Lars Knoll [Thu, 14 Mar 2013 13:03:04 +0000 (14:03 +0100)]
Remove the parent pointer from ExecutionContext
It's cleaner to have an explicit stack of contexts
in the Engine, esp. as the global context can get
pushed onto the stack several times. This avoids
an ugly hack in eval() where we created a 'copy'
of the global context.
Change-Id: I3936443fba6c1829a60a8e0e9a106ec75293274f
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
Lars Knoll [Thu, 14 Mar 2013 11:59:11 +0000 (12:59 +0100)]
Move all context management methods into the ExecutionEngine
Centralizes this functionality, so it can be more easily handled
and modified in the future.
Change-Id: I733fb791b816a117343615ddd7d7a754f42d3dbb
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
Lars Knoll [Thu, 14 Mar 2013 11:48:00 +0000 (12:48 +0100)]
Rename qmljs_engine.* to qv4engine.*
Change-Id: I452fc11972c802acc14e0a3358d8c408b6dbded8
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
Lars Knoll [Thu, 14 Mar 2013 11:43:27 +0000 (12:43 +0100)]
Rename qmljs_environment.* to qv4context.*
Change-Id: I6fe62b501803422f78b47a87c55e4278820725aa
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
Lars Knoll [Fri, 8 Mar 2013 08:38:46 +0000 (09:38 +0100)]
Optimise put and defineOwnProperty a bit
Change-Id: Ie944ce76a744c2bbadce20da28ceb9104f9c95c2
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
Lars Knoll [Fri, 8 Mar 2013 08:17:20 +0000 (09:17 +0100)]
Remove unused method
Change-Id: Ie96da445a7c94e43db1d293f254315eed6af6713
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
Lars Knoll [Fri, 8 Mar 2013 08:17:11 +0000 (09:17 +0100)]
Remove unused variable
Change-Id: I83cab99dbb6905b4f480cf1584cdb96246fd987d
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
Lars Knoll [Wed, 13 Mar 2013 21:50:00 +0000 (17:50 -0400)]
Implement another method in the v8 API
Change-Id: I54918914a8c9b395211ccbff891a90ab069be4a5
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
Lars Knoll [Wed, 13 Mar 2013 21:41:42 +0000 (17:41 -0400)]
Message is refcounted
Change-Id: Ic5097ea1eae6503f906157b3d4bf51a028c94280
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
Lars Knoll [Wed, 13 Mar 2013 21:39:21 +0000 (17:39 -0400)]
Catch some more exceptions in the v8 API
Change-Id: I23428f7dce0771576ae1243812eeed7a534dc833
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
Lars Knoll [Wed, 13 Mar 2013 21:09:18 +0000 (17:09 -0400)]
Remove some debug output
Change-Id: I9178da579194479eac8996290cda1ae056720b15
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
Lars Knoll [Wed, 13 Mar 2013 21:05:34 +0000 (17:05 -0400)]
Cleanup and unify context creation
* Always create contexts on the heap. When embedding native methods
we can't create contexts on the stack anymore without running the
risk of them being used in some scope chain.
* Unify context creation for call contexts, share the code
* Add a hack for indirect calls to eval and create a new 'fake'
global context there, so we don't mess up the context stack (it
broke badly when unwinding exceptions before).
Change-Id: I5804224dc26582f24ec79518639ceb13a8a3e967
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
Lars Knoll [Wed, 13 Mar 2013 21:03:51 +0000 (17:03 -0400)]
Pass the current context to evalCall
Passing the root context will seriously mess up our
JS call stack.
Change-Id: I513a9eb61bdc24b7a4542bc5f8556bf927af6c75
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
Tor Arne Vestbø [Wed, 13 Mar 2013 13:15:50 +0000 (14:15 +0100)]
Fix a few warnings
Change-Id: I82efe66a28a76eb64b1254a7a9c6c60a0fce6228
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
Lars Knoll [Tue, 12 Mar 2013 19:34:09 +0000 (15:34 -0400)]
Be conservative and assume that functions need activation
A function that doesn't need it's context allocated on the heap
should state so explicitly. If not it's better to assume that we
need a heap allocated context.
