STW marking is no longer O(new-space), so there's no longer a reason to delay finalizing marking hoping for a scavenge to make new-space mostly empty.
TEST=ci
Change-Id: Ie782e88852714d30e0c75aa9aecac62e56c434ce
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/319880
Reviewed-by: Martin Kustermann <kustermann@google.com>
Commit-Queue: Ryan Macnak <rmacnak@google.com>
- Initial and final marking no longer visit all of new-space, reducing the STW pause for major GC.
- A scavenge during concurrent marking must forward / filter objects in the marking worklist that are moved / collected, increasing the STW pause for minor GC.
- Unreachable intergenerational cycles and weak references are collected in the next mark-sweep instead of first requiring enough scavenges to promote the whole cycle or weak target into old-space.
- Artificial minor GCs are no longer needed to avoid memory leaks from back-to-back major GCs.
- reachabilityBarrier is now just a count of major GCs.
TEST=ci
Change-Id: Ic7754e8d972763654eae2b7faa8670735d9cda3f
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/340644
Reviewed-by: Siva Annamalai <asiva@google.com>
Commit-Queue: Ryan Macnak <rmacnak@google.com>
This reverts commit 5daaa7d9eb.
Reason for revert: internal crashes
Original change's description:
> [vm, gc] Mark through new-space.
>
> - Initial and final marking no longer visit all of new-space, reducing the STW pause for major GC.
> - A scavenge during concurrent marking must forward / filter objects in the marking worklist that are moved / collected, increasing the STW pause for minor GC.
> - Unreachable intergenerational cycles and weak references are collected in the next mark-sweep instead of first requiring enough scavenges to promote the whole cycle or weak target into old-space.
> - Artificial minor GCs are no longer needed to avoid memory leaks from back-to-back major GCs.
> - reachabilityBarrier is now just a count of major GCs.
>
> TEST=ci
> Change-Id: I4a6a23273d8ecb78c640f054731d4ceb737bfc4d
> Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/325840
> Reviewed-by: Siva Annamalai <asiva@google.com>
> Commit-Queue: Ryan Macnak <rmacnak@google.com>
Change-Id: I8a50074db343c63c14f0487ae8b4f5fee2c4ae76
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/330720
Reviewed-by: Ryan Macnak <rmacnak@google.com>
Commit-Queue: Slava Egorov <vegorov@google.com>
Bot-Commit: Rubber Stamper <rubber-stamper@appspot.gserviceaccount.com>
- Initial and final marking no longer visit all of new-space, reducing the STW pause for major GC.
- A scavenge during concurrent marking must forward / filter objects in the marking worklist that are moved / collected, increasing the STW pause for minor GC.
- Unreachable intergenerational cycles and weak references are collected in the next mark-sweep instead of first requiring enough scavenges to promote the whole cycle or weak target into old-space.
- Artificial minor GCs are no longer needed to avoid memory leaks from back-to-back major GCs.
- reachabilityBarrier is now just a count of major GCs.
TEST=ci
Change-Id: I4a6a23273d8ecb78c640f054731d4ceb737bfc4d
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/325840
Reviewed-by: Siva Annamalai <asiva@google.com>
Commit-Queue: Ryan Macnak <rmacnak@google.com>
Instead of making a StaticCall to _TypedListBase.nativeSetRange
inside _memMoveN, make a CCall to the memmove leaf runtime entry.
Rename _TypedListBase._nativeSetRange to _setClampedRange, since
it's now only used when per-element clamping is necessary.
Fix the load optimizer so that loads of unboxed fields from freshly
allocated objects do not have the tagged null value forwarded
as their initial post-allocation value.
