With parallel recompilation enabled, objects made accessible by handles may
have changed between graph construction and graph optimization. Therefore
we must not assume that information on those objects remain the same between
those two phases. To police this, we forbid handle dereferencing during
graph optimization.
Exceptions to this rule are:
- Dereferencing the handle to obtain the raw location of the object. This
is safe since parallel recompilation acquires RelocationLock
- Some places that dereference the handle for a type check. These are checked
to be safe on a case-by-case basis.
R=jkummerow@chromium.org
BUG=
Review URL: https://chromiumcodereview.appspot.com/12049012
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@13475 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
HCheckPrototypeMaps currently records the prototype and the holder of the
prototype chain (both ends of the chain) and assumes that the chain elements
and their maps did not change in during the entirety of Crankshaft. The actual
traversal of the prototype chain happens in Lithium at code generation.
With parallel compilation, this assumption is not longer correct.
R=mstarzinger@chromium.org
BUG=
Review URL: https://chromiumcodereview.appspot.com/11864013
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@13454 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
This reduces the time take for mjsunit/limit-locals from 56.8s to 15.1s in debug
mode and from 12.0s to 1.6s in release mode.
Note that GrowableBitVector and BitVector should really be merged, and probably
have their allocation strategy parmeterized. The current state of affairs
involving tons of checks and delegation is extremely ugly, and it is far from
clear if all that special casing is a clear win. STL FTW! :-P
Review URL: https://codereview.chromium.org/11775016
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@13327 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
Modules now have their own local scope, represented by their own context.
Module instance objects have an accessor for every export that forwards
access to the respective slot from the module's context. (Exports that are
modules themselves, however, are simple data properties.)
All modules have a _hosting_ scope/context, which (currently) is the
(innermost) enclosing global scope. To deal with recursion, nested modules
are hosted by the same scope as global ones.
For every (global or nested) module literal, the hosting context has an
internal slot that points directly to the respective module context. This
enables quick access to (statically resolved) module members by 2-dimensional
access through the hosting context. For example,
module A {
let x;
module B { let y; }
}
module C { let z; }
allocates contexts as follows:
[header| .A | .B | .C | A | C ] (global)
| | |
| | +-- [header| z ] (module)
| |
| +------- [header| y ] (module)
|
+------------ [header| x | B ] (module)
Here, .A, .B, .C are the internal slots pointing to the hosted module
contexts, whereas A, B, C hold the actual instance objects (note that every
module context also points to the respective instance object through its
extension slot in the header).
To deal with arbitrary recursion and aliases between modules,
they are created and initialized in several stages. Each stage applies to
all modules in the hosting global scope, including nested ones.
1. Allocate: for each module _literal_, allocate the module contexts and
respective instance object and wire them up. This happens in the
PushModuleContext runtime function, as generated by AllocateModules
(invoked by VisitDeclarations in the hosting scope).
2. Bind: for each module _declaration_ (i.e. literals as well as aliases),
assign the respective instance object to respective local variables. This
happens in VisitModuleDeclaration, and uses the instance objects created
in the previous stage.
For each module _literal_, this phase also constructs a module descriptor
for the next stage. This happens in VisitModuleLiteral.
3. Populate: invoke the DeclareModules runtime function to populate each
_instance_ object with accessors for it exports. This is generated by
DeclareModules (invoked by VisitDeclarations in the hosting scope again),
and uses the descriptors generated in the previous stage.
4. Initialize: execute the module bodies (and other code) in sequence. This
happens by the separate statements generated for module bodies. To reenter
the module scopes properly, the parser inserted ModuleStatements.
R=mstarzinger@chromium.org,svenpanne@chromium.org
BUG=
Review URL: https://codereview.chromium.org/11093074
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@13033 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
Previously Crankshaft emitted a generic load for these, now we emit a load of a
named field, guarded by a proto chain check.
LCheckPrototypeMaps now returns the holder, which is for free, because it
already had to check its map as the last step, anyway. This is in sync with what
StubCompiler::CheckPrototype does.
Review URL: https://codereview.chromium.org/11338030
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@12847 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
We iteratively remove all dead Hydrogen instruction until we reach a fixed point. We consider an instruction dead if it is unused, has no observable side effects and is deletable. The last part of the condition is currently not very nice: We basically have to whitelist "safe" instructions, because we are missing more detailed dependencies and/or more detailed tracking of side effects.
We disable dead code elimination for now in our test runners, because we have tons of poorly written tests which wouldn't test anymore what they are supposed to test with this phase enabled. To get test coverage for dead code elimination itself, we should enable it on a few build bots. This is not really a perfect state, but the best we can do for now.
This patch includes a few const-correctness fixes, most of them were necessary for this CL.
Review URL: https://codereview.chromium.org/11088027
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@12697 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
This fixes materialization of arguments objects for strict mode functions during
deoptimization. We materialize arguments from the stack area where optimized
code pushes the arguments when entering the inlined environment. For adapted
invocations we use the arguments adaptor frame for materialization.