This fixes a crash on the QML clocks demo.
Change-Id: I0a67e6ed38c99e7dd75dd03d6cc2dda10d8b5ad7
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
Lars Knoll [Tue, 12 Mar 2013 19:06:55 +0000 (15:06 -0400)]
Remove unused method
The method was only used to destroy the global context
and it's usage was wrong and needless there.
Change-Id: If97a0c94862b5f7f7ac4dff5ccb58dc2ef8cec4b
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
Lars Knoll [Tue, 12 Mar 2013 18:49:13 +0000 (14:49 -0400)]
Add valgrind support to our garbage collector
Tell valgrind about the memory we allocate and free in the
garbage collector. Also mark stack variables as initialized
before walking the stack.
This avoids all valgrind warnings related to GC and gives
proper warnings if we should attempt to access garbage collected
memory.
Change-Id: I7e923a163c25e7a5b4409b1b9c2191f314675f2d
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
Erik Verbruggen [Tue, 12 Mar 2013 13:18:04 +0000 (14:18 +0100)]
Add debugging function to dump Value contents.
Change-Id: I45e1d1d13fe7dee22809734595a3fbbfd03b123d
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
Erik Verbruggen [Tue, 12 Mar 2013 11:46:44 +0000 (12:46 +0100)]
Moved temp compression to codegen to use for all backends.
Change-Id: I3afba14741f18782e71210b14d9d091fe0bebc51
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
Erik Verbruggen [Wed, 6 Mar 2013 10:52:40 +0000 (11:52 +0100)]
Added the constant propagation optimization.
Change-Id: I4fe680ff334a13085df425ad2badad08e548f21f
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
Erik Verbruggen [Fri, 1 Mar 2013 13:27:50 +0000 (14:27 +0100)]
Fix & re-enable Temp compression for the interpreter.
Change-Id: Ie9e767d36d3d810bf0d2a173d04954277041f68f
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
Simon Hausmann [Tue, 12 Mar 2013 09:00:06 +0000 (10:00 +0100)]
Fix crash when calling Object.create(null)
O is allowed to be null here, so don't use objectValue() unconditionally
and instead use asObject().
Change-Id: I3e868d3486d2de109c94aef8a7de16110a3a8326
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
Simon Hausmann [Mon, 11 Mar 2013 12:02:59 +0000 (13:02 +0100)]
Add support for ARM exception unwinding
This is used only on Linux/ARM. iOS uses setjmp/longjmp.
Change-Id: I06f907ce702c1b8e06a50ac31b5486d11ef392ec
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
Simon Hausmann [Sat, 9 Mar 2013 11:51:12 +0000 (12:51 +0100)]
Make it possible to debug exception handling with libunwind
Run qmake with CONFIG+=debug-with-libunwind.
Change-Id: I02f5820a1c8ee9132dac521d182007800c250103
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
Erik Verbruggen [Mon, 11 Mar 2013 11:00:13 +0000 (12:00 +0100)]
Inline the two most common parameter fetch cases.
By inlining temp and local fetching, the interpreter is 30% faster
on x86 on crypto.js. x86_64 gains about 35%.
As the comment in the source already indicates: this might need some
tweaking for QML.
Change-Id: I9aa51036d42d2749f9aeec39a479d9e5f6233a23
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
Erik Verbruggen [Mon, 11 Mar 2013 10:29:01 +0000 (11:29 +0100)]
Mark asm math functions as const.
This allows gcc to inline slightly better, gaining 1% in the
interpreter.
Change-Id: Ia0ebc9eb698f12de4ff7860e9f62a62dbd782ccd
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
Lars Knoll [Sun, 10 Mar 2013 17:43:06 +0000 (18:43 +0100)]
Catch more exceptions in the v8 API layer
Change-Id: I7af3abbbf8cae86abe31a0d55d3d4c7e8c91a803
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
Lars Knoll [Sun, 10 Mar 2013 16:53:29 +0000 (17:53 +0100)]
Write properties back to the qml Object
The object acts as a replacement for the QML scope chain.
We need to write values to it even if __hasProperty__ returns
false.
