This change does the following:
a.) Remove unused fields from the Typer.
b.) Move some interesting unions to types.h.
c.) Reduce Typer constructor overhead.
d.) Avoid heap allocation in the Typer.
R=jarin@chromium.org
Review URL: https://codereview.chromium.org/1199903002
Cr-Commit-Position: refs/heads/master@{#29178}
This fixes a terrible interaction of code flushing and the clearing of
optimized code maps hanging off a SharedFunctionInfo. The following is
what happened:
1) Incremental marking cleared map in SharedFunctionInfo s, however it
was not enqueued as a flushing candidate because one JSFunction f1
still had optimized code.
2) Deoptimization of f1 made s eligible for code flushing.
3) Optimization of f2 added new entry to optimized code map of s.
4) The JSFunction f2 became unreachable and hence is never marked.
5) Incremental marking now visits f1, finds it eligible for flushing,
also s is eligible for flushing, both are enqueued.
6) Marking finishes, code flusher clears f1 and s, but the optimized
code map of s still contains an entry.
7) Boom!
R=ulan@chromium.org,hpayer@chromium.org
TEST=mjsunit/es6/generators-iteration
BUG=v8:3803
LOG=N
Review URL: https://codereview.chromium.org/1197713004
Cr-Commit-Position: refs/heads/master@{#29177}
This is another missing piece in the puzzle towards general inlining.
The fact that we can combine this with context specialization is a nice
bonus, and not necessarily a requirement.
R=jarin@chromium.org
Review URL: https://codereview.chromium.org/1198193002.
Cr-Commit-Position: refs/heads/master@{#29176}
This way we don't need the separate late control reduction pass over the
graph, plus we can also reduce dead code recognized by generic lowering.
R=jarin@chromium.org
Review URL: https://codereview.chromium.org/1198923002
Cr-Commit-Position: refs/heads/master@{#29174}
Turbofan needs to pass vector slots around for named and keyed stores.
Also, the CL addresses a missing slot for ClassLiterals.
BUG=
Review URL: https://codereview.chromium.org/1178363002
Cr-Commit-Position: refs/heads/master@{#29173}
Object.setPrototypeOf() throws a TypeError if value would create a
cycle. Previously a plain Error was thrown rather than a TypeError.
BUG=v8:4197
R=mike@bocoup.com
LOG=N
Review URL: https://codereview.chromium.org/1198523002
Cr-Commit-Position: refs/heads/master@{#29169}
Reason:
Regressions in various benchmarks.
Revert "Revert of Revert of [strong] Implement strong mode restrictions on property access (patchset #1 id:1 of https://codereview.chromium.org/1189153002/)"
This reverts commit 41405c0470.
Revert "X87: Revert of Revert of [strong] Implement strong mode restrictions on property access."
This reverts commit 48de5f4d6b.
Revert "Fix overlapping KeyedLoadIC bitfield."
This reverts commit 4e6c956abf.
Revert "MIPS64: Fix 'Revert of Revert of [strong] Implement strong mode restrictions on property access'."
This reverts commit 74f97b0d2a.
BUG=
Review URL: https://codereview.chromium.org/1199493002
Cr-Commit-Position: refs/heads/master@{#29166}
Reason for revert:
Fails on V8 Linux - isolates (http://build.chromium.org/p/client.v8/builders/V8%20Linux%20-%20isolates/builds/4128)
Original issue's description:
> Add d8 API for spawning function on a new thread (Second try)
>
> This API closely matches the Worker API. The differences:
>
> 1) The argument to the Worker constructor is a function to run, not a script.
> 2) Receiving a message from a worker is a synchronous API (as there is no event
> loop).
>
> The serialization done here is not robust as the real DOM implementation. For
> example, recursive data structures or otherwise duplicated objects are not
> allowed.
