This ensures that we always have a kind of "brace" around the actual compilation
with --trace-opt. Previously this was coupled to --trace-hydrogen for
Crankshaft, and there was no output at the start of the compilation for
TurboFan.
Removed redundant "[completed...]" output: Whenever the compilation was
successful, we have already printed "[optimizing...took...]".
Output total TurboFanning time with --trace-opt, too (basically as graph
building time, this is rather arbitrary).
R=bmeurer@chromium.org
Review URL: https://codereview.chromium.org/819613002
Cr-Commit-Position: refs/heads/master@{#25900}
Denver supports a coherent cache mechanism. There is no need to clean
the D cache and invalidate I cache. MTS has to check the translation
anytime there is an I cache invalidate and this time can be saved by
making FlushICache a NOP.
The patch improves Octane by roughly 3-4% on Denver.
Review URL: https://codereview.chromium.org/797233002
Cr-Commit-Position: refs/heads/master@{#25898}
I picked a more conservative start value this time of half max old generation size.
BUG=
Review URL: https://codereview.chromium.org/804323004
Cr-Commit-Position: refs/heads/master@{#25890}
This keeps dying maps alive for FLAG_retain_maps_for_n_gc garbage collections to increase chances of them being reused for new objects in future.
BUG=v8:3664
LOG=N
TEST=cctest/test-heap/MapRetaining
Review URL: https://codereview.chromium.org/794583003
Cr-Commit-Position: refs/heads/master@{#25887}
Reason for revert:
[sheriff] Still crashes on win32 (XP):
http://build.chromium.org/p/client.v8/builders/V8%20Win32%20-%201/builds/1380
Original issue's description:
> ES6 computed property names
>
> This adds support for computed property names, under the flag
> --harmony-computed-property-names, for both object literals and
> classes.
>
> This is a revert of the revert, a76419f0f4.
>
> This changes to do an early bailout in
> HOptimizedGraphBuilder::VisitObjectLiteral instead of doing that in the later
> loop.
>
> BUG=v8:3754
> LOG=Y
> TBR=dslomov@chromium.orgTBR=dslomov@chromium.org,arv@chromium.org
NOTREECHECKS=true
NOTRY=true
BUG=v8:3754
Review URL: https://codereview.chromium.org/811593004
Cr-Commit-Position: refs/heads/master@{#25872}
This adds support for computed property names, under the flag
--harmony-computed-property-names, for both object literals and
classes.
This is a revert of the revert, a76419f0f4.
This changes to do an early bailout in
HOptimizedGraphBuilder::VisitObjectLiteral instead of doing that in the later
loop.
BUG=v8:3754
LOG=Y
TBR=dslomov@chromium.org
Review URL: https://codereview.chromium.org/792233008
Cr-Commit-Position: refs/heads/master@{#25868}
The GraphReducer should recurse into new inputs after a node was
changed. Most of the reducers already assume this behavior, and it
mostly worked by accident in many cases.
R=svenpanne@chromium.org
Review URL: https://codereview.chromium.org/800073004
Cr-Commit-Position: refs/heads/master@{#25861}
The spec ended up using Get(unscopables, propertyName) and
comparing the result to undefined instead of using Has.
BUG=v8:3632
LOG=Y
R=adamk, dslomov@chromium.org
Review URL: https://codereview.chromium.org/807893002
Cr-Commit-Position: refs/heads/master@{#25854}
Reason for revert:
Crashes Win32. It was not flake.
Original issue's description:
> ES6 computed property names
>
> This adds support for computed property names, under the flag
> --harmony-computed-property-names, for both object literals and
> classes.
>
> This is a revert of the revert, a76419f0f4 with
> no changes. I cannot reproduce the issue on Win8.1 or WinXP. Letting the bots
> try again.
>
> BUG=v8:3754
> LOG=Y
> TBR=dslomov@chromium.orgTBR=dslomov@chromium.org
NOTREECHECKS=true
NOTRY=true
BUG=v8:3754
Review URL: https://codereview.chromium.org/807033003
Cr-Commit-Position: refs/heads/master@{#25853}
This adds support for computed property names, under the flag
--harmony-computed-property-names, for both object literals and
classes.
This is a revert of the revert, a76419f0f4 with
no changes. I cannot reproduce the issue on Win8.1 or WinXP. Letting the bots
try again.
BUG=v8:3754
LOG=Y
TBR=dslomov@chromium.org
Review URL: https://codereview.chromium.org/807173002
Cr-Commit-Position: refs/heads/master@{#25851}
This greatly reduces the number of nodes in the graph (by more than 20x in
some extreme cases) for the Emscripten python interpreter main function.
BUG=v8:3763
LOG=y
TEST=cctest,mjsunit,unittests
R=svenpanne@chromium.org
Review URL: https://codereview.chromium.org/802353003
Cr-Commit-Position: refs/heads/master@{#25840}
The bug would occur when we try to Reset() to a position already at the end.
