Data properties don't contain valid data in the set field
if they are being stored in Objects. Thus we should never
access that field unless we are dealing with accessor
properties.
Change-Id: I19dcbaee7ebd042ae24387f92a93571d75ca578a
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
Make sure FunctionObjects always have the prototype property at
index 0. This way we can speed up the instanceOf operator even more,
and at the same time save 16-28 bytes of memory per FunctionObject.
Change-Id: I8527bc8f9dc7b04a9db8395b5f05bab47ddc21ce
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
The data is anyway stored in the name property of
the FunctionObject, and is not performance critical.
Change-Id: If1784b0ec6f368bc474c246bb9c2c50d5e56b689
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
The debugger should only have one breakpoint that can
be set per line. Nevertheless, we should have proper
line number information available in case we stop at
other places.
We also need a debug instruction before the return
statement, so that step out will always find a last
stopping point in the parent frame.
Change-Id: I86145fc244148f106a4a97ce69ab60b568c8dac6
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
This unifies the way we handle line numbers in the
JIT and Interpreter.
Remove the now unused lineNumberMapping code and data.
Change-Id: I1d60b1fbb77e70b531fa73d93410683e84dd1e3c
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
Saves 12 bytes per Function
Change-Id: I9a495805f9201eb6162a520ff5c2defeb73dc37a
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
Reviewed-by: Michael Brasser <michael.brasser@live.com>
* Don't create prototype Object for bindings and
signal handlers. It is inaccessible and not required.
This saves one Object-sized allocation per binding.
* Shrink the size of QQmlContextWrapper by removing
the v8 member variable.
* Shrink the size of QObjectWrapper by moving the destroy
identifier to the engine.
Change-Id: I76e84e4c0581e97a19d2e959f814ac84d9c431fa
Task-number: QTBUG-37134
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
Self-contained profiler for V4. By itself it doesn't have any
connections to qqmlprofilerservice.
Change-Id: I471a6119e07eab9c5f4712a16835be49c8886d1a
Reviewed-by: Kai Koehne <kai.koehne@digia.com>
Add an exported C-function dumping the JS stack trace which can be
invoked by a debugger with the address of an execution context
it finds in a complete stack trace.
Task-number: QTCREATORBUG-11144
Change-Id: I5314f6b24868f12d4f9dedd1c261658957e581ba
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
Move to a class hierarchy that mirrors the main classes. This will
allow moving functionality over into the Ref classes, as the current
Managed classes become mainly something that holds the data. This
is required to make objects movable by the GC.
Change-Id: I4ca88ab0e5d8c88c8dc56d51937990500a33e0d9
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
First step of removing the templates here and turning this
into a class hierarchy. This is required, so we can move all
member methods into the Ref classes and make objects movable
during GC.
Change-Id: Ie14af07fd3e72a7d84a528d0042189ff12ba21bb
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
Move commonly used variables in the ExecutionEngine
to the beginning of the struct to increase cache locality.
Keep the engine pointer in a register in the interpreter to
save one memory load per instruction.
Change-Id: If2540c66b62685701511f410aff495c0a20ca694
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
Remove SafeValue, it was used to port over to an exact GC. Since
we now have that, we can now safely merge it with QV4::Value
again. Also rename SafeString to StringValue for better naming
consistency.
Change-Id: I8553d1bec5134c53996f6b0d758738a0ec8a2e4d
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
Rename qv4value_def_p.h -> qv4value_p.h and qv4value_p.h to
qv4value_inl_p.h.
It makes more sense to have the class definition in the file
that is named after the class and move the inline methods into
a _inl file. Doing this now, as I expect we'll be needing a few
more _inl files soon.
Change-Id: Ib59e9380e9e976254c6b4369574157f39b1b5f51
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
Reduce the amount of allocations required for Arrays, and
allows freeing the array data more easily in the GC.
Change-Id: I3e3213f089c45c83a227038ce444aa60b2735b7f
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
Keep the basic methods in ManagedVTable, but have
the Object related stuff in an ObjectVTable class.
Change-Id: I9b068acf3caef813686227b8d935e7df1a7d1a6e
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
[ChangeLog][Platform Specific Changes][BSD] Fixed compile errors on
DragonFly BSD and potentially FreeBSD.
Task-number: QTBUG-35867
Change-Id: Iea90b93672c34f8a4b56e9afc4dbfb82cc993548
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
When proc is not mounted pthread_getattr_np fails, so default
to 1MB stack in getStackLimit and to exactGC in MemoryManager
Change-Id: Ic7515fd420f2d39a656808d24a3915a657722891
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
Move the type flag into the vtable to free up these
bits in the Managed class, and not have to set them
at object construction time.