TEST=co19{,_2}/LibTest/typed_data lib{,_2}/typed_data
corelib{,_2}/list_test
vm/cc/LoadOptimizer_LoadDataFieldOfNewTypedData
Issue: https://github.com/dart-lang/sdk/issues/42072
Change-Id: Ib82e24a5b3287fa53099fffd3b563a27d777507e
Cq-Include-Trybots: luci.dart.try:vm-aot-linux-debug-simarm_x64-try,vm-aot-linux-debug-x64-try,vm-aot-linux-debug-x64c-try,vm-kernel-linux-debug-x64-try,vm-kernel-precomp-linux-debug-x64-try,vm-linux-debug-x64-try,vm-linux-debug-x64c-try,vm-mac-debug-arm64-try,vm-aot-linux-release-simarm_x64-try,vm-aot-linux-release-x64-try,vm-aot-mac-release-arm64-try,vm-aot-msan-linux-release-x64-try,vm-msan-linux-release-x64-try,vm-aot-tsan-linux-release-x64-try,vm-tsan-linux-release-x64-try,vm-linux-release-x64-try,vm-mac-release-arm64-try,vm-kernel-precomp-linux-release-x64-try
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/324080
Reviewed-by: Alexander Markov <alexmarkov@google.com>
Commit-Queue: Tess Strickland <sstrickl@google.com>
This reverts commit 3fb88e4c66.
Reason for revert: b/297175670
Original change's description:
> [vm, gc] Mark through new-space.
>
> - Initial and final marking no longer visit all of new-space, reducing the STW pause for major GC.
> - A scavenge during concurrent marking must forward / filter objects in the marking worklist that are moved / collected, increasing the STW pause for minor GC.
> - Unreachable intergenerational cycles and weak references are collected in the next mark-sweep instead of first requiring enough scavenges to promote the whole cycle or weak target into old-space.
> - Artificial minor GCs are no longer needed to avoid memory leaks from back-to-back major GCs.
> - reachabilityBarrier is now just a count of major GCs.
>
> TEST=ci
> Change-Id: I6362802cd93ba5ba9c39f116ddff82e4feb4c312
> Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/321304
> Commit-Queue: Ryan Macnak <rmacnak@google.com>
> Reviewed-by: Siva Annamalai <asiva@google.com>
Change-Id: I33075156160dc35861355d738a5776b74dce88b9
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/322344
Reviewed-by: Martin Kustermann <kustermann@google.com>
Auto-Submit: Ivan Inozemtsev <iinozemtsev@google.com>
Commit-Queue: Martin Kustermann <kustermann@google.com>
- Initial and final marking no longer visit all of new-space, reducing the STW pause for major GC.
- A scavenge during concurrent marking must forward / filter objects in the marking worklist that are moved / collected, increasing the STW pause for minor GC.
- Unreachable intergenerational cycles and weak references are collected in the next mark-sweep instead of first requiring enough scavenges to promote the whole cycle or weak target into old-space.
- Artificial minor GCs are no longer needed to avoid memory leaks from back-to-back major GCs.
- reachabilityBarrier is now just a count of major GCs.
TEST=ci
Change-Id: I6362802cd93ba5ba9c39f116ddff82e4feb4c312
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/321304
Commit-Queue: Ryan Macnak <rmacnak@google.com>
Reviewed-by: Siva Annamalai <asiva@google.com>
This reverts commit e95e7b8e96.
Reason for revert: suspected cause to non-deterministic AOT binaries crashes, b/296014654
Original change's description:
> [vm, gc] Mark through new-space.
>
> - Initial and final marking no longer visit all of new-space, reducing the STW pause for major GC.
> - A scavenge during concurrent marking must forward / filter objects in the marking worklist that are moved / collected, increasing the STW pause for minor GC.
> - Unreachable intergenerational cycles and weak references are collected in the next mark-sweep instead of first requiring enough scavenges to promote the whole cycle or weak target into old-space.
> - Artificial minor GCs are no longer needed to avoid memory leaks from back-to-back major GCs.
> - reachabilityBarrier is now just a count of major GCs.
>
> TEST=ci
> Change-Id: I8c2c64b120766571b62d3bd8dab37ae81c2dca98
> Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/319583
> Commit-Queue: Ryan Macnak <rmacnak@google.com>
> Reviewed-by: Siva Annamalai <asiva@google.com>
Change-Id: Idda542c7c657d4f14c836423b173c9b067132212
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/320820
Reviewed-by: Martin Kustermann <kustermann@google.com>
Bot-Commit: Rubber Stamper <rubber-stamper@appspot.gserviceaccount.com>
Commit-Queue: Ilya Yanok <yanok@google.com>
- Initial and final marking no longer visit all of new-space, reducing the STW pause for major GC.
