I collected statistics for the sizes and capacities of growable arrays which are promoted to old-space or survive an old-space gc when running dart2js and Fasta. For these applications, the vast majority of arrays stay empty. More than half of the total object size of promoted backing arrays is backing for empty growable arrays.
Furthermore, since the overhead for an array is 3 words (header, type parameters and length), and object sizes are rounded up to an even number of words, we waste one word for all even-sized arrays.
This CL changes the growth strategy so that empty growable arrays are created with a shared, zero-sized array as backing, avoiding the allocation of a backing array if no elements are added. When the array needs to grow, it starts out at 3 and grows to double size plus one each time: 7, 15, 31, ...
A few places in the VM code need to handle these shared, zero-sized arrays specially. In particular, the Array::MakeArray function needs to allocate a new, empty array if its result is to be returned to Dart code.
Benchmarks suggest that the change improves memory usage by a few percent overall and does not significantly affect run time.
BUG=
R=erikcorry@google.com
Review-Url: https://codereview.chromium.org/2949803002 .
Use that to sync variables to predefined stack
slots inside of exception handler, not before
each MayThrow() instruction.
In JIT mode we can avoid storing extra metadata
by using deopt info instead.
Introduce caching of metadata and
exception handler based on PC.
R=fschneider@google.com, vegorov@google.com
Review-Url: https://codereview.chromium.org/2734323003 .
Use that to sync variables to predefined stack
slots inside of exception handler, not before
each MayThrow() instruction.
In JIT mode we can avoid storing extra metadata
by using deopt info instead.
Introduce caching of metadata and
exception handler based on PC.
R=fschneider@google.com
Review-Url: https://codereview.chromium.org/2640853002 .
Original CL: https://codereview.chromium.org/2692803006/
Original commit message:
Tracking the awaiter return call stack:
- [x] Each async function closure now knows who is awaiting on their
return. This is effectively the asynchronous equivalent of the 'frame pointer'.
- [x] Each async* function closure now knows how is listening on their
stream. This is effectively the asynchronous equivalent of the 'frame pointer'.
Detecting uncaught exceptions in async functions:
- [x] Code object keeps a map from :await_jump_var to token position
- [x] Exception Handlers keep track if they are generated (as part of compilation) or directly from user code
- [x] Debugger maps :await_jump_var to a specific try index
R=johnmccutchan@google.com
Review-Url: https://codereview.chromium.org/2725623003 .
Tracking the awaiter return call stack:
- [x] Each async function closure now knows who is awaiting on their
return. This is effectively the asynchronous equivalent of the 'frame pointer'.
- [x] Each async* function closure now knows how is listening on their
stream. This is effectively the asynchronous equivalent of the 'frame pointer'.
Detecting uncaught exceptions in async functions:
- [x] Code object keeps a map from :await_jump_var to token position
- [x] Exception Handlers keep track if they are generated (as part of compilation) or directly from user code
- [x] Debugger maps :await_jump_var to a specific try index
Fixes#27242R=rmacnak@google.com
Review-Url: https://codereview.chromium.org/2692803006 .
Also fix stack trace collection to always include invisible frames. It can happen that a visible function is inlined into an invisible function, and we don't expand inlined frames until we print a stack trace.
dart2js product x64:
compile time: 12.459s
VMIsolate(CodeSize): 152292
Isolate(CodeSize): 3343117
ReadOnlyData(CodeSize): 3728928
Instructions(CodeSize): 8677600
Total(CodeSize): 15901937
->
compile time: 14.195s (+13%)
VMIsolate(CodeSize): 174034
Isolate(CodeSize): 3892418 (+16%)
ReadOnlyData(CodeSize): 5036320 (+35%)
Instructions(CodeSize): 8682624
Total(CodeSize): 17785396 (+12%)
R=asiva@google.com
Review-Url: https://codereview.chromium.org/2687143005 .
CodeSourceMap now contains a small bytecode that describes changes to the token position and inlining stack as one moves through a function. This information is used by the disassembler to display inlining comments and by the profiler to properly blame ticks against source code.
Slightly shrinks the total size of code metadata; on dart2js compiling hello world, the final heap is
before 59.6MB total, 12.8MB List, 1.1MB CodeSourceMap
after 59.4MB total, 12.1MB List, 1.5MB CodeSourceMap
To be filtered and included in AOT code to expand inline frames in stack traces.
R=johnmccutchan@google.com
Review-Url: https://codereview.chromium.org/2670843006 .
- 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 .
- Use TokenDescriptor instead of intptr_t for all token positions.
- Use TokenDescriptor in raw_object instead of int32_t.
- TokenDescriptor is a POD with an int32_t (this shrinks the size of AST and IR nodes by 32-bits on 64-bit architectures).
There are some cleanups I plan on doing as a follow up CL:
- Replace TokenDescriptor::value() with TokenDescriptor::TokenPos()
R=iposva@google.com, rmacnak@google.com
Review URL: https://codereview.chromium.org/1644793002 .
Replace always_optimize_ and allow_recompilation_ with
FLAG_precompilation and !FLAG_precompilation.
They are equivalent and set only once according to the
--precompilation flag
BUG=
R=rmacnak@google.com
Review URL: https://codereview.chromium.org/1657153002 .
The function's code pointer is either a lazy-compile stub
or the compiled code. The function is compiled on invocation.
There is no need to check for null when invoking a function. This means
one check less when invoking functions via a stub. For this I added a LazyCompile
stub.
Additional cleanup in Stackmap and PcDescriptors (remove unused code pointer,
and changed smi-lengh into a raw integer).
R=zra@google.com
Review URL: https://codereview.chromium.org//203523011
git-svn-id: https://dart.googlecode.com/svn/branches/bleeding_edge/dart@34643 260f80e4-7a28-3924-810f-c04153c831b5
which have different implementations of 'operator=' and 'operator^='.
In the case of FINAL_OBJECT_IMPLEMENTATION we do not do the vtable setting
in these methods (Note the |= operator functionality is now subsumed into
the new implementation of "operator^=")
Review URL: https://codereview.chromium.org//12052033
git-svn-id: https://dart.googlecode.com/svn/branches/bleeding_edge/dart@17491 260f80e4-7a28-3924-810f-c04153c831b5
Add flag --deoptimize-alot, which will lazily deoptimize all live optimized frames. Currently only the presence of deoptimization information is checked.
Add kDeoptAfter deoptimization point, which is the continuation for lazy deoptimization, after a call. Rename kDeopt to kDeoptBefore.
Removed a tests case that called into a native without properly setting up a Dart frame (Ok-d by Siva).
Native functions are not optimizable.
TODO: Split deoptimization information from DeoptimizationStubs. Check for redundant PcDescriptor information (what can be merged, especially at calls).
Review URL: https://chromiumcodereview.appspot.com//10885039
git-svn-id: https://dart.googlecode.com/svn/branches/bleeding_edge/dart@11642 260f80e4-7a28-3924-810f-c04153c831b5
Zachary Anderson