As we often need to know whether a Managed object is a
Object, FunctionObject or String, add some bitflags to test
for these to the vtable.
Change-Id: I7d08ca044544debb307b55f124f34cb086ad9e84
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
vtable changes need to happen when the internal class
is being constructed, not later on.
Change-Id: Ibb9515745c9fc3507a5a90b4cc50a33e2e0d3f99
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
Bionic pthreads reports too small stack size for main thread.
Change-Id: I3d33229e76101a847309c723d534844ffb2d2042
Reviewed-by: Andy Nichols <andy.nichols@digia.com>
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
Encapsulate accesses to the current context, and rework
the way we push and pop this context from the context
stack.
Largely a cleanup, but simplifies the code in the long term
Change-Id: I409e378490d0ab027be6a4c01a4031b2ea35c51d
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
Remove all the calls to setVTable that were in performance
critical parts of the code. This now brings performance
back to the level we had with the vtable inlined in the
Managed objects.
Change-Id: I76317cc5c53b5b700d1d3883b954407142a4c424
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
This saves one pointer per object, and willmake other optimizations
easier in the future.
Change-Id: I1324cad31998896b5dc76af3c8a7ee9d86283bfe
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
The wrapper method for Function::code() was still there
from the times we used C++ exceptions. It's not needed
any more, so get rid of it.
Change-Id: I2ec25fbca71eeef9d7a94a38b5adfa42e4de3a84
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
Instead of storing the current instruction pointer in the
ExecutionContext, we might as well directly store the current
line number there.
Leads to simpler code, works cross platform and should also
be faster.
Change-Id: Ifb7897cf8dbe8a962505fe876aa3ed43283ebb06
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
This finally gives proper memory management for ExecutionContexts.
So far they had been garbage collected but where still allocated
using standard malloc/free(). This allows us to collect the
contexts faster and speed up context creation.
Change-Id: I02e642391d55eaa59ab3f4c2720a2ac71259caf4
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
Turns out pthread_get_size does not return the correct size when it is
called from the main thread, so to workaround you call getrlimit instead
Without this change, most QML applications are broken on iOS.
Change-Id: I9a61494de26caa3d7be7e46a991e6d6d0514ce17
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
pthread_getattr_np is not available there, so we use the default implementation (1MB stack limit) for now.
Change-Id: Ia1f3cbbcd846998b688541c0b576ef42f22ac8b6
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
Setup limits for both the C and the JS stack, and check
them before entering functions. If we run out of space,
throw a RangeError exception.
Be careful and recheck the stack bounds when things go
outside. This catches the case where the engine got
moved to another thread changing the stack boundaries.
Windows currently uses an unsafe fallback implementation,
this needs to be fixed later on.
Task-number: QTBUG-34568
Change-Id: I22fbcbec57b28f9cc8a49e12f1cc6e53e4f07888
Reviewed-by: Erik Verbruggen <erik.verbruggen@digia.com>
Reviewed-by: Gunnar Sletta <gunnar.sletta@digia.com>
Currently missing, but coming in subsequent patches:
- evaluating expressions
- evaluating breakpoint conditions
Change-Id: Ib43f2a3aaa252741ea7ce857a274480feb8741aa
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
Get rid of the SimpleCallContext, instead simply
use the CallContext data structure, but don't
initialize the unused variables.
Change-Id: I11b311986da180c62c815b516a2c55844156d0ab
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
Don't use recursive function calls anymore. Instead, push marked
objects onto the JS stack, and then pop them off when their children
are being marked.
Should reduce stack memory usage, and improves performance by ~5%.
Change-Id: I2d37d97579144fcba87ec8e9fd545dd220c01fbb
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
This class is not required anymore to generate stack traces, as
we now store the required information in the JS context stack.
Change-Id: I3893c805ca89dda70efde07fdd120e7dfaf3639f
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
This method was throwing a C++ exception in the old exception
handling and is not needed anymore.
Change-Id: If67696cdbd260225ae60720a1035941fe7e1e650
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
Don't write to objects if we have a pending exception to
avoid any side effects.
Change-Id: I9f93a9195a652dbae7033cc6ebb355d5d86e9b5e
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
We don't want to check for exceptions after every single
line on our runtime methods. A better way to handle this
is to add the check in all methods that have direct side
effects (as e.g. writing to a property of the JS stack).
We also need to return whereever we throw an exception.
To simplify the code, ExecutionContext::throwXxx methods now
return a ReturnedValue (always undefined) for convenience.
Change-Id: Ide6c804f819c731a3f14c6c43121d08029c9fb90
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
Replace all try/catch statements used when parsing
with checks for engine->hasException.