- A scavenge during concurrent marking must forward / filter objects in the marking worklist that are moved / collected, increasing the STW pause for minor GC.
- Unreachable intergenerational cycles and weak references are collected in the next mark-sweep instead of first requiring enough scavenges to promote the whole cycle or weak target into old-space.
- Artificial minor GCs are no longer needed to avoid memory leaks from back-to-back major GCs.
- reachabilityBarrier is now just a count of major GCs.
TEST=ci
Change-Id: I8c2c64b120766571b62d3bd8dab37ae81c2dca98
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/319583
Commit-Queue: Ryan Macnak <rmacnak@google.com>
Reviewed-by: Siva Annamalai <asiva@google.com>
This reverts commit 6194209b28.
Reason for revert: issues on arm32
Original change's description:
> [vm, gc] Mark through new-space.
>
> - Initial and final marking no longer visit all of new-space, reducing the STW pause for major GC.
> - A scavenge during concurrent marking must forward / filter objects in the marking worklist that are moved / collected, increasing the STW pause for minor GC.
> - Unreachable intergenerational cycles and weak references are collected in the next mark-sweep instead of first requiring enough scavenges to promote the whole cycle or weak target into old-space.
> - Artificial minor GCs are no longer needed to avoid memory leaks from back-to-back major GCs.
> - reachabilityBarrier is now just a count of major GCs.
>
> TEST=ci
> Change-Id: I3668a2e56821f9eadf96e38c228dab27be656016
> Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/309826
> Reviewed-by: Siva Annamalai <asiva@google.com>
> Commit-Queue: Ryan Macnak <rmacnak@google.com>
Change-Id: I434eb595c9e7858efc8c9b07cbca954e5649f506
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/319321
Commit-Queue: Ryan Macnak <rmacnak@google.com>
Reviewed-by: Siva Annamalai <asiva@google.com>
- Initial and final marking no longer visit all of new-space, reducing the STW pause for major GC.
- A scavenge during concurrent marking must forward / filter objects in the marking worklist that are moved / collected, increasing the STW pause for minor GC.
- Unreachable intergenerational cycles and weak references are collected in the next mark-sweep instead of first requiring enough scavenges to promote the whole cycle or weak target into old-space.
- Artificial minor GCs are no longer needed to avoid memory leaks from back-to-back major GCs.
- reachabilityBarrier is now just a count of major GCs.
TEST=ci
Change-Id: I3668a2e56821f9eadf96e38c228dab27be656016
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/309826
Reviewed-by: Siva Annamalai <asiva@google.com>
Commit-Queue: Ryan Macnak <rmacnak@google.com>
Currently every invocation of a dart function will set and later on
reset the stack limits. Doing so requires acquiring locks.
Similarly because we have async ffi callbacks (another way to invoke
dart code) the logic was duplicated there.
Though the stack limit never changes for a given [OSThread]. Isolates
can run on different [OSThread]s throughtout its lifetime. But an
isolate always has to be entered on a native thread before it can
execute dart code.
=> We initialize the stack limit when we scheduling an isolate on a
thread and re-set it when unscheduling it.
=> That centralizes the place to one where we have to deal with stack
limits and avoids repeated acquiring of locks on each embedder dart
function invocation.
TEST=ci
Change-Id: Ia59ba8f92b93c58a990010ec75dfcd879aea2c43
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/311960
Commit-Queue: Martin Kustermann <kustermann@google.com>
Reviewed-by: Ryan Macnak <rmacnak@google.com>
The name is more accurately reflects what this stack trace
contains: despite what causal implies it does *not* actually
reflect the stack which initiated the async operation. Instead
it contains the chain of listeners which will run when
the async operation completes - its awaiters.
TEST=ci
Change-Id: Ie7309c9b1c39246e0fd4c14f7b9c515bcdfbbe10
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/311384
Reviewed-by: Derek Xu <derekx@google.com>
The round default value used in f7e26c50 means that we allocate
a new backing array shortly before hitting the cap, and once we
hit the cap the max load factor for the new backing array is only ~49%.
Instead, given a desired maximum, calculate an appropriate default
value that ensures the cap is only hit shortly before the array reaches
the max load factor that causes reallocation.