Change-Id: I97856ab8cc3a7a728bb244f40b9f230a415b5a33
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
Lars Knoll [Sun, 10 Mar 2013 13:59:44 +0000 (14:59 +0100)]
Add a proper constructor to v8::FunctionTemplate
Change-Id: I59b414484c05d8292ee7300860ef9e0611da49e6
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
Lars Knoll [Sun, 10 Mar 2013 13:58:35 +0000 (14:58 +0100)]
Properly mark objects at the beginning of a chunk
Objects at the beginning of a chunk didn't get marked
properly as qLowerBound in the mark() method returned an
even number for them. Fix this by subtracting one to the
beginning of the chunk.
Change-Id: I1abd7e24f5b7f374fc0776b7d891a3124fee7270
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
Lars Knoll [Sat, 9 Mar 2013 22:18:25 +0000 (23:18 +0100)]
Fix v8::Persistent
Implement it in a similar way as PersistentValue
Change-Id: If4adba61a8bfedce657f07ee24662c3b13384fd9
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
Lars Knoll [Fri, 8 Mar 2013 21:52:55 +0000 (22:52 +0100)]
Add exception handling support to the v8 API
Not all calls are yet propertly protected by try/catch
statements, but it should work in principle.
Change-Id: I5ddaf433b0a454106dc2743fe2ccfc5822fb3926
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
Simon Hausmann [Fri, 8 Mar 2013 14:48:50 +0000 (15:48 +0100)]
Implement more missing bits & pieces including a fake "QML mode"
Change-Id: Ifbbc5004ada03b6efa7a9884938ff5d59220c5fb
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
Simon Hausmann [Fri, 8 Mar 2013 13:11:42 +0000 (14:11 +0100)]
Fix name lookup for named function expressions
Change-Id: Ia36b2b5c5b40475450fe369c7d6cb5e3965a4488
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
Simon Hausmann [Fri, 8 Mar 2013 12:02:26 +0000 (13:02 +0100)]
Implemented a whole more API...
Change-Id: Ica73c21e58b406abd6310c59ec0f541ae2cc9155
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
Simon Hausmann [Fri, 8 Mar 2013 10:47:26 +0000 (11:47 +0100)]
Implemented invocation callback on FunctionTemplate
Change-Id: I718f10f28a053eda78e086407183706e8fad317b
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
Simon Hausmann [Fri, 8 Mar 2013 09:53:41 +0000 (10:53 +0100)]
Implemented v8::Arguments
Change-Id: I02f7072c13c319b1bdf77471360308bd0d0d3d2a
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
Simon Hausmann [Fri, 8 Mar 2013 09:47:18 +0000 (10:47 +0100)]
Implemented Template::Set
Change-Id: I237fdf6acdfad9805306e4fc5f437457efa8eb7f
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
Simon Hausmann [Fri, 8 Mar 2013 09:41:38 +0000 (10:41 +0100)]
Minor API change for easier implementation
Change v8::Template::Set to take Handle<Value> instead of Handle<Data>
and remove incorrect Data inheritance from Value
Change-Id: Iaa71d4437c9913c7a79bc5322f05dafce6e3c89c
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
Simon Hausmann [Fri, 8 Mar 2013 09:34:19 +0000 (10:34 +0100)]
Various fixes
* Make V4V8 object a template in preparation for future changes
* Fix handle <> local conversions
* Implemented v8::External
Change-Id: I27b57c767a31393e1b8b05f3150c16158b178f0d
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
Simon Hausmann [Thu, 7 Mar 2013 15:35:46 +0000 (16:35 +0100)]
Remaining implementation of ObjectTemplate
Change-Id: I299b252346b2c5696fb11eb7b8dc527d94dbaf3b
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
Simon Hausmann [Thu, 7 Mar 2013 13:59:25 +0000 (14:59 +0100)]
First implementation of ObjectTemplate::SetAccessor
Change-Id: Iee403e3b542423726f95a0b516a4e2d5938ea2ed
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
Erik Verbruggen [Fri, 8 Mar 2013 11:17:33 +0000 (12:17 +0100)]
Fix use/def for exception var.