Change-Id: I4493cb600d5a3eb095c2003bb88bd031403e47c9
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
Simplify the generated code. Add a special block to catch
exceptions thrown inside a catch() statement.
store the exception on the stack when entering finally and
rethrow it at the end. This ensure correct behavior for
break/continue/return statements inside finally.
Don't check for exceptions after calling push_catch_scope
and pop_scope in the JIT'ed code. This can lead to infinite
loops when throwing inside an exception handler.
Change-Id: I67e9325794e2fd25b0773b21e02fbaadb43faab0
Change-Id: Ic1ea9c0c43eec1d49177dc1ab4552a1da04e96fe
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
Start the work to remove c++ exceptions from our JS
exception handling. Rather rely on engine->hasException.
Check the flag after we return from any runtime call in the
JIT.
Implement new try/catch handling code in qv4codegen and
for the JIT that doesn't rely on exceptions. As an added
bonus, we can remove the Try statement in the IR.
Change-Id: Ic95addd6ae03371c43c47e04cac26afdce23a061
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
QQmlError is public API and shouldn't expose an internal
method.
Change-Id: I7caf06af9340fefec5c96103395fe74acbf19497
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
Avoid catch (...) with re-throw as it turns that this is very slow because it
throws a new exception and the unwinder starts from scratch. Instead use stack
allocated objects and cleaning destructors to restore state before continuing
with the propagation of exceptions.
Change-Id: I6d95026bcd60b58cb6258a9dae28623a46739532
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
It turns out that in QML it is not unusual that during early binding
evaluations due to the undefined order, the evaluation tries to look up
properties in objects that aren't initialized yet and thus exceptions are
thrown. Eeach thrown exception saves a stack trace, which is expensive to
generate when using the JIT, as it does full stack unwinding.
This patch implements a more light-weight approach by storing the instruction
pointer in the context before leaving JIT generated code.
Change-Id: I95e1cfd01179247dfc2c1df949828f474a23161b
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
Properly protect them through Scoped values.
Change-Id: I5a0a1d5580d55ecff493419baa8959751a65f1d3
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
This makes pretty much all test cases pass with exact
garbage collection.
Change-Id: Ia874e3c17c3984afb7cfe370f9bd3ad8fe46699a
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
The code in the Exception class operates entirely on the engine's data,
so move it into ExecutionEngine instead. This eliminates the need for
a QV4::Exception class and catches and old code that tries to still do
catch (Exception &) instead of catch (...)
Change-Id: Ie608bec6af652038aca6c9423c225a4d7eb13b39
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
This patch changes the exception handling API in the engine slightly, encapsulating
any use of direct throw statements and catch blocks with concrete types. In the future
we need to be able to change the way these are implemented, in order to ensure that
the correct stack unwinding code is triggered for throw and re-throw.
This patch separates the C++ exception object thrown from the V4 exception
(that includes value, throwing context pointer) and stores the latter inside
the engine.
In order for that to compile, ExecutionEngine::StackTrace and StackFrame had to
move into the QV4 namespace directly.
In addition the syntax for catching exceptions changes from
try {
...
} catch (QV4::Exception &ex) {
ex.accept(context);
QV4::ScopedValue exceptionValue(scope, ex.value());
}
to
try {
...
} catch (...) {
QV4::ScopedValue exception(scope, context->catchException());
}
Context::catchException() checks if there's a "current" exception in the engine,
and if not assumes that we caught an unrelated exception and consequently re-throws.
partiallyUnwind() is also gone and replaced with rethrowException(), in order to
encapsulate the re-throw.
Lastly, in the future nesting try/catch blocks isn't going to be possible due to
limitations in the common C++ ABI with regards to foreign exceptions.
Change-Id: Ic81c75b057a2147e3176d8e0b4d326c14278b47d
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
Fix some usages in qv4engine, and fix return types
in methods in qqmlxmlhttprequest.
Change-Id: I3d6225ca01bf7ea77fcc424914c8392bb6c3a454
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
Remove Value::fromString(String *), and make
Encode safe against encoding raw Managed * pointers.
Change-Id: Ibca4668e1cbeaf85c78169d14386281659d33ef6
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
This will simplify finding the remaining direct usages of
QV4::Value that need fixing.
Change-Id: I223099727436d5748027c84c53d9dfc4028e38ed
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
Don't use unprotected Values in the API anymore.
Change-Id: I8851628227fca374de24701bc8ee0908b5ae3923
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
While objects are being constructed, we don't have a reference to them
on the JS stack yet. So the constructor needs to protect itself against
being collected by putting the this object onto the JS stack.
Added an environment switch MM_EXACT_GC to test exact garbage
collection.
