The interpreter used a ZoneVector<WasmVal> to model the value stack.
Thus, at each single pop to the stack, a bounds check was performed,
and the storage was potentially extended.
This CL changes this to pre-allocate enough space for the stack of a
function when a new frame is entered. This avoids any checks for pushs
and pops.
Instead of storing a ZoneVector<WasmVal>, we store WasmVal* directly.
The maximum value stack size is precomputed together with the control
transfer side table.
This CL speeds up interpreted execution by 15% on average (measured
locally on a Z840).
R=ahaas@chromium.org
BUG=v8:5822
Change-Id: If949f7ee5233d874cd6a04b7dde2d7b4a95e45ea
Reviewed-on: https://chromium-review.googlesource.com/488061
Commit-Queue: Clemens Hammacher <clemensh@chromium.org>
Reviewed-by: Andreas Haas <ahaas@chromium.org>
Cr-Commit-Position: refs/heads/master@{#44902}
This still doesn't cover all the paths yet, since some paths are
impossible to trigger at this point due to the way the CanInlineCall
predicate works on the AllocationSite, which says multiple things:
- In case of Array(len), the len was always a Smi so far.
- In case of Array(...args), storing the args didn't change the
elements kind.
- In case of Array(len), the len was always less than the initial
maximum fast element array size.
These conditions are tailored towards Crankshaft and don't really
make a lot of sense in the TurboFan world. We'd need more fine
grained protections, which we will achieve by refactoring the Array
constructor.
BUG=chromium:715404,v8:6262
TBR=machenbach@chromium.org
Review-Url: https://codereview.chromium.org/2843033002
Cr-Commit-Position: refs/heads/master@{#44901}
For holey Smi and double source arrays, we would go to the general
case, which is much slower than before. We already check that there
are no prototype chain changes in IterableToListCanBeElided, and
there is no JS-code run between that check and the copying of the
elements, so we can safely check for the hole and convert it to
undefined, which is then converted to 0/NaN appropriately for the
given TypedArray.
Bug: chromium:713570,chromium:711275
Change-Id: I5b21c915907d71eebb73b7b1eea8eb58b4a5436d
Reviewed-on: https://chromium-review.googlesource.com/485520
Commit-Queue: Peter Marshall <petermarshall@chromium.org>
Reviewed-by: Camillo Bruni <cbruni@chromium.org>
Reviewed-by: Benedikt Meurer <bmeurer@chromium.org>
Cr-Commit-Position: refs/heads/master@{#44899}
At the moment the AsyncCompileJob object is deallocated after one of its
task functions return false. This mechanism is, however, not documented,
potentially error-prone, and I think there are already some cases where
I think that we got it wrong.
This CL moves the deallocation of the AsyncCompileJob object to the
place where the promise which belongs to the AsyncCompileJob is either
resolved or rejected. This is a more appropriate place to deallocate the
object, because conceptionally, at the end of every an AsyncCompileJob
its promise should either be resolved or rejected.
R=clemensh@chromium.org, mtrofin@chromium.org
Change-Id: I87618c5619a3ac923645d1c3f6acaee9b0b14a83
Reviewed-on: https://chromium-review.googlesource.com/486884
Reviewed-by: Clemens Hammacher <clemensh@chromium.org>
Reviewed-by: Mircea Trofin <mtrofin@chromium.org>
Commit-Queue: Andreas Haas <ahaas@chromium.org>
Cr-Commit-Position: refs/heads/master@{#44896}
Reason for revert:
Field representation is not preserved
Original issue's description:
> [turbofan] Set proper representation for initial arguments length.
>
> The JSArgumentsObject::length representation is initially Smi, so we can
> record that on the initial map and use it to optimize the accesses in
> TurboFan based on that. Similar for JSSloppyArgumentsObject::caller.
>
> BUG=v8:6262
> R=yangguo@chromium.org
>
> Review-Url: https://codereview.chromium.org/2810333004
> Cr-Commit-Position: refs/heads/master@{#44644}
> Committed: 5eec7df9b3TBR=yangguo@chromium.org,bmeurer@chromium.org
# Not skipping CQ checks because original CL landed more than 1 days ago.
BUG=v8:6262
Review-Url: https://codereview.chromium.org/2825323002
Cr-Commit-Position: refs/heads/master@{#44893}
This makes sure that the observable property order of the module export
maintains insertion order. Now that properties are configurable, we no
longer need to reverse the export processing.
R=clemensh@chromium.org
TEST=mjsunit/asm/asm-validation
BUG=chromium:715420
Change-Id: Ib2024254c07bdad7fee1cf2fa0bd3e847721f5b5
Reviewed-on: https://chromium-review.googlesource.com/488022
Reviewed-by: Clemens Hammacher <clemensh@chromium.org>
Commit-Queue: Michael Starzinger <mstarzinger@chromium.org>
Cr-Commit-Position: refs/heads/master@{#44891}
This fixes the bounds checking of "unsigned" numeric literals (those
that do not contains dots) by the parser. In particular this fixes a
bogus truncation to 32-bit in the scanner. It also makes the scanner
more robust by limiting the range of those numeric literals, hence
completely avoiding rounding loss or truncation errors.