A try defines the exception var, because it is the first time it is
mentioned. It might also use it, but that could be seen as
an initialization.
Change-Id: Ibacd12edc02f99c8049003655373122d7916978b
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
Erik Verbruggen [Fri, 8 Mar 2013 09:44:04 +0000 (10:44 +0100)]
Fix liveness analysis.
When a move has a temp that is either a scoped local or a formal,
then instead of "returning" from the visit, the source expr still has
to be visited. It might contain a use.
Change-Id: Ibd54fdc3488b1348e63ecd5a0c1b1036ae111c8a
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
Erik Verbruggen [Fri, 8 Mar 2013 08:57:33 +0000 (09:57 +0100)]
Fix MSVC build.
Suppress error for methods that have to return a value, but are marked
as Q_UNIMPLEMENTED by adding a Q_UNREACHABLE.
Change-Id: I2f7d3027fd5a8949b7c330d8cab3f3f095e11450
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
Erik Verbruggen [Fri, 8 Mar 2013 08:56:44 +0000 (09:56 +0100)]
Compile fixes for MSVC.
When a class is dllexported, the methods cannot also be marked as such.
They are exported automatically. The error message is:
d:\dev\v4vm\src\v4\qv4v8.h(1525) : error C2487: 'New' : member of dll interface class may not be declared with dll interface
Change-Id: Ib44ee401e35c723b20c8c9f2dceab02736c77a3f
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
Erik Verbruggen [Thu, 7 Mar 2013 16:00:17 +0000 (17:00 +0100)]
"Fix" llvm build.
In a way that it the compiler doesn't complain anymore.
Change-Id: I877b1a051b2ffcbd63a04cf219733b2b21cb9128
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
Simon Hausmann [Thu, 7 Mar 2013 11:53:34 +0000 (12:53 +0100)]
Fix exception handling on Mac OS X
Re-introduce the old register_frame and deregister_frame approach
on Mac OS X until we have something faster.
Change-Id: Idf524857c91175535184b29c11265c7d26f064a7
Reviewed-by: Erik Verbruggen <erik.verbruggen@digia.com>
Erik Verbruggen [Wed, 6 Mar 2013 10:48:05 +0000 (11:48 +0100)]
Simple but effective dead-assignment removal.
At the moment we only generate a dead assignment when initializing the
return value to undefined (and subsequently assign to it). With upcoming
patches, esp. constant propagation, there will be more dead assignments.
Change-Id: I423ed77fb0693ca6c2855d1e0200e6a205d01f4f
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
Lars Knoll [Thu, 7 Mar 2013 12:41:21 +0000 (13:41 +0100)]
Fix String::get and start using it
Change-Id: I80cc7533effad2a1fa3fe60c2596ed9d27e843c9
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
Lars Knoll [Thu, 7 Mar 2013 12:14:34 +0000 (13:14 +0100)]
And use the vtbl methods for delete as well.
Change-Id: I7632382ccb8aa2f8ae733f638bb42eddbc739c6a
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
Lars Knoll [Thu, 7 Mar 2013 11:57:49 +0000 (12:57 +0100)]
Implement Object::query/queryIndexed
Change-Id: I7feea63d0e9eb7c1e633536aea4c4a4dbd03b462
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
Lars Knoll [Thu, 7 Mar 2013 11:55:02 +0000 (12:55 +0100)]
Cleanup put and properly implement it for strings
Change-Id: I10dda09c8a087bc7ef34c0315c698e08c35ec28d
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
Simon Hausmann [Thu, 7 Mar 2013 07:40:36 +0000 (08:40 +0100)]
Fix build with clang
It suggests a template keyword here to make it clear that As() is template dependent
Change-Id: I0a228ad51b7a7a09e3774465f7118e3e6e247e93
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
Lars Knoll [Thu, 7 Mar 2013 11:25:59 +0000 (12:25 +0100)]
First part of moving get/put etc. into the Managed vtbl.
