegorov@google.com
Review URL: https://codereview.chromium.org/2388093003 .
When throwing to a frame scheduled for lazy deopt, update the continuation pc for that frame to be the catch handler.
Weaken new assert that the deopt pc belongs to the frame's code as the deopt pc for the last eager deopt in a function lies outside the code, after the call to the deopt stub.
R=fschneider@google.com
Review URL: https://codereview.chromium.org/2392613002 .
A switchable call in the monomorphic state is now basically the same cost as a test and call, so this is as performant without ht code size cost of the slow path megemorphic call.
Flutter gallery before:
VMIsolate(CodeSize): 114052
Isolate(CodeSize): 2250283
ReadOnlyData(CodeSize): 1744760
Instructions(CodeSize): 5679488
Total(CodeSize): 9788583
After:
VMIsolate(CodeSize): 114085
Isolate(CodeSize): 2047224
ReadOnlyData(CodeSize): 1666808
Instructions(CodeSize): 5405424 -4.8%
Total(CodeSize): 9233541 -5.6%
R=fschneider@google.com
Review URL: https://codereview.chromium.org/2376893003 .
Catch block entries need to be considered when doing lazy deoptimization.
In addition to the return, also patch all catch entry blocks in a function scheduled for lazy deoptimization.
Also, move restoring the pool pointer to the jump-to-handler stub to simplify patching the catch entry.
BUG=#27419
R=rmacnak@google.com
Review URL: https://codereview.chromium.org/2357343003 .
The call sequence is very similar to a classic IC call, except the guarded class and the target are loaded indirectly from the constant pool instead of as immediates. In the monomorphic case, we call directly to the expected target with a class check in the callee. In the unlinked, polymorphic and megamorphic cases, we call a stub; these case are now call-through instead of call-and-return.
Every code, except stubs involved in switchable calls, includes the class check sequence at the beginning. So we now distinguish between a checked and an unchecked entry point. Generated code except the switchable call continues to use the unchecked entry point.
PC offsets are calculated relative to the beginning of the instruction stream, rather than either entry point.
BUG=
R=fschneider@google.com
Review URL: https://codereview.chromium.org/2226893002 .
- [x] Start reloading even earlier.
- [x] Keep the value of FLAG_reload_every per-isolate (so spawned isolates reload eagerly too).
- [x] Adjust the reload every back off curve to be linear in the beginning.
- [x] Cap the reload every threshold at 1,000,000
- [x] Test status file updates.
- [x] Stop disabling the background compiler / OSR when running tests.
- [x] Skip multi tests with compilation errors when reloading.
BUG=
R=rmacnak@google.com
Review URL: https://codereview.chromium.org/2133823002 .
- Removed the flag use_megamorphic_stub and
- Removed StubCode::EmitMegamorphicLookup (inlined version of megamorphic call).
- Removed the null load in GenerateMegamorphicLookupStub for arm64 to match other platforms.
BUG=
R=fschneider@google.com
Review URL: https://codereview.chromium.org/2132803002 .
- [x] Add an is static call bit to ICData
- [x] Rewrite the reset ICData iterators to rely on that bit rather than the function's ic data map
BUG=
Review URL: https://codereview.chromium.org/2064693003 .
Most of the infrastructure is fixed to work with DBC stack layout:
- register allocator allocates DBC registers with the limitation that we allocate only 20 registers and bail out if anything needs spilling (there is no use implementing spilling on DBC because registers are memory locations themselves). We should be able to bump number of CPU registers on DBC up to 256 but this requires major surgery in some parts - so I postponed this;
- lazy deoptimization is implemented, eager deoptimization is not - because we don't emit any code that actually requires it. it's a minor change to support it once we have a target;
- stack scanning respects stack maps built by registers allocator;
We bailout from all unsupported instructions.
R=zra@google.com
Review URL: https://codereview.chromium.org/1992963002 .
The original IC data are only needed when we reference them from the generated code.
EmitMegamorphicInstanceCall does not emit them into the code, but calls
EmitMegamorphicInstanceCall -> ShouldInlineSmiStringHashCode -> NumberOfChecks,
which fails if the mutator thread grows the IC data concurrently.
BUG=
R=vegorov@google.com
Review URL: https://codereview.chromium.org/1994993002 .
We can't iterate over the original IC data (like NumberOfUsedChecks does) since
the mutator thread may grow the IC data array concurrently.
This works around this issue by using the cloned IC data.
BUG=
R=vegorov@google.com
Review URL: https://codereview.chromium.org/1993673002 .
This is a cut of the work that Todd and I collaborated on in the reload branch.
In this CL, we've dropped the loader port hacks, in other words, on stack reloading in the standalone embedder does not work yet.
- [x] Support for hot reloading of isolate source code
- [x] Unit test harness and many tests
- [x] Service protocol and Observatory support
- [x] Product build does not include support for hot reloading.
R=rmacnak@google.com
Review URL: https://codereview.chromium.org/1965823002 .
This generalizes the use of switchable calls to the case where
the number of arguments checked is > 1 and improves performance
of calls to the various operator methods.
BUG=
R=rmacnak@google.com
Review URL: https://codereview.chromium.org/1888783002 .
API changes:
- Remove isolate specific Timeline APIs.
- Remove Dart_EmbedderTimelineGetTimeline callback.
- Add Dart_TimelineEvent so that arbitrary trace events can be added.
Internal changes:
- Remove isolate specific streams which were never used or controllable from outside the VM.
R=rmacnak@google.com
Review URL: https://codereview.chromium.org/1811613002 .
For now, change compilation-related flags to be regular
command line options. They won't affect the size of the
precompiled runtime anyway since the compiler is not included
there.
Also introduce a separate switch for precompiled runtime. This can be
used instead of the macro since it is a compile-time constant in the
precompiled runtime.
BUG=
Review URL: https://codereview.chromium.org/1759913002 .
- Fix bug in inlined call token positios.
- Add ProfileFunctionSourcePosition for tracking source positions within functions.
- Keep a list of ProfileFunctionSourcePosition in each function.
- Add unit test helper functions.
- Add unoptimized unit tests that verify source positions in function profiles.
- Add optimized/inlined unit tests that verify source positions in function profiles.
R=srdjan@google.com
Review URL: https://codereview.chromium.org/1758653003 .
- Add helpers for building CodeSourceMap to FlowGraphCompiler. Ensure that no duplicate pc offsets are added to a CodeSourceMap.
- Add a CodeSourceMap debug printer which incorporates inlining information.
- Add some new classifying token positions.
R=rmacnak@google.com
Review URL: https://codereview.chromium.org/1740503002 .
Move optimizations for precompilation into separate file.
Remove precompilation-specific code from flow_graph_optimizer.cc.
Add precompilation-flags: They are const in the precompiled runtime. Started moving flags to use the new flag-definition macros.
Add libdart_lib_precompiled target.
Define PRECOMPILED_RUNTIME_MACRO in more places (dart_precompiled_runtime)
BUG=
R=rmacnak@google.com
Review URL: https://codereview.chromium.org/1714743002 .
The unoptimized code cannot depend on guarded cid when using code collection
since it may be lazily regenerated and has to match the original version.
BUG=dartbug.com/25771
R=vegorov@google.com
Review URL: https://codereview.chromium.org/1693843002 .
Vyacheslav Egorov