R=clemensh@chromium.org
TEST=unittests/AsmJsScannerTest.UnsignedNumbers
BUG=v8:6298
Change-Id: Id31ab3c652e99fa8d3d6663315768e1bfaf3b773
Reviewed-on: https://chromium-review.googlesource.com/486881
Reviewed-by: Clemens Hammacher <clemensh@chromium.org>
Commit-Queue: Michael Starzinger <mstarzinger@chromium.org>
Cr-Commit-Position: refs/heads/master@{#44890}
Instead of calculating the OSR entry point both in the bytecode analysis
and in the bytecode graph builder, calculate it once in the analysis and
use that calculation in the graph builder.
Old TODO from https://codereview.chromium.org/2558093005.
Change-Id: I071bc622beb55dc5eddaee25ef28e21fc4b477f0
Reviewed-on: https://chromium-review.googlesource.com/485899
Commit-Queue: Michael Starzinger <mstarzinger@chromium.org>
Reviewed-by: Michael Starzinger <mstarzinger@chromium.org>
Cr-Commit-Position: refs/heads/master@{#44888}
This makes e.g., load(file) work within Realm.eval(realm, "load(file)") to load files into that realm.
Bug:
Change-Id: I85738f0dfab621f2a8c9e2703f4ce4b39dd882bf
Reviewed-on: https://chromium-review.googlesource.com/484379
Reviewed-by: Yang Guo <yangguo@chromium.org>
Commit-Queue: Toon Verwaest <verwaest@chromium.org>
Cr-Commit-Position: refs/heads/master@{#44887}
Only create a singleton array for Array(len) if Type(len) cannot be
Number, otherwise we might need to throw an exception instead.
BUG=chromium:715404
R=jarin@chromium.org
Review-Url: https://codereview.chromium.org/2838123004
Cr-Commit-Position: refs/heads/master@{#44886}
The only users of the LoadStoreOpcodeOf function were a number of
macros in wasm-macro-gen.h, and three test functions using it directly.
This CL refactors those functions to also use the macros.
In one case, this requires storing the value in a local variable first.
R=ahaas@chromium.org
Change-Id: Ia2fbf67a3831fafc9345e155eb240cf1bf6feb5d
Reviewed-on: https://chromium-review.googlesource.com/486842
Reviewed-by: Andreas Haas <ahaas@chromium.org>
Commit-Queue: Clemens Hammacher <clemensh@chromium.org>
Cr-Commit-Position: refs/heads/master@{#44885}
Update the json file in js-perf-test with compare bytecode handler tests.
This cl (https://chromium-review.googlesource.com/c/485522/) adds new
tests but not all of them are updated in the json file.
Bug:v8:4280
Change-Id: Ifd1f479b770a4277fbba1de51ca2f7cbc26003cb
Reviewed-on: https://chromium-review.googlesource.com/487961
Reviewed-by: Ross McIlroy <rmcilroy@chromium.org>
Commit-Queue: Mythri Alle <mythria@chromium.org>
Cr-Commit-Position: refs/heads/master@{#44879}
Previously API function calls would only be optimized in TurboFan when
the receiver was a (compile-time) known constant, which was probably
only true for certain cases where functions where called on the global
proxy (the window object).
BUG=v8:5267,v8:6304
R=jarin@chromium.org
Review-Url: https://codereview.chromium.org/2839953002
Cr-Commit-Position: refs/heads/master@{#44877}
Although we currently only support up to 1GB memory, we want to raise
this issue in the future. This test illustrates several issues we need
to be sure to fix first.
Bug: v8:6306
Change-Id: I362b7a9e51e8eb33a50e3b172a6f01d41995c3cb
Reviewed-on: https://chromium-review.googlesource.com/487047
Commit-Queue: Brad Nelson <bradnelson@chromium.org>
Reviewed-by: Brad Nelson <bradnelson@chromium.org>
Cr-Commit-Position: refs/heads/master@{#44876}
Port 46d0e4818a
Original Commit Message:
The CallApiCallbackStub can avoid loading undefined in case the
call_data is already undefined, which doubles the number of versions of
the stub and adds unnecessary complexity (at the benefit of saving one
stupid load). The idea is to turn the CallApiCallbackStub into a single
builtin instead, which does the right thing, so this is the first step
towards that goal.
