At the moment, the whole WebAssembly compilation may run in a single
background task. On a low-end device, this can mean that the background
thread is busy for seconds and thereby blocks other tasks, see e.g.
https://crbug.com/914757.
With this CL we re-schedule compilation tasks after every 50ms. These
50ms are an arbitrary number. I don't want to introduce too much
overhead, but since this is in the background we also don't have to
make tasks super short.
Tasks which are going to compile with TurboFan will be posted with
lower priority.
This change requires changes in the CancelableTaskManager. At the
moment it is not possible that a background task posts a new task
which is managed by the same task manager as itself. The problem is
about how to deal with another thread which calls CancelAndWait
concurrently. At the moment, if a new task gets posted after the call
to CancelAndWait, then `CHECK(!canceled_)` in
CancelableTaskManager::Register will fail. If we used a lock to
synchronize the calls to CancelAndWait and Register, then there would
be a deadlock, where the thread which calls CancelAndWait waits for
the task which wants to call Register, but at the same time blocks that
task by holding the lock.
With the change here, posting a task after the call to CancelAndWait
will just immediately cancel the new task. This matches the behavior
you would get if CancelAndWait is called right after calling Register.
Bug: chromium:914757
Change-Id: I6d57aba161db8a915ec0d745658e0c28d25219a8
Reviewed-on: https://chromium-review.googlesource.com/c/1411884
Commit-Queue: Andreas Haas <ahaas@chromium.org>
Reviewed-by: Clemens Hammacher <clemensh@chromium.org>
Reviewed-by: Michael Lippautz <mlippautz@chromium.org>
Cr-Commit-Position: refs/heads/master@{#58898}
It's too easy to implicitly cast it to bool, as we did in several tests.
Also, move TryAbortResult out of CancelableTaskManager to avoid too much
typing when referencing one of the enum values.
R=mstarzinger@chromium.org
Bug: v8:8238
Change-Id: Ia3fa8597428876217bc86f9b8b31c21ae4846fa1
Reviewed-on: https://chromium-review.googlesource.com/c/1326027
Commit-Queue: Clemens Hammacher <clemensh@chromium.org>
Reviewed-by: Michael Starzinger <mstarzinger@chromium.org>
Cr-Commit-Position: refs/heads/master@{#57363}
LockGuard is mostly used with Mutex. Since both are defined outside the
internal namespace, we often have to write
{base::LockGuard<base::Mutex>}. This CL shortens this to
{base::MutexGuard} across the code base
R=mlippautz@chromium.org
Bug: v8:8238
Cq-Include-Trybots: luci.chromium.try:linux_chromium_headless_rel;master.tryserver.blink:linux_trusty_blink_rel
Change-Id: I020d5933b73aafb98c4b72e3bb2dfd07c979ba73
Reviewed-on: https://chromium-review.googlesource.com/c/1278796
Reviewed-by: Michael Lippautz <mlippautz@chromium.org>
Reviewed-by: Yang Guo <yangguo@chromium.org>
Commit-Queue: Clemens Hammacher <clemensh@chromium.org>
Cr-Commit-Position: refs/heads/master@{#56612}
The compilation state of a native module is now extracted into its own
datastructure. It reflects which functions are left to compile, and contains
task managers to accomplish parallel and asynchronous compilation.
Bug:
Change-Id: I45308c7b32ba78e6c83f2d260990846a653bbd9c
Reviewed-on: https://chromium-review.googlesource.com/958865
Commit-Queue: Kim-Anh Tran <kimanh@google.com>
Reviewed-by: Ben Titzer <titzer@chromium.org>
Reviewed-by: Andreas Haas <ahaas@chromium.org>
Reviewed-by: Clemens Hammacher <clemensh@chromium.org>
Cr-Commit-Position: refs/heads/master@{#52013}
They were only limited to 32 bit when using the internal Hashmap. Since
this has changed alreay some time ago, we can switch to 64 bit ids and
check that we never overflow.
Bug:
Change-Id: Ia6c6d02d6b5e555c6941185a79427dc4aa2a1d62
Reviewed-on: https://chromium-review.googlesource.com/598229
Commit-Queue: Michael Lippautz <mlippautz@chromium.org>
Reviewed-by: Ulan Degenbaev <ulan@chromium.org>
Cr-Commit-Position: refs/heads/master@{#47085}
This CL removes the isolate_ property from the Cancelable class. The
isolate is actually not used in the class itself, only in sub-classes
which can store the isolate themselves.
R=jochen@chromium.org, mlippautz@chromium.org
Change-Id: If102fdea41b81fb3cb25782728f7fcbce9642815
Reviewed-on: https://chromium-review.googlesource.com/516704
Reviewed-by: Jochen Eisinger <jochen@chromium.org>
Commit-Queue: Andreas Haas <ahaas@chromium.org>
Cr-Commit-Position: refs/heads/master@{#45535}
Reason for revert:
Timeout on gc stress bot.
Original issue's description:
> [heap] Cancel tasks before tearing down the heap.
>
> BUG=chromium:654343
>
> Committed: https://crrev.com/9365463d6f1fc11b2369fff63ca0ac825c61eae4
> Cr-Commit-Position: refs/heads/master@{#40272}
TBR=mlippautz@chromium.org
# Skipping CQ checks because original CL landed less than 1 days ago.
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
BUG=chromium:654343
Review-Url: https://codereview.chromium.org/2414073002
Cr-Commit-Position: refs/heads/master@{#40279}
Since {CancelAndWait} blocks on the tasks that are still present in the internal
hashmap, we are not allowed to remove the task upon trying to cancel it using
{TryAbort}.
The previous implementation suffered from a bug where:
1) The task was created and handed over to the platform.
2) The task was started by the platform, setting it to running state.
3) We called {TryAbort}, effectively removing it from the manager, but failing
to cancel (as it was already running)
4) All tasks finished running, indicating this with their own semaphore.
5) The platform was stuck (scheduling) before destroying the task.
6) Main thread finished its work, waiting for all the necessary tasks, and the
isolate terminated.
7) The platform destroyed the task, calling the destructor, calling into an
already freed isolate.
BUG=chromium:524425
LOG=N
Review URL: https://codereview.chromium.org/1460763004
Cr-Commit-Position: refs/heads/master@{#32118}
This change binds each {Cancelable} task to a so-called {CancelableTaskManager},
which is then used to handle concurrent cancelation as well as synchronizing
shutdown for already running tasks. Since ownership of tasks is transferred to
the platform executing a task (destructor), handling in the manager uses integer
ids. Note that this also mitigates (modulo integer size) the ABA problem.
All handling of {Cancelable} tasks is now encapsulated into the corresponding
manager, which is instantiated for each isolate.
R=hpayer@chromium.org
BUG=chromium:524425
LOG=N
CQ_EXTRA_TRYBOTS=tryserver.v8:v8_linux_arm64_gc_stress_dbg;tryserver.v8:v8_linux_gc_stress_dbg;tryserver.v8:v8_mac_gc_stress_dbg;tryserver.v8:v8_linux64_msan_rel;tryserver.v8:v8_linux64_tsan_rel;tryserver.v8:v8_mac64_asan_rel
Review URL: https://codereview.chromium.org/1409993012
Cr-Commit-Position: refs/heads/master@{#31997}
This CL us a pure refactoring that makes an empty compilation unit
including just "isolate.h" or "contexts.h" but not "objects-inl.h"
compile without warnings or errors. This is needed to further reduce
the header dependency tangle.
R=bmeurer@chromium.org
Review URL: https://codereview.chromium.org/1322883002
Cr-Commit-Position: refs/heads/master@{#30500}