This happens e.g., when the regexp ends with \u. What used to happen in that
case: 1) Advance past \ and u (to the end) (which wouldn't increase next_pos_
enough) 2) Try to parse 4 hex digits 3) When that failed, Reset() to the
position which should've been at the end but wasn't.
To be able to properly Reset() to a position at the end, we need to allow
next_pos_ to move beyond the end (since position() is next_pos_ - 1).
Minimal repro case:
var r = /foo\u/
r.test("foou") // should be true, was false.
(Note that \u not followed by 4 hex didits should be interpreted as an identity
escape. It already worked unless \u was at the end of the regexp.)
BUG=v8:3756
LOG=NO
Review URL: https://codereview.chromium.org/802313003
Cr-Commit-Position: refs/heads/master@{#25838}
port d211608a3e
original commit message:
[turbofan] Remove the no-context hack for JSToNumber.
The ToNumberStub is now able to handle all plain primitives (Numbers,
Booleans, Null, Undefined and Strings) without context access.
BUG=
Review URL: https://codereview.chromium.org/810683002
Cr-Commit-Position: refs/heads/master@{#25833}
Relanded with fix for always returning Change for PlainPrimitive even
if there was no change. The performance regression on primes.js and
corrections.js is due to unlucky loop header alignment; will be addressed
separately.
TEST=unittests
R=svenpanne@chromium.org
Committed: 75484e8d16
Review URL: https://codereview.chromium.org/799413002
Cr-Commit-Position: refs/heads/master@{#25832}
There are no users of this infrastructure yet, so it's behind an off-by-default flag.
Review URL: https://codereview.chromium.org/768633002
Cr-Commit-Position: refs/heads/master@{#25829}
Port d211608a3e
Original commit message:
The ToNumberStub is now able to handle all plain primitives (Numbers,
Booleans, Null, Undefined and Strings) without context access.
TEST=cctest,mjsunit,unittests
BUG=
Review URL: https://codereview.chromium.org/803973002
Cr-Commit-Position: refs/heads/master@{#25824}
This adds support for computed property names, under the flag
--harmony-computed-property-names, for both object literals and
classes.
BUG=v8:3754
LOG=Y
Review URL: https://codereview.chromium.org/795573005
Cr-Commit-Position: refs/heads/master@{#25821}
build is the third number of the V8 version, and very likely to change
(in contrast to the patch level which typically is zero on canaries).
BUG=chromium:440984
R=mvstanton@chromium.org,yangguo@chromium.org
LOG=n
Review URL: https://codereview.chromium.org/802363002
Cr-Commit-Position: refs/heads/master@{#25820}
Keyed loads should not unconditionally be compiled to element loads. Update KeyedLoadICs to keep track of the key type, so that Hydrogen can emit ICs for string-keyed loads it doesn't have inline support for.
BUG=v8:3167
R=jkummerow@chromium.org
Review URL: https://codereview.chromium.org/755513003
Cr-Commit-Position: refs/heads/master@{#25817}
This functionality is now used by both object visitor and store buffer.
TEST=cctest/test-unboxed-doubles
Review URL: https://codereview.chromium.org/726713003
Cr-Commit-Position: refs/heads/master@{#25816}
Reason for revert:
Fix Windows nosnap.
Original issue's description:
> Call DisableInlineAllocation() in heap setup when flag inline_new is off.
>
> BUG=
TBR=ulan@chromium.org,mstarzinger@chromium.org
NOTREECHECKS=true
NOTRY=true
BUG=
Review URL: https://codereview.chromium.org/794053003
Cr-Commit-Position: refs/heads/master@{#25815}
The ToNumberStub is now able to handle all plain primitives (Numbers,
Booleans, Null, Undefined and Strings) without context access.
TEST=cctest,mjsunit,unittests
Review URL: https://codereview.chromium.org/801333002
Cr-Commit-Position: refs/heads/master@{#25814}
Add support for Symbol.isConcatSpreadable in Array.prototype.concat. This enables spreading non-Array objects with the symbol.
LOG=N
R=dslomov@chromium.org
BUG=
Review URL: https://codereview.chromium.org/771483002
Cr-Commit-Position: refs/heads/master@{#25808}
Such properties never notified prior to r21558, but the combination of
that change and r23163 led to sending notifications when they were
set via Object.defineProperty (but not when set via other means).
This also allows some cleanup in v8natives.js and objects.cc,
both of which were doing unnecessary contortions to produce the right
change records.
BUG=v8:3745
LOG=n
Review URL: https://codereview.chromium.org/791243002
Cr-Commit-Position: refs/heads/master@{#25806}
The Mac version of GetCurrentThreadId should be used when building the host
build of V8 on Android for Mac.