TEST=ci
Change-Id: I4c4fc3f5726aea3b260cf288cd00f886c2a1bd96
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/311280
Reviewed-by: Martin Kustermann <kustermann@google.com>
Commit-Queue: Tess Strickland <sstrickl@google.com>
This also lowers the threshold for converting from a linear cache
to a hash-based cache from 100 to 30 on non-IA32 architectures.
(IA32 still goes to runtime if there's a hash-based cache.)
For SubtypeTestCache benchmarks above this threshold, we see an
improvement varying from ~50-100% to ~700-800% on all architectures,
from lowest number of checks (50) to highest (1000).
For SubtypeTestCache benchmarks below this threshold, no major
changes are seen: generally <5% improvement or <5% regression per
check at most.
TEST=vm/cc/TTS
Change-Id: I83aa7c085ab5a411e944ec660d6b8eba7a788ee0
Cq-Include-Trybots: luci.dart.try:vm-aot-linux-debug-simriscv64-try,vm-aot-linux-debug-x64-try,vm-aot-linux-release-x64-try,vm-aot-linux-product-x64-try,vm-aot-linux-release-simarm64-try,vm-aot-linux-release-simarm_x64-try,vm-aot-linux-debug-x64c-try,vm-aot-tsan-linux-release-x64-try,vm-aot-mac-release-arm64-try,vm-kernel-precomp-linux-release-x64-try,vm-kernel-precomp-linux-debug-x64-try,vm-aot-dwarf-linux-product-x64-try,vm-linux-release-ia32-try,vm-linux-debug-x64c-try,vm-linux-debug-x64-try,vm-linux-debug-simriscv64-try,vm-linux-release-simarm64-try,vm-linux-release-simarm-try,vm-mac-release-arm64-try,vm-mac-release-x64-try,vm-tsan-linux-release-x64-try,vm-reload-rollback-linux-release-x64-try,vm-reload-linux-release-x64-try,vm-ffi-qemu-linux-release-arm-try
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/308941
Reviewed-by: Martin Kustermann <kustermann@google.com>
Commit-Queue: Tess Strickland <sstrickl@google.com>
In particular, avoid storing/comparing the destination type unless the
SubtypeTestCache was created for a dynamic AssertAssignable instruction.
TEST=vm/cc/STC_{Linear,Hash}Lookup, vm/cc/TTS
Bug: https://github.com/dart-lang/sdk/issues/48345
Change-Id: I279c69a668ce19785785cf71707acf951f2b2133
Cq-Include-Trybots: luci.dart.try:vm-aot-linux-debug-simriscv64-try,vm-aot-linux-debug-x64-try,vm-aot-linux-release-x64-try,vm-aot-linux-product-x64-try,vm-aot-linux-release-simarm64-try,vm-aot-linux-release-simarm_x64-try,vm-aot-linux-debug-x64c-try,vm-aot-tsan-linux-release-x64-try,vm-aot-mac-release-arm64-try,vm-kernel-precomp-linux-release-x64-try,vm-kernel-precomp-linux-debug-x64-try,vm-aot-dwarf-linux-product-x64-try,vm-linux-release-ia32-try,vm-linux-debug-x64c-try,vm-linux-debug-x64-try,vm-linux-debug-simriscv64-try,vm-linux-release-simarm64-try,vm-linux-release-simarm-try,vm-mac-release-arm64-try,vm-mac-release-x64-try,vm-tsan-linux-release-x64-try,vm-reload-rollback-linux-release-x64-try,vm-reload-linux-release-x64-try,vm-ffi-qemu-linux-release-arm-try
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/310340
Reviewed-by: Martin Kustermann <kustermann@google.com>
Reviewed-by: Alexander Markov <alexmarkov@google.com>
Commit-Queue: Tess Strickland <sstrickl@google.com>
We added a hash-based cache check in DRT_TypeCheck until the
SubtypeNTestCache stubs handle hash-based caches to avoid doing
the full runtime check when unnecessary, but did not do so for
DRT_InstanceOf. This fixes that until an upcoming CL that adds
support for hash-based caches in the SubtypeNTestCacheStubs lands.
