Three bugs were fixed:
1. BitVector::Equals was not fully fixed by the original CL.
2. We need to add old objects to the deferred marking queue during
RememberLiveTemporaries().
3. The thread being scanned in RestoreWriteBarrierInvariant may not
be scheduled, so we cannot use its store buffer block.
In addition, this changed uncovered another bug fixed in:
https://dart-review.googlesource.com/c/sdk/+/138960.
Original CL is in patchset 3.
This reverts commit 30a12a349e.
Change-Id: I36169b09563998ed5b3c3eac70ee0ebe78853e62
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/138920
Reviewed-by: Ryan Macnak <rmacnak@google.com>
Commit-Queue: Samir Jindel <sjindel@google.com>
This reverts commit eff1a9ff97.
Reason for revert:
Causes flaky hits of RELEASE_ASSERT in marker.cc, see b/151131634.
Original change's description:
> Re-land "[vm] Aggressive write-barrier elimination."
>
> The original revision is in Patchset 3.
>
> Four bugs were fixed:
>
> 1. JoinEntryInstr::SuccessorCount() is not the correct way to get the
> number of successor blocks from the Join block;
> JoinEntryInstr::last_instruction()->SuccessorCount() must be used
> instead.
>
> 2. BitVector::Equals() was non-deterministically returning 'false'
> for equal vectors.
>
> 3. All blocks need to be processed at least once during the Analysis
> phase (not only in the SaveResults phase).
>
> 4. We were not removing write barriers from StoreIndexed instructions,
> even though we had support for it.
>
> This reverts commit 7fd8ad5a2d.
>
> Fixes https://github.com/dart-lang/sdk/issues/40780
>
> Change-Id: I9650ec2c547ec49cf88ca0524e14f6c245621f6a
> Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/138086
> Commit-Queue: Samir Jindel <sjindel@google.com>
> Reviewed-by: Martin Kustermann <kustermann@google.com>
> Reviewed-by: Ryan Macnak <rmacnak@google.com>
TBR=kustermann@google.com,rmacnak@google.com,sjindel@google.com
Change-Id: If9afd84465175fad2431405a97e9293c8bd5e476
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/138808
Reviewed-by: Martin Kustermann <kustermann@google.com>
Commit-Queue: Martin Kustermann <kustermann@google.com>
The original revision is in Patchset 3.
Four bugs were fixed:
1. JoinEntryInstr::SuccessorCount() is not the correct way to get the
number of successor blocks from the Join block;
JoinEntryInstr::last_instruction()->SuccessorCount() must be used
instead.
2. BitVector::Equals() was non-deterministically returning 'false'
for equal vectors.
3. All blocks need to be processed at least once during the Analysis
phase (not only in the SaveResults phase).
4. We were not removing write barriers from StoreIndexed instructions,
even though we had support for it.
This reverts commit 7fd8ad5a2d.
Fixes https://github.com/dart-lang/sdk/issues/40780
Change-Id: I9650ec2c547ec49cf88ca0524e14f6c245621f6a
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/138086
Commit-Queue: Samir Jindel <sjindel@google.com>
Reviewed-by: Martin Kustermann <kustermann@google.com>
Reviewed-by: Ryan Macnak <rmacnak@google.com>
Begin concurrent marking when available space in old-space is less than 5% or than the size of new-space, whichever comes first.
Avoids concurrent marking starting too late in larger heaps, where the program would exceed the hard threshold before concurrent marking would finish.
Change-Id: I6a1991331a1450ffa647bf0e21dfc19e82537268
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/134466
Commit-Queue: Ryan Macnak <rmacnak@google.com>
Reviewed-by: Siva Annamalai <asiva@google.com>
Some embedder code relies on the presence of a current isolate equal to the isolate that created the handle. This code will be need to be updated to be isolate-group aware before we can run finalizers only in the context of the isolate group instead of an isolate.
Bug: https://github.com/dart-lang/sdk/issues/40836
Change-Id: I38956c556df551e48ee7e4d0822aa5def8dc4fc5
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/138013
Commit-Queue: Ryan Macnak <rmacnak@google.com>
Reviewed-by: Siva Annamalai <asiva@google.com>
It incorrectly assumed that all stores in Dart code write to Instances.
There is actually one exception, Contexts, which do not inherit from Instance.
I've added asserts to ensure this kind of bug cannot resurface.
The original change is in patchset 4.
Change-Id: Ic2d8d05e70a4de738eb9fb5980487b4f27111b8c
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/136221
Commit-Queue: Samir Jindel <sjindel@google.com>
Reviewed-by: Martin Kustermann <kustermann@google.com>
Reviewed-by: Ryan Macnak <rmacnak@google.com>
This CL:
* Moves [Heap]/[SharedClassTable] from [Isolate] to [IsolateGroup], which
will make all isolates in the group use the same heap. The GC will use
the shared class table for object size information.