Change-Id: I6fe14b02205901dbffa25c6c1b4883fb99586417
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
Lars Knoll [Thu, 7 Mar 2013 09:57:48 +0000 (10:57 +0100)]
Make Object::hasProperty() inline
Simply call getPropertyDescriptor instead
Change-Id: I9e156a45dd10bf250fa156820ec2f3d5bbe80bbc
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
Lars Knoll [Thu, 7 Mar 2013 09:31:29 +0000 (10:31 +0100)]
Implement v8::AccessorInfo
Change-Id: I72598ad9941b35e7268f6a90c26c91f37593bed4
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
Lars Knoll [Thu, 7 Mar 2013 09:26:03 +0000 (10:26 +0100)]
Remove v8 internal namespace
Change-Id: I78d0ddb1b5b790cb74cfa2d115e7dd47ebb30ab1
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
Lars Knoll [Wed, 6 Mar 2013 21:21:26 +0000 (22:21 +0100)]
A few more API methods implemented
Change-Id: I4168ad5bb934a74e6957969417fdcf491a7374a1
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
Lars Knoll [Wed, 6 Mar 2013 20:43:58 +0000 (21:43 +0100)]
More v8 API
Change-Id: If62d4a0dbe8f59d62bef2cce0bb9002be47957bc
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
Lars Knoll [Wed, 6 Mar 2013 20:28:03 +0000 (21:28 +0100)]
Implement more of the v8 API
Change-Id: Ibc8feb95b9473712105cf79e98fc0cef8c3b0fe7
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
Simon Hausmann [Wed, 6 Mar 2013 19:04:47 +0000 (20:04 +0100)]
Fix slowness in unwind table registration
__register_frame and __deregister_frame are very slow when called
with many functions many times, like in some of the tests that use
a lot of eval(). Change the approach to interposing a function symbol
that is called in libgcc whenever a PC -> FDE lookup is required.
This also changes the code to create the unwind info on-demand.
Change-Id: I99dfcc921b087831ad00f79bad9e7bcfb0cc8c33
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
Simon Hausmann [Wed, 6 Mar 2013 19:04:43 +0000 (20:04 +0100)]
Make sure that the value stored in the Exception object is not garbage collected
Change-Id: I7e3c4e0d1d9a23a54922a3254962ac2dc42bf67b
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
Simon Hausmann [Wed, 6 Mar 2013 19:04:39 +0000 (20:04 +0100)]
Get rid of 'Value ExecutionEngine::exception' member
Instead the JS exception value is now part of the C++ Exception object.
This also allows getting rid of some run-time functions.
Change-Id: I43ff773cacd5e925ba96601f3633ccf3b62273be
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
Simon Hausmann [Wed, 6 Mar 2013 19:04:36 +0000 (20:04 +0100)]
Get rid of the builtin_get_exception call in the IR
Instead allocate a temp to hold the exception to rethrow and pass the address
of that to the TRY statement. Then it can be set in the run-time.
Change-Id: Ic15869e8e5ab5119b26f98fc45dbdb1a2ad9d21e
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
Simon Hausmann [Wed, 6 Mar 2013 19:04:31 +0000 (20:04 +0100)]
Cleanup: Get rid of calls to builtin_{push,pop}_catch_scope in the IR
These calls aren't needed anymore and can easily be done on the run-time level
instead right before and after calling the catch block.
Change-Id: I445361688782d12269764578e83ebb21ae184190
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
Lars Knoll [Wed, 6 Mar 2013 17:12:39 +0000 (18:12 +0100)]
Implement a few mere methods in the v8 API
Change-Id: Ia95866060e27c4311ce819c7aec04f328fa8eca1
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
Simon Hausmann [Wed, 6 Mar 2013 19:04:21 +0000 (20:04 +0100)]
Optimize exception throwing
Introduce a try statement in the IR that allows the back-end to jump to the try
and catch blocks directly instead of having a conditional jump for each entry.
For the case where we have no catch but only try { ... } finally { ... } we
create a synthetic catch block that sets hasException to true, to ensure that
when an exception is thrown during try we execute finally and then rethrow the
exception.