Review URL: https://codereview.chromium.org/799943003
Cr-Commit-Position: refs/heads/master@{#25805}
We managed to access the scope_ member of CompilationInfo before its
containing object was actually constructed.
Rule of thumb: When constructing an object, never ever pass around
pointers to members which come later in the member initializer list,
you simply can't see locally if this might cause trouble or not.
Review URL: https://codereview.chromium.org/796363002
Cr-Commit-Position: refs/heads/master@{#25796}
Avoid explicitly branching to the false label on unordered when the condition
on the true branch will not catch the unordered case and let the code fall
through.
R=bmeurer@chromium.org
Review URL: https://codereview.chromium.org/794203003
Cr-Commit-Position: refs/heads/master@{#25787}
It could happen that we shrink a live object on the main thread (e.g.
MigrateFastToSlow) while we're sweeping the same page. The main
thread first creates a filler object that the release-stores the new
map. Therefore it's important to barrier load the map word of live
objects from the sweeper thread.
BUG=none
R=ulan@chromium.org,hpayer@chromium.org
LOG=n
Review URL: https://codereview.chromium.org/797623002
Cr-Commit-Position: refs/heads/master@{#25780}
To make space in the type bitset, remove Function, RegExp, and Buffer
types for now, since they aren't really relied upon anyway.
R=bmeurer@chromium.org
BUG=
Review URL: https://codereview.chromium.org/795993002
Cr-Commit-Position: refs/heads/master@{#25776}
The instruction selector now selects pseudo instructions: CompareAndBranch or
TestAndBranch which are associated with their continuations so that generic
code in the code generator will treat them as branch instruction and will be
able to apply optimization like avoiding branches when the code can falltrhough.
R=bmeurer@chromium.org
Review URL: https://codereview.chromium.org/798553002
Cr-Commit-Position: refs/heads/master@{#25773}
V8 didn't report compile events on pause before this patch. These events can be important for listener. For example, DevTools allows user to execute some JS code on pause and needs to show correct stack trace in message from it.
BUG=396013
R=yurys@chromium.org
Review URL: https://codereview.chromium.org/781623004
Cr-Commit-Position: refs/heads/master@{#25767}
This completes the first round of optimizations for Map and Set.
All non-key-dependent methods now have a Hydrogen version, and
for keyed methods, string versions are optimized.
Review URL: https://codereview.chromium.org/796503002
Cr-Commit-Position: refs/heads/master@{#25763}
They both now run fast (due to utilizing transitions instead of always
creating new maps) and sealed or non-extensible objects can stay in
fast mode after transitioning.
This almost entirely reuses the code for transitioning objects
frozen by Object.freeze(), with the added benefit of freeing
up a bit on the map (we no longer keep track of frozen-ness,
as that bit wasn't used for anything interesting).
BUG=v8:3662,chromium:115960
LOG=y
Review URL: https://codereview.chromium.org/776143005
Cr-Commit-Position: refs/heads/master@{#25759}
Reason for revert:
For breaking the waterfall (run-json-stringify test).
Original issue's description:
> Avoid number range holes in bitset types.
>
> BUG=
TBR=rossberg@chromium.org
NOTREECHECKS=true
NOTRY=true
BUG=
Review URL: https://codereview.chromium.org/794663002
Cr-Commit-Position: refs/heads/master@{#25756}
This makes tests runnable with the external snapshot, and should be the
last step before enabling external snapshot on >=1 bots.
R=yangguo
BUG=
Review URL: https://codereview.chromium.org/780333004
Cr-Commit-Position: refs/heads/master@{#25753}
Also fix the pushing of JSToBoolean into Phis and generalize it to
also include pushing into Selects.
TEST=cctest,unittests
Review URL: https://codereview.chromium.org/792463003
Cr-Commit-Position: refs/heads/master@{#25718}
This was previously landed in commit 8599f5f, but failed the
collections mjsunit test on ia32 with --deopt-every-n=1. The fixed
patch adds a ClearFlag(HValue::kCanOverflow) call after every
arithmetic operation, which should remove all the deopt points in these
intrinsics.
Ideally, it seems like there should be a way to verify that there are
no deopt points for these inline optimized functions, since there's
nothing to deopt to. But I don't currently know of such a thing.
Review URL: https://codereview.chromium.org/782073002
Cr-Commit-Position: refs/heads/master@{#25715}
Blink already has code for handling invalid cached data. The attached test
ensures that cached data is gracefully rejected if it cannot be used.
This also unifies parser cache and code cache handling.
R=yangguo@chromium.org
BUG=439889
LOG=N
Review URL: https://codereview.chromium.org/781203003
Cr-Commit-Position: refs/heads/master@{#25708}
When compiling with the macro DCHECK_ALWAYS_ON defined, DCHECKs and
supporting code gets compiled and enabled.