* Adds support for entering/leaving an isolate group as a helper thread
(e.g. via [Thread::EnterIsolateGroupAsHelper]). The current active
isolate group can be accessed via TLS `IsolateGroup::Current()` or
`Thread::isolate_group_`. When entering as a helper thread there will be
no current isolate.
* Changes the GC to use the above mechanism and ensures GC works without
a currently active isolate. The GC will use information purely available via
[IsolateGroup]. The GC will iterate all isolates within an isolate
group e.g. for scanning roots.
* Makes spawning of new isolates start in their own isolate group.
Once the isolate is fully functional it's heap will be merged into
the original isolate group
* Moves ApiState, containing persistent and weak persistent handles,
from [Isolate] to [IsolateGroup], plus adds appropriate locking.
Issue https://github.com/dart-lang/sdk/issues/36097
Change-Id: Ia8e1d8aa78750e8400864200f4825395a182c004
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/126646
Commit-Queue: Martin Kustermann <kustermann@google.com>
Reviewed-by: Ryan Macnak <rmacnak@google.com>
The final TryAllocateDataLocked will consider all free list buckets and a new page. These cases are all covered by the preceding TryAllocateSmallLock (free list small buckets) and TryAllocateDataBumpLocked (free list large buckets and new pages), so this retry will always fail.
Change-Id: I9ba3174d588085130273ee6e7920888d3abca0ab
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/135581
Reviewed-by: Siva Annamalai <asiva@google.com>
Reviewed-by: Martin Kustermann <kustermann@google.com>
Commit-Queue: Ryan Macnak <rmacnak@google.com>
When these were originally added, we thought we would be able to use them in place of ifdefs and rely on the compiler optimizations and linker GC to remove things. This turned out not to reliably remove what we wanted removed, so we ended up with the ifdefs anyway.
Change-Id: I62e74d60d92b18a688b9dffaf77b1440c10a07ff
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/134402
Commit-Queue: Ryan Macnak <rmacnak@google.com>
Reviewed-by: Ben Konyi <bkonyi@google.com>
This scope can sneak objects larger than kAllocatablePageSize directly into non-large heap pages.
This scope also holds a heap lock outside of the GC, which is likely to create issues when multiple isolates run on the same heap.
Change-Id: Ic56277c83ffccfe2e1396ed773dd30b8d172bbba
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/134040
Reviewed-by: Martin Kustermann <kustermann@google.com>
Reviewed-by: Siva Annamalai <asiva@google.com>
This reverts commit 9eb531bde4.
Reason for revert: Bots are red. Some tests are failing.
https://ci.chromium.org/p/dart/builders/ci.sandbox/vm-kernel-precomp-obfuscate-linux-release-x64/6039https://ci.chromium.org/p/dart/builders/ci.sandbox/vm-kernel-precomp-android-release-arm_x64/957
Original change's description:
> [vm] Add support for real unboxed floating point fields in AOT
>
> Non-nullable floating point fields (double, Float32x4, Float64x2)
> are fully unboxed in their classes.
>
> A bitmap for each class was added to the shared class table in order to keep
> track of the pointers of the classes. Since all classes in Flutter Gallery
> have less than 64 fields, the bitmap is represented by a 64 bit integer and
> fields whose offset is more than 64 words are not unboxed.
>
> The instance sizes and field offsets might change between target and host
> in cross-compilation, since the number of words used to store unboxed fields
> may differ.
>
> dart-aot Xeon
>
> SplayLatency -4.62%
> SplayHarderLatency -4.17%
> NavierStokes -2.20%
> Tracer 8.72%
> ParticleSystemPaint 2.90%
> NBodySIMD 8.35%
> NBody 25.59%
>
> With hack TFA to make doubles in Rect/Offset/Size classes in flutter non-nullable:
>
> flutter arm-v8:
>
> gallery total size: -1%
>
> matrix_utils_transform_rect_perspective -16.70% (less is better)
> matrix_utils_transform_rect_affine -31.82% (less is better)
> matrix_utils_transform_point_perspective -24.90% (less is better)
> matrix_utils_transform_point_affine) -27.26% (less is better)
> rrect_contains_bench -4.719% (less is better)
>
> Change-Id: I9ae09c9c3167d99f9efd071a92937aa51093fd1d
> Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/131824
> Commit-Queue: Victor Agnez Lima <victoragnez@google.com>
> Reviewed-by: Martin Kustermann <kustermann@google.com>
> Reviewed-by: Ryan Macnak <rmacnak@google.com>
> Reviewed-by: Samir Jindel <sjindel@google.com>
TBR=kustermann@google.com,rmacnak@google.com,sjindel@google.com,victoragnez@google.com
Change-Id: Ic73858f6adb7f55c4129d4f46ff4731b378cb634
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/134020
Reviewed-by: Zichang Guo <zichangguo@google.com>
Commit-Queue: Zichang Guo <zichangguo@google.com>
With concurrent sweeping, the amount of capacity freed by a major GC is not known at the time growth is evaluated. This consistently biases the policy into growing more than it would with an accurate capacity from stop-the-world sweeping.