TEST=Manual inspection of profiles before and after change. Not
adding an automated test as this change is a temporary bandaid
similar to the tested path in DRT_TypeCheck that will be removed soon.
Change-Id: I905a0fc6f501c68d66ae1ee848631c40cad04af2
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/310882
Reviewed-by: Martin Kustermann <kustermann@google.com>
Commit-Queue: Tess Strickland <sstrickl@google.com>
Previously, SlowTypeTestStub would use the incoming instance and
destination type to determine which `SubtypeNTestCacheStub` to call,
where `N` is the number of inputs checked for each entry. The code
for generating `InstanceOf` instructions performs similar work,
inlining some checks and then falling back to checking and updating
an appropriate `SubtypeNTestCacheStub` if the inlined checks fail.
Meanwhile, the runtime would use all available information to
create a new entry when needed, depending on the `SubtypeNTestCacheStub`
called at runtime to ignore the parts of the entry it didn't need.
To illustrate a situation where the runtime might use more
information to create an entry than needed by the stub, consider
the following class header:
```
class C<T> extends D
```
If no subtype of `D` requires knowing the specific type arguments for
an instance to determine whether that instance is an instance of `D`,
then the compiler can generate a `Subtype1TestCacheStub` call, where
only the class id of the instance is checked.
However, in the runtime, when updating the cache, the runtime will
create an entry that includes the specific type arguments used to
instantiate `C` when the initial cache miss occurred, as the runtime
does not know how many inputs the code that checks that SubtypeTestCache
uses. If the cache has grown to the point that it is hash-based, that
means the runtime uses the instance type arguments of the first cache
miss to determine the correct place to create a new entry.
When the stub only performs a linear search over the cache contents,
this works, as every entry is checked until a matching entry is found,
and thus the stub can ignore unneeded information. However, if a cache
is hash-based, this doesn't work: if the first miss was an instance
of `C<int>`, then the runtime includes all that information in a new
entry. This means the stub has to calculate the instance type
arguments, which it did not have to do before, and if a subsequent
check uses different instance type arguments (e.g., an instance of
`C<String>`), the result may still be a cache miss, and the runtime
adds another entry which has the same information as the previous
entry as far as the stub was originally concerned.
Thus, to limit itself to considering only the inputs used by the
appropriate `SubtypeNTestCacheStub`, the runtime needs to be told how
many inputs are used. This CL adds an additional field to
`SubtypeTestCache`s which contains the number of used inputs, and
changes the places where `SubtypeTestCache`s are allocated in the
flow graph compiler and the `DRT_TypeCheck` runtime entry to pass the
appropriate number of inputs. To ensure that the runtime and stubs
agree on how many inputs should be used, the `SlowTypeTestStub` now
uses the value of this field to determine which `SubtypeNTestCacheStub`
to call.
TEST=vm/cc/STC_{Linear,Hash}Lookup
Bug: https://github.com/dart-lang/sdk/issues/48345
Change-Id: Ie16f2d12c4c5eae856b668f28178818203662d1c
Cq-Include-Trybots: luci.dart.try:vm-aot-linux-debug-simriscv64-try,vm-aot-linux-debug-x64-try,vm-aot-linux-release-x64-try,vm-aot-linux-product-x64-try,vm-aot-linux-release-simarm64-try,vm-aot-linux-release-simarm_x64-try,vm-aot-linux-debug-x64c-try,vm-aot-mac-release-arm64-try,vm-kernel-precomp-linux-release-x64-try,vm-kernel-precomp-linux-debug-x64-try,vm-aot-dwarf-linux-product-x64-try,vm-linux-release-ia32-try,vm-linux-debug-x64c-try,vm-linux-debug-x64-try,vm-linux-debug-simriscv64-try,vm-linux-release-simarm64-try,vm-linux-release-simarm-try,vm-mac-release-arm64-try,vm-mac-release-x64-try,vm-tsan-linux-release-x64-try,vm-reload-rollback-linux-release-x64-try,vm-reload-linux-release-x64-try,vm-ffi-qemu-linux-release-arm-try
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/310180
Reviewed-by: Martin Kustermann <kustermann@google.com>
Commit-Queue: Tess Strickland <sstrickl@google.com>
Reviewed-by: Alexander Markov <alexmarkov@google.com>
Use `#if defined(TESTING)` for defining and using test-only flag
variables.