>
> BUG=chromium:497295
> R=jochen@chromium.org
> LOG=n
>
> Review URL: https://codereview.chromium.org/1185643004
>
> Cr-Commit-Position: refs/heads/master@{#29126}
>
> Committed: https://crrev.com/ec2eaf712ecee6b4891c0458f2397e04a1f9b339
> Cr-Commit-Position: refs/heads/master@{#29158}
TBR=jochen@chromium.org
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
BUG=chromium:497295
Review URL: https://codereview.chromium.org/1191373005
Cr-Commit-Position: refs/heads/master@{#29161}
This API closely matches the Worker API. The differences:
1) The argument to the Worker constructor is a function to run, not a script.
2) Receiving a message from a worker is a synchronous API (as there is no event
loop).
The serialization done here is not robust as the real DOM implementation. For
example, recursive data structures or otherwise duplicated objects are not
allowed.
BUG=chromium:497295
R=jochen@chromium.org
LOG=n
Review URL: https://codereview.chromium.org/1185643004
Cr-Commit-Position: refs/heads/master@{#29126}
Review URL: https://codereview.chromium.org/1195613003
Cr-Commit-Position: refs/heads/master@{#29158}
Each Script object now keeps a WeakFixedArray of SharedFunctionInfo
objects created from this script.
This way, when compiling a function, we do not create duplicate shared
function info objects when recompiling with either compiler.
This fixes a class of issues in the debugger, where we set break points
on one shared function info, but functions from duplicate shared function
infos are not affected.
LOG=N
BUG=v8:4132
Review URL: https://codereview.chromium.org/1183733006
Cr-Commit-Position: refs/heads/master@{#29151}
The three different concerns that the ControlReducer used to deal with
are now properly separated into
a.) DeadCodeElimination, which is a regular AdvancedReducer, that
propagates Dead via control edges,
b.) CommonOperatorReducer, which does strength reduction on common
operators (i.e. Branch, Phi, and friends), and
c.) GraphTrimming, which removes dead->live edges from the graph.
This will make it possible to run the DeadCodeElimination together with
other passes that actually introduce Dead nodes, i.e. typed lowering;
and it opens the door for general inlining without two stage fix point
iteration.
To make the DeadCodeElimination easier and more uniform, we basically
reverted the introduction of DeadValue and DeadEffect, and changed the
Dead operator to produce control, value and effect. Note however that
this is not a requirement, but merely a way to make dead propagation
easier and more uniform. We could always go back and decide to have
different Dead operators if some other change requires that.
Note that there are several additional opportunities for cleanup now,
i.e. OSR deconstruction could be a regular reducer now, and we don't
need to use TheHole as dead value marker in the GraphReducer. And we can
actually run the dead code elimination together with the other passes
instead of using separate passes over the graph. We will do this in
follow up CLs.
R=jarin@chromium.org, mstarzinger@chromium.org
Review URL: https://codereview.chromium.org/1193833002
Cr-Commit-Position: refs/heads/master@{#29146}
Adds a script for converting gc nvp output into csv files. Factors out common
code in gc-nvp-trace-processor.py to gc_nvp_common.py to be shared by both
scripts. Fixes a couple of issues in nvp parsing code.
Review URL: https://codereview.chromium.org/1175113008
Cr-Commit-Position: refs/heads/master@{#29141}
Not doing so could result in this scenario:
- We ensure that the string table is large enough. It is.
- We compile code stubs, which triggers a GC.
- The GC clears string table entries.
- This increases the number of deleted entries in the table.
- When the deserializer hooks up internalized strings into the
table, we ensure that the table is large enough every time.
- Due to changed number of deleted entries, the heuristic
decides to expand the string table.
- Allocation during deserialization causes assertion to fail.
BUG=chromium:502085
LOG=N
Review URL: https://codereview.chromium.org/1197633002
Cr-Commit-Position: refs/heads/master@{#29134}
We block constant pool up to the call stub instruction, but the check for code size
includes the next instruction after the call instruction. That instruction can
emit constant pool.
BUG=chromium:500831
LOG=NO
TEST=mjsunit/regress/regress-500831
Review URL: https://codereview.chromium.org/1189123003
Cr-Commit-Position: refs/heads/master@{#29132}