R=svenpanne@chromium.org
BUG=v8:2261
TEST=mjsunit/regress/regress-2261,mjsunit/compiler/inline-arguments
Review URL: https://chromiumcodereview.appspot.com/10908194
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@12489 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
xmm0 is not saved across runtime call on x64 because MacroAssembler::EnterExitFrameEpilogue preserves only allocatable XMM registers unlike on ia32 where it preserves all registers.
Cleanup handling of shifts: SHR can deoptimize only when its a shift by 0, all other shift never deoptimize.
Fix type inference for i-to-t change instruction. On X64 this ensures that write-barrier is generated correctly.
R=danno@chromium.org
Review URL: https://chromiumcodereview.appspot.com/10868032
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@12373 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
Safe operations are those that either do not observe unsignedness or have special support for uint32 values:
- all binary bitwise operations: they perform ToInt32 on inputs;
- >> and << shifts: they perform ToInt32 on left hand side and ToUint32 on right hand side;
- >>> shift: it performs ToUint32 on both inputs;
- stores to integer external arrays (not pixel, float or double ones): these stores are "bitwise";
- HChange: special support added for conversions of uint32 values to double and tagged values;
- HSimulate: special support added for deoptimization with uint32 values in registers and stack slots;
- HPhi: phis that have only safe uses and only uint32 operands are uint32 themselves.
BUG=v8:2097
TEST=test/mjsunit/compiler/uint32.js
Review URL: https://chromiumcodereview.appspot.com/10778029
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@12367 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
in anticipation of the upcoming lexical global scope.
Mostly automatised as:
for FILE in `egrep -ril "global[ _]?context" src test/cctest`
do
echo $FILE
sed "s/Global context/Native context/g" <$FILE >$FILE.0
sed "s/global context/native context/g" <$FILE.0 >$FILE.1
sed "s/global_context/native_context/g" <$FILE.1 >$FILE.2
sed "s/GLOBAL_CONTEXT/NATIVE_CONTEXT/g" <$FILE.2 >$FILE.3
sed "s/GlobalContext/NativeContext/g" <$FILE.3 >$FILE
rm $FILE.[0-9]
done
R=mstarzinger@chromium.org
BUG=
TEST=
Review URL: https://chromiumcodereview.appspot.com/10832342
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@12325 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
Currently only simple setter calls are handled (i.e. no calls in count
operations or compound assignments), and deoptimization in the setter is not
handled at all. Because of the latter, we temporarily hide this feature behind
the --inline-accessors flag, just like inlining getters.
We now use an enum everywhere we depend on the handling of a return value,
passing around several boolean would be more confusing.
Made VisitReturnStatement and the final parts of TryInline more similar, so
matching them visually is a bit easier now.
Simplified the signature of AddLeaveInlined, the target of the HGoto can simply
be retrieved from the function state.
Review URL: https://chromiumcodereview.appspot.com/10836133
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@12286 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
When inlining is being done, it is crucial to use the correct type feedback
oracle with a given type feedback ID. To ensure this, TestContext now carries an
oracle which is associated with the context's condition, and these are both used
together in TestContext::BuildBranch.
Note that in VisitReturnStatement and TryInline we are currently lucky that the
oracles don't go out of sync in an observable way, but this will change when we
inline setters. Therefore, there is no separate test case...
Review URL: https://chromiumcodereview.appspot.com/10834247
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@12284 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
The underlying problem is that for compound/count operations we use the *load*
type feedback for storing, too. For normal properties this doesn't matter, but
for accessor properties we should better use the *store* type feedback, which
would be available, too. This consistent feedback usage could be guaranteed if
we removed the heavy copy-n-paste in the crankshaft code generation for
compound/count operations and assignments/property loads.
To be on the safe side, we postpone this refactoring and do a quick and easily
mergeable fix.
BUG=140083
TEST=mjsunit/regress/regress-crbug-140083.js
Review URL: https://chromiumcodereview.appspot.com/10828146
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@12252 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
Currently only simple getter calls are handled (i.e. no calls in count
operations or compound assignments), and deoptimization in the getter is not
handled at all. Because of the latter, we temporarily hide this feature behind a
new flag --inline-accessors, which is false by default.
Review URL: https://chromiumcodereview.appspot.com/10828066
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@12223 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
This is a refactoring-only CL and the third one in a series for enabling
inlining of accessors. The goal of this CL is to move the builders for accessors
to the places where we might be able to inline them later, i.e. the VisitFoo and
HandleBar member functions of HGraphBuilder.
Extracted duplicate code into LookupAccessorPair.
Review URL: https://chromiumcodereview.appspot.com/10831047
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@12209 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
This is a refactoring-only CL and the first one in a series for enabling
inlining of accessors. The naming and argument order has been unified a bit, and
some tests have been pushed to the caller in order to get a simpler
signature. Note that the latter temporarily introduces some code redundancy, but
this will be cleaned up in one of the next CLs.
Review URL: https://chromiumcodereview.appspot.com/10826028
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@12198 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
The previous fix was for "real" calls, this one is for getters. It is a bit
unfortunate that this has to be fixed twice: We should really break up
Call::ComputeTarget into a predicate and 1 or 2 getters, so code can be reused.
The regression test has been modified a bit to make things more uniform.
Review URL: https://chromiumcodereview.appspot.com/10702164
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@12053 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