Change-Id: If4be5421f9731522beab80e843283b517d4aa41c
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
Lars Knoll [Wed, 6 Mar 2013 13:54:30 +0000 (14:54 +0100)]
Implement most of the New() methods in the v8 API
Change-Id: Icf88354656ae12ae38a24eaebfd1b1d22e755c52
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
Lars Knoll [Wed, 6 Mar 2013 13:01:07 +0000 (14:01 +0100)]
Implement external object comparison
Change-Id: Ieec34aa1b410766510919a99c5964c288b4addc1
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
Lars Knoll [Wed, 6 Mar 2013 11:31:46 +0000 (12:31 +0100)]
Implement the various v8 Cast methods
Change-Id: I056fbcb229ef3b487e28e14c6d58007aeb7ba8a0
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
Lars Knoll [Wed, 6 Mar 2013 10:19:02 +0000 (11:19 +0100)]
Fix a bug in Persistent::New
Change-Id: Iac13aa897e64bca6b50ea90bd5e2804e2212e008
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
Lars Knoll [Wed, 6 Mar 2013 10:18:56 +0000 (11:18 +0100)]
Cleanups
Change-Id: Ifbe085eea9f1f605af0b802f299f82f24a14e9b1
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
Lars Knoll [Wed, 6 Mar 2013 08:55:15 +0000 (09:55 +0100)]
Add stubs for all the methods in the v8 API
Qt Qml and Qt Quick now link against V4 :)
Change-Id: I8320875bdeab91d800fbf530b0d57a67dc872188
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
Lars Knoll [Tue, 5 Mar 2013 21:41:49 +0000 (22:41 +0100)]
Simplify conversion between VM::Value and V8::Value
Also implement the Exception class and ThrowException.
Change-Id: Ib9ca804cb82288d9da8b74f6cd3c38ecd61dbc1b
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
Lars Knoll [Tue, 5 Mar 2013 21:21:40 +0000 (22:21 +0100)]
Implement v8::Undefined/Null/True/False
Change-Id: I831e3b5e53d06edcbce57d37fc43995154b26305
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
Lars Knoll [Tue, 5 Mar 2013 19:28:06 +0000 (20:28 +0100)]
Remove API not used by Qml
Change-Id: Ibd202369f320c2db501d7a900941f0734326db7d
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
Lars Knoll [Tue, 5 Mar 2013 16:13:15 +0000 (17:13 +0100)]
Make a V8 hello world actually work.
Change-Id: Ia80004d2104e72e7c871be3582cab26ca9e493bf
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
Simon Hausmann [Tue, 5 Mar 2013 15:37:08 +0000 (16:37 +0100)]
Implement v8::Script::Run()
Change-Id: I7456e0a5e389216aac84d3ca9d66cc1e7f688008
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
Simon Hausmann [Tue, 5 Mar 2013 15:11:22 +0000 (16:11 +0100)]
Move function execution into a little ExecutionEngine::run helper function
This will allow for re-use from the v8 wrapper API.
Change-Id: I2330432dad43d7fd7af6147e96a0cbbb340cd917
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
Simon Hausmann [Fri, 15 Feb 2013 13:59:34 +0000 (14:59 +0100)]
Initial v8'ish API wrappers for QML integration
The goal is for these wrappers to exist only for a limited period
of time until all of QML is ported to use pure v4vm.
Change-Id: I4568420a2116420ae6f45bc5eb1e3d464a7c4cc3
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
Simon Hausmann [Tue, 5 Mar 2013 12:12:37 +0000 (13:12 +0100)]
Add support for associating external resources with garbage collected JS objects
Change-Id: Ib09096b684997ac77902e3cc50458092e2f569a2
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
Simon Hausmann [Fri, 15 Feb 2013 13:58:01 +0000 (14:58 +0100)]
Remove unused wtf inclusion from qv4managed.h
This allows including this header file outside of the module, i.e. in
a qml integration.
Change-Id: I0aed18d76dfacfdf5e615cc12f3d2904fcef9091
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
Simon Hausmann [Tue, 5 Mar 2013 11:49:35 +0000 (12:49 +0100)]
Rename qv4ir_p.h to qv4jsir_p.h to avoid conflict with qtdeclarative
Change-Id: I0e222141232f191e47950318f8262144de486703
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
Simon Hausmann [Mon, 4 Mar 2013 14:30:40 +0000 (15:30 +0100)]
Cleanup unwind handler
Store the platform specific unwind info (CIE/FDE on Linux/Mac x86/x86-64) directly
inside the VM::Function and therefore make all the functions of UnwindHelper static.