This increases test coverage for chromium release buildbots
BUG=v8:3731
R=jkummerow@chromium.org
LOG=y
Review URL: https://codereview.chromium.org/760213005
Cr-Commit-Position: refs/heads/master@{#25701}
Reason for revert:
Deopt fuzzer reports that something's still broken, reverting until
I can create a minimal repro.
Original issue's description:
> Optimize add/set/delete operations for string keys in Maps and Sets
>
> This was previously landed in commit 66e2f60, but failed the
> collections mjsunit test with --deopt-every-n=1 due to a typo
> in the shrinking code. This patch corrects and simplifies the
> shrinking logic, and the tests now pass.
>
> R=dslomov@chromium.org
>
> Committed: 8599f5f047TBR=dslomov@chromium.org
NOTREECHECKS=true
NOTRY=true
Review URL: https://codereview.chromium.org/742353006
Cr-Commit-Position: refs/heads/master@{#25696}
This was previously landed in commit 66e2f60, but failed the
collections mjsunit test with --deopt-every-n=1 due to a typo
in the shrinking code. This patch corrects and simplifies the
shrinking logic, and the tests now pass.
R=dslomov@chromium.org
Review URL: https://codereview.chromium.org/777663003
Cr-Commit-Position: refs/heads/master@{#25695}
This introduces Hydrogen for %_GetPrototype. The code falls back on
runtime if the object needs access checks or if its prototype is a
hidden prototype.
BUG=None
LOG=Y
R=dslomov@chromium.org
Review URL: https://codereview.chromium.org/756423006
Cr-Commit-Position: refs/heads/master@{#25694}
Port 35ff259b63
Original commit message:
This combines Map::DoneInobjectSlackTracking and Map::ConstructionCount into one more generic 4-bit counter.
Counter values from 15 down to 8 are used for in-object slack tracking, values from 7 down to 0 are free to be used for a new counter when in-object slack tracking is inactive
BUG=
Review URL: https://codereview.chromium.org/780303003
Cr-Commit-Position: refs/heads/master@{#25692}
This combines Map::DoneInobjectSlackTracking and Map::ConstructionCount into one more generic 4-bit counter.
Counter values from 15 down to 8 are used for in-object slack tracking, values from 7 down to 0 are free to be used for a new counter when in-object slack tracking is inactive.
Review URL: https://codereview.chromium.org/767253002
Cr-Commit-Position: refs/heads/master@{#25689}
Achieve more than parity with modes currently handled on ia32 in preparation for
porting the entire mechanism to ia32, including supporting mul of constants 3,
5 and 9 with "leal" instructions.
TEST=unittests
Review URL: https://codereview.chromium.org/774193003
Cr-Commit-Position: refs/heads/master@{#25677}
This is an initial version of redundant load elimination, currently
limited to LoadField operators, and implemented by walking the effect
chain.
TEST=unittests
R=svenpanne@chromium.org
Review URL: https://codereview.chromium.org/782473002
Cr-Commit-Position: refs/heads/master@{#25673}
Reason for revert:
Caused test failures in mjsunit/es6/collections with --deopt-every-n-times=1
Original issue's description:
> Optimize add/set/delete operations for string keys in Maps and Sets
TBR=dslomov@chromium.org,arv@chromium.org
NOTREECHECKS=true
NOTRY=true
Review URL: https://codereview.chromium.org/777233003
Cr-Commit-Position: refs/heads/master@{#25670}
These methods for used for compressed libraries, where GetSource* functions
contained the compressed sources and [GS]etRawSource* the uncompressed
sources. This is dead code since the API no longer supports compression.
(If you need/want compressed sources, use the external startup data and
compress/uncompress on the Embedder's side.)
BUG=
Review URL: https://codereview.chromium.org/772853003
Cr-Commit-Position: refs/heads/master@{#25666}
This patch introduces a new inline runtime function,
%_HasFastPackedElements(), and uses it both in the implementation
of the 'in' operator and in the array builtins to speed
up testing for the existence of an index in an array.
In testing with the microbenchmark on the attached bug,
for example, the runtime goes from 326ms to 66ms.
A reviewer might ask whether the HAS_INDEX macro is worthwhile,
and I tried the same example without it, which pushed the
microbenchmark up to 157ms. So it seems it's worth it to
avoid the function call to IN() if we know we're dealing
with arrays and numbers.
BUG=v8:3701
LOG=n
Review URL: https://codereview.chromium.org/754863002
Cr-Commit-Position: refs/heads/master@{#25665}
In the gc-tracer, we check whether we're marking to figure out which
part of the mark compact we're in. If we aborted incremental marking for
whatever reason, the check fails and we might later run into trouble
BUG=none
R=hpayer@chromium.org
LOG=n
Review URL: https://codereview.chromium.org/783453003
Cr-Commit-Position: refs/heads/master@{#25663}