Change-Id: Ie37665a954de800359c272ffbebbe1488e7a8ace
Reviewed-by: Gunnar Sletta <gunnar.sletta@digia.com>
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
Change-Id: I4fda83a0832760c277e629d4e658da718c0bf92b
Reviewed-by: Gunnar Sletta <gunnar.sletta@digia.com>
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
also store "toString" and "valueOf" as identifiers
in the engine and fix two places where we compared
strings the wrong way.
Change-Id: I70612221e72d43ed0e3c496e4209681bf254cded
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
Added some convenience typedefs (StringRef, ObjectRef, ReturnedString,
ScopedString, ...)
Used StringRef in newBuiltinFunction() for testing.
Cleaned up the duplicated code for thrower functions.
Change-Id: I7b7676690cbe70d9eabb0a5afd0d922f0be3aefd
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
Add some more convenience in the helper classes
in qscopedvalue_p.h
Make accesses to CallData safer, and change
ExecutionEngine::newObject() to return a safe
pointer.
Change-Id: I980909754ce9681cf6faa1355bab3a1e5d6dd186
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
Now that we store them per QML file, they live in the type cache and
can outlive the engine. Therefore the engine needs to free to unlink
any remaining units upon destruction. This needs to be done after the
"death" of the memory manager, which is likely to sweep away any
function objects that also hold a reference to the compilation units.
Change-Id: I3968d5995289e8d2bc1e3abbb1f8be88a0ab4e03
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
Add a Returned<T> that we can return instead of raw pointers
to Managed objects.
Start using the Returned<T> for a few methods.
Also clean up all our classes to use the Q_MANAGED macro instead
of manually defining their vtable.
Change-Id: I0a2962e47f3de955cd2cd8474f8f3fcc9e36d084
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
This is the correct abstraction for a builtin method.
Time to get rid of the 'old' in the name.
Change-Id: Ia386d2cc8ef0e6c269ab24da6fc8ebfac52d7f69
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
Surround all calls into generated code with a
try {} catch {} statement that resets the jstack
to the correct position.
Like this we properly unwind the js stack in all cases, and
can also use stricter assertions in our ScopedCallData, etc.
classes to check that the stack is healthy.
Change-Id: I7ca03e06ea55007be683305d9c2a6898cf5fc689
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
First step towards being able to do an exact GC.
Create a stack for JS Values that is separate from the C++
stack.
Use the stack for generated methods (masm and moth).
Change-Id: I80ac0e5b5d86439dda5e9ea2b21fa0c57d8aef22
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
By setting this variable, the interpreter will always be used, even when
the JIT is available.
Change-Id: I66a439b649fd22e0d10be4f07f1e7be392b012f5
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
This speeds up the v8 regexp benchmark by a factor 2.5 :)
Change-Id: Ibd6b18ee28181aa712429cbec4598984e0c69820
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
Rather use the correct internalClass directly when constructing
the objects.
Change-Id: I8e916f1ce8f83d291c08ca6332fe85b1f57b90b5
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
The prototype is actually the same for most objects. By
moving it into the internal class, we can save 8 bytes
per object, as well as allowing for some future
optimizations.
Also fix a bug in the implementation of the Error
prototype objects.
Change-Id: I4d4b641055f644a9b088f27be34bfdb0446279b7
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
This avoids one indirection when calling the methods and cleans up
the engine a bit.
Change-Id: I426f41e23f6a7262af95b9807b00920530fef642
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
Choose whether we use a stack based context for a function, when
the actual closure is generated, not at call time. This speeds up
function calling for leaf functions.
Change-Id: Ibcbf3acb5610a7f59b6474e982122df03c1c5298
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
At the moment we collect a lot of compilation units (one per binding
expression!), which for long running QML accumulates and creates a horrible
performance when trying to retrieve back traces. There is work in progress
to reduces the number of units down to one per QML file, and then the
fixed sorted QVector might proof to be a more efficient data structure
for the lookups.
But until that code lands, this patch proposes to use a QMap instead for the
time being, that tracks all functions. This brings down the qtquickcontrols
auto-test run from 2.5 minutes to just under a minute on my machine.
Change-Id: I45bf609055877081daa984de90f291a030f2f24f
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
QQmlError provides the same functionality, so let's rather
use that where required. Remove the dependency of
codegen onto the ExecutionContext that was only
required for error handling.
Change-Id: Ib0b61c0e138f89ff989c32996c93c339e4b62223
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
Lars Knoll
Michael Brasser
Ulf Hermann
Friedemann Kleint
Simon Hausmann
Alex Blasche
Albert Astals Cid
aavit
Andy Nichols
Fabian Bumberger
Erik Verbruggen
Erik Verbruggen