This is more interesting for larger heaps and when there is a tighter memory limit.
For a very large dart2js compile
Max Heap Capacity: 13.057 -> 11.861 GB
Compile Time: 545 -> 472 seconds
For dart2js self-compile
Max Heap Capacity: 1.343 -> 1.385 GB
Compile Time: 43.4 -> 43.5 seconds
Bug: https://github.com/dart-lang/sdk/issues/21364
Change-Id: If047e22a9cd9da6b34c8e499a39fde97427a69b9
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/133660
Commit-Queue: Ryan Macnak <rmacnak@google.com>
Reviewed-by: Siva Annamalai <asiva@google.com>
Non-nullable floating point fields (double, Float32x4, Float64x2)
are fully unboxed in their classes.
A bitmap for each class was added to the shared class table in order to keep
track of the pointers of the classes. Since all classes in Flutter Gallery
have less than 64 fields, the bitmap is represented by a 64 bit integer and
fields whose offset is more than 64 words are not unboxed.
The instance sizes and field offsets might change between target and host
in cross-compilation, since the number of words used to store unboxed fields
may differ.
dart-aot Xeon
SplayLatency -4.62%
SplayHarderLatency -4.17%
NavierStokes -2.20%
Tracer 8.72%
ParticleSystemPaint 2.90%
NBodySIMD 8.35%
NBody 25.59%
With hack TFA to make doubles in Rect/Offset/Size classes in flutter non-nullable:
flutter arm-v8:
gallery total size: -1%
matrix_utils_transform_rect_perspective -16.70% (less is better)
matrix_utils_transform_rect_affine -31.82% (less is better)
matrix_utils_transform_point_perspective -24.90% (less is better)
matrix_utils_transform_point_affine) -27.26% (less is better)
rrect_contains_bench -4.719% (less is better)
Change-Id: I9ae09c9c3167d99f9efd071a92937aa51093fd1d
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/131824
Commit-Queue: Victor Agnez Lima <victoragnez@google.com>
Reviewed-by: Martin Kustermann <kustermann@google.com>
Reviewed-by: Ryan Macnak <rmacnak@google.com>
Reviewed-by: Samir Jindel <sjindel@google.com>
Previously large (>256k) typed arrays would be initialized 3 times:
- The OS initializing the new virtual memory to zero
- Object::InitializeObject initializing to object null
- TypedData::New reinitializing to zero
Change-Id: I432944919ab68aecb2038bba245594bbbaed9093
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/130928
Commit-Queue: Ryan Macnak <rmacnak@google.com>
Reviewed-by: Alexander Aprelev <aam@google.com>
Reviewed-by: Martin Kustermann <kustermann@google.com>
In some workloads, the heap is mostly large pages.
Move large executable pages from the large list to the executable list. They do not require special treatment because they are large during sweeping because we never truncate an Instructions object. They still require being swept immediately because they are executable so that code protections can be flipped before resuming Dart execution.
Fix off-by-one errors in SnapshotGraph.objects and SnapshotClass.instances discovered as a result of the large page list changing from prepending to appending.
Bug: https://github.com/flutter/flutter/issues/48360
Change-Id: If6d9137fd36c9b18205493166d586b9a7cee4b1c
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/130823
Commit-Queue: Ryan Macnak <rmacnak@google.com>
Reviewed-by: Siva Annamalai <asiva@google.com>
The VM does not include image pages in its usual accounting of heap usage and capacity because they are immortal and don't contribute to marking or sweeping work. The heap snapshot does include objects on these pages, so it should include the pages in its notion of usage and capacity, otherwise the heap-snapshot can effectively claim it has more usage than capacity.
Fixes negative fragmentation reported by the heap snapshot page.
Change-Id: Ib9f3c395ee5a6e51ee9b4c9a834843851426910d
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/128565
Reviewed-by: Siva Annamalai <asiva@google.com>
Reviewed-by: Ben Konyi <bkonyi@google.com>
Samir Jindel