Also calculate the function size correctly.
Change-Id: If6a6b92a250044c6799218ef3506f34d630674c7
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
Simon Hausmann [Mon, 4 Mar 2013 11:14:54 +0000 (12:14 +0100)]
Cleanup: Rename __builtin_delete_exception_handler
This IR builtin function no more deletes an exception handler, instead it tells
the back-end to stop try catching exceptions. In MASM and Moth this is
implemented by returning from the nested function call initiated by the try.
Change-Id: Ia8336c998817a73aeed03f4a05d4b592cc9143ad
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
Simon Hausmann [Mon, 4 Mar 2013 11:12:21 +0000 (12:12 +0100)]
Cleanup: Get rid of manual unwind stack
Instead of doing the ExecutionContext unwinding at the time of throw, do it
at the time of catch, conveniently through an accept() method of the Exception
object. That allows us to get rid of the separate handler registration stack.
The only tricky part are that some execution contexts are allocated on the stack.
When exceptions are thrown through that, we have to catch, unwinding up until that
point (or rather the parent of the stack allocated context that's going to be
deleted) and then simply rethrow.
This patch also gets rid of the __builtin_delete_exception_handler. The next patch
will rename the remainder to what it really does now.
Change-Id: I00bb113b3a2fe24f7054c03fdfb8fed5cc1258b1
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
Erik Verbruggen [Mon, 4 Mar 2013 10:17:55 +0000 (11:17 +0100)]
Fix IR control flow after a throw statement.
A throw statement in a try block will now jump to the catch/finally, and
a throw statement outside will jump to the exit block. This follows what
happens during run-time, which fixes liveness issues with the
return-temp of calls to builtin_create_exception_handler.
Change-Id: I70f14f03f0431d3fc962c22984d4f486a1d81c7d
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
Erik Verbruggen [Mon, 4 Mar 2013 08:32:16 +0000 (09:32 +0100)]
Fix off-by-one in assertion for argc/args in calls.
When a call is done with 1 argument, it is not copied into the outgoing
args array, but passed by-reference. So when it happens to be the last
element in the stack frame, then args + argc == stackSize (because argc
is 1).
Change-Id: Idb769c95e9066c24a9d93cdcc24e13d3a9acc995
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
Lars Knoll [Mon, 4 Mar 2013 10:48:49 +0000 (11:48 +0100)]
Optimise array construction
Cache the InternalClass for ArrayObject and
use it directly in the constructor.
Change-Id: I17346045cd14852c400c7f3886414953e4dff869
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
Lars Knoll [Mon, 4 Mar 2013 10:47:59 +0000 (11:47 +0100)]
Small optimisation for string concatenation
The real fix will require a string class that can hold
several substrings.
Change-Id: I0e23f248048781b0f623d01087397afaeefcc97a
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
Erik Verbruggen [Fri, 1 Mar 2013 13:34:56 +0000 (14:34 +0100)]
Do not re-use temp for both parameter and result.
Change-Id: Iffd50459bc55960ac5cef1e246cd3d2664565a8a
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
Erik Verbruggen [Fri, 1 Mar 2013 13:28:46 +0000 (14:28 +0100)]
Moved isNegative(double) into another header.
So it can be used in codegen.
Change-Id: Iba1294a58cf6785eb9c37b34c39c4ec0281cd694
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
Erik Verbruggen [Fri, 1 Mar 2013 13:26:35 +0000 (14:26 +0100)]
Remove shared subexpressions from the function's IR code.
Change-Id: I4502eb9c86aba14142f1ae15f1ba560255a2cb6a
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
Erik Verbruggen [Mon, 4 Mar 2013 08:05:38 +0000 (09:05 +0100)]
Fix warning about unused parameter.