R=bmeurer@chromium.org, joransiu@ca.ibm.com, jyan@ca.ibm.com, michael_dawson@ca.ibm.com
BUG=v8:6304
LOG=N
Review-Url: https://codereview.chromium.org/2837283004
Cr-Commit-Position: refs/heads/master@{#44872}
The CallApiCallbackStub can avoid loading undefined in case the
call_data is already undefined, which doubles the number of versions of
the stub and adds unnecessary complexity (at the benefit of saving one
stupid load). The idea is to turn the CallApiCallbackStub into a single
builtin instead, which does the right thing, so this is the first step
towards that goal.
R=yangguo@chromium.org
BUG=v8:6304
Review-Url: https://codereview.chromium.org/2838143003
Cr-Commit-Position: refs/heads/master@{#44869}
In preparation for adding another verifier that only considers a subset
of the graph.
BUG=chromium:651354
Review-Url: https://codereview.chromium.org/2844473002
Cr-Commit-Position: refs/heads/master@{#44867}
Also add more local variables to regress-v8-6077 to force
register spill on platform with 32 float registers.
BUG=
Review-Url: https://codereview.chromium.org/2822073003
Cr-Commit-Position: refs/heads/master@{#44865}
Evacuators shoud know their associated collector and thus figure out the
marking state themselves.
BUG=chromium:651354
Review-Url: https://codereview.chromium.org/2840863002
Cr-Commit-Position: refs/heads/master@{#44864}
This makes an ObjectVisitor as powerful as a StaticVisitor and allows
slots recording in ObjectVisitor.
This patch also renames VisitCell method of ObjectVisitor to
VisitCellPointer, so that VisitCell is free to be used for actually
visiting a cell.
BUG=chromium:709075
Review-Url: https://codereview.chromium.org/2810653002
Cr-Commit-Position: refs/heads/master@{#44860}
This reverts commit 56a6fda316.
Reason for revert: Makes tsan flaky:
https://build.chromium.org/p/client.v8/builders/V8%20Linux64%20TSAN/builds/15038
Original change's description:
> [parser] Inital parallel parse tasks implementation.
>
> While parsing top-level code eager functions are skipped just like lazy
> ones, but also a parse task is created for each.
>
> The parse tasks are run by the compiler dispatcher and can be executed
> either on background thread or in idle time.
> After parsing of top-level code finishes it waits for all unfinished
> parser tasks - possibly picking up and executing them on current thread.
> Afterwards parse task results are stitched together with top-level AST,
> in case of failures eager functions are treated just like lazy -
> parsing/compilation is retriggered for them in the runtime and proper
> errors are generated (performance is not optimized for error case at
> all).
>
> BUG=v8:6093
>
> Change-Id: I718dd2acc8a70ae1b09c2dea2616716605d7b05d
> Reviewed-on: https://chromium-review.googlesource.com/483439
> Commit-Queue: Wiktor Garbacz <wiktorg@google.com>
> Reviewed-by: Marja Hölttä <marja@chromium.org>
> Reviewed-by: Jochen Eisinger <jochen@chromium.org>
> Reviewed-by: Daniel Vogelheim <vogelheim@chromium.org>
> Cr-Commit-Position: refs/heads/master@{#44849}
TBR=marja@chromium.org,vogelheim@chromium.org,jochen@chromium.org,wiktorg@google.com
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
BUG=v8:6093
Change-Id: I17e689efee7d216d28a94a5c8147022ae7e830dd
Reviewed-on: https://chromium-review.googlesource.com/486883
Reviewed-by: Michael Achenbach <machenbach@chromium.org>
Commit-Queue: Michael Achenbach <machenbach@chromium.org>
Cr-Commit-Position: refs/heads/master@{#44859}
With this CL SloppyArguments immediately go to dictionary elements on
deletion, keeping the arguments backing store packed.
Bug: v8:6251
Change-Id: I2afa4fb5f0af9942eee0a1606942f5f289539330
Reviewed-on: https://chromium-review.googlesource.com/480379
Commit-Queue: Camillo Bruni <cbruni@chromium.org>
Reviewed-by: Jakob Kummerow <jkummerow@chromium.org>
Cr-Commit-Position: refs/heads/master@{#44857}
At the moment all tasks which are spawned during asynchronous
compilation are CancelableTasks. However, we don't ever really cancel
tasks, and in the cases where we do it actually makes no sense.
Additionally, using CancelableTasks causes problems when V8 shuts down.
Therefore this CL switches to normal v8::Tasks instead of
CancelableTasks.
R=clemensh@chromium.org, mtrofin@chromium.org
BUG=v8:6253
Change-Id: Idf972fa042e2614a3b25faa4537416a772990bd3
Reviewed-on: https://chromium-review.googlesource.com/485760
Reviewed-by: Mircea Trofin <mtrofin@chromium.org>
Reviewed-by: Clemens Hammacher <clemensh@chromium.org>
Commit-Queue: Andreas Haas <ahaas@chromium.org>
Cr-Commit-Position: refs/heads/master@{#44855}