Change-Id: Ia94e1ac073dc16a0eb9841677343dbaddad923ae
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
Simon Hausmann [Fri, 1 Mar 2013 16:04:21 +0000 (17:04 +0100)]
Implement JavaScript exceptions using C++ exceptions
Instead of registering catch handlers with setjmp and throwing JS exceptions
with longjmp, they are now thrown and caught as C++ exceptions. This allows for
tight interoperability between C++ and JS in the future and allows for clear
semantics with regards to cleaning up memory in the engine when throwing
exceptions. (destructors are guaranteed to be called, unlike with
setjmp/longjmp).
The recent unwind table additions allow for the exceptions to be thrown through
JIT generated code.
Catching the exception is done by re-using the existing IR semantics where the
beginning of a try block is marked by registering an exception handler.
Execution after the registration continues conditionally, based on the return
value of builtin_create_exception_handler. A return value of is 0 the try
block(s) are executed. If an exception is thrown during that time, execution
resumes at the point where builtin_create_exception_handler returns, but with a
return value of 1. If an exception is thrown within the catch handler, the
execution resumes again at the same point, but the inCatch IR variable will
guide execution straight to the finally block(s), which calls
delete_exception_handler.
In the JIT as well as the interpreter this is implemented by entering a C++
code section that contains a C++ try {} catch {} block, in which the calling
function is called again and continues right at the next instruction (or the
interpreter loop is recursively entered). An exception will throw us out of
that scope and back into the try {} catch {} wrapper, which can call again
into the calling function.
The IR guarantees that delete_exception_handler is always called, regardless of
how the try or catch blocks are terminated. That is where in the JIT and
interpreter we return from the nested function call and return back into the
original stack frame, effectively unregistering the catch handler.
Further cleanups with regards to the naming and the exception handler stack
will come in subsequent patches, this is merely the minimal patch set to
change to the new mechanism.
This patch set breaks ARM until ARM exception handler tables are implemented.
The interpreter changes are based on a patchset from Erik
from https://codereview.qt-project.org/#change,45750
Change-Id: I543f2bd37b2186f7e48ffcab177d57b5ce932a0c
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
Lars Knoll [Sat, 2 Mar 2013 23:50:33 +0000 (00:50 +0100)]
Fix builtin_define_property
Commit
72c1fe5822aa65f4a3f70f78e058fb7e3154a4b6 broke
object literals that uses numbers as keys (e.g.
{ "2": "bla" }. This fixes it while keeping the faster
code path.
Change-Id: I0e89eb6e03da6a2e55d833ac0ad956f35e597297
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
Simon Hausmann [Sun, 3 Mar 2013 09:20:05 +0000 (10:20 +0100)]
Fix crashes on ia32 with array literals
prepareVariableArguments uses copyValue to copy/initialize the
arguments for a variable arguments call. Previously it called
copyValue with only temps as source, which on ia32 is implemented
as
(1) load temp from memory into FP reg
(2) store FP reg into destination memory location
After Gerrit change Id5757dd7a910be13db4958da7e449172ec1b04ae the
source of copyValue can now also be an IR::Expr and hence a constant
value. On ia32 it is unfortunately not possible to load a constant
floating point value into an FP reg right away. However in this
very situation it's sufficient to store the correct VM::Value in
the destination memory address right away. This is now done using
a template specialization of copyValue.
The old code compiled because we accidentally picked the
loadDouble(const void *address) overload of MASM, with address
pointing to the IR::Expr :)
Change-Id: Ie43f9be432cfcb844cc5e706e63bd3e91e857bac
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
Lars Knoll [Fri, 1 Mar 2013 15:47:07 +0000 (16:47 +0100)]
More efficient initialization of array literals
Initialize array literals in one go instead of going
through repeated calls to the runtime for each value.
This gives quite a nice speed improvement on the splay
tree benchmark.
Change-Id: Id5757dd7a910be13db4958da7e449172ec1b04ae
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
Lars Knoll [Sat, 2 Mar 2013 23:54:48 +0000 (00:54 +0100)]
Skip some really slow tests
The test cases themself finish in ~10-20 secs. Unfortunately
__deregister_frame seems to be extremely slow if you have many
functions (as is the case in these tests). It takes many minutes
to then clean up the engine and deregister all frames.
Change-Id: Idaa829f64c91d324e650b1d22b94dec6becad3d4
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>