TTSTestCase had several flag fields where most of the flag fields should
not be set at the same time. Instead, use a separate `expected_result`
field that precisely states what type of result we expect from running
the test, and perform more precise checking of the expected state prior
to invocation and after invocation based on that.
Remove the old `TESTING_runtime_fail_on_existing_STC_entry` flag
variable and the abort it caused when set if an STC entry was found
when entering the runtime. Instead, replace it with two new flag
variables:
* `TESTING_runtime_entered_on_TTS_invocation`: whether the runtime
was entered for any reason during TTS invocation.
* `TESTING_found_hash_STC_entry`: whether we found an entry in a
hash-based STC, and so didn't perform runtime assignability checking.
Also modify most of the tests to check that either running the same
test case twice gives the same results, or in the case of
`kNewSTCEntry`, that a subsequent run with `kExistingSTCEntry` finds
the added entry.
TEST=vm/cc/TTS
Change-Id: Ibc3f49766494d0c61e2f78835637b9596b7f2db7
Cq-Include-Trybots: luci.dart.try:vm-linux-debug-x64-try,vm-linux-release-x64-try
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/309285
Commit-Queue: Tess Strickland <sstrickl@google.com>
Reviewed-by: Martin Kustermann <kustermann@google.com>
Also eliminates the need to store the delayed type arguments on the
stack during the search loop on ARM.
TEST=vm/cc/TTS
Change-Id: Ife01cc8d255724c272bc277eec397486cedc07a7
Cq-Include-Trybots: luci.dart.try:vm-aot-linux-debug-simriscv64-try,vm-aot-linux-debug-x64-try,vm-aot-linux-release-x64-try,vm-aot-linux-product-x64-try,vm-aot-linux-release-simarm64-try,vm-aot-linux-release-simarm_x64-try,vm-aot-linux-debug-x64c-try,vm-aot-tsan-linux-release-x64-try,vm-aot-mac-release-arm64-try,vm-kernel-precomp-linux-release-x64-try,vm-kernel-precomp-linux-debug-x64-try,vm-aot-dwarf-linux-product-x64-try,vm-linux-release-ia32-try,vm-linux-debug-x64c-try,vm-linux-debug-x64-try,vm-linux-debug-simriscv64-try,vm-linux-release-simarm64-try,vm-linux-release-simarm-try,vm-mac-release-arm64-try,vm-mac-release-x64-try,vm-tsan-linux-release-x64-try,vm-reload-rollback-linux-release-x64-try,vm-reload-linux-release-x64-try,vm-ffi-qemu-linux-release-arm-try
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/307681
Reviewed-by: Martin Kustermann <kustermann@google.com>
Commit-Queue: Tess Strickland <sstrickl@google.com>
Reviewed-by: Alexander Markov <alexmarkov@google.com>
We use specialized type testing stubs in `as` checks to quickly answer
whether the type of an object is assignable to a given type. However,
perfect specialization of the type testing stub is not possible in all
cases. Thus, we use SubtypeTestCaches (STCs) to cache the result of
checking assignability in the runtime to avoid repeated runtime calls
for the same check.
Previously, the STC was linear and capped at 100 entries. This cap was
for two reasons:
* STC are created per `as` check, not per checked type, so there could
be many STCs created for a given type if assignability to that type
is checked many times throughout the code.
* STC were strictly linear, and thus eventually the cost of
traversing the cache outweighs the cost of returning a false
negative that requires going to the runtime and doing the check there.
This CL provides a partial solution to the second reason by performing
similar work to that done for instantiation caches in TypeArguments
(4f925105), where the STC is linear up to a certain threshold, and then
converted to a hash-based STC. Like that CL, only the runtime currently
handles the lookup of cached results in hash-based STCs, with the
changes needed for the `SubtypeNTestCacheStub`s coming in a followup CL.
Since the stubs have not yet been adjusted, we keep the maximum size of
linear caches the same at 100, and allow hash-based STCs to grow to a
maximum size of 1000 since the STCs are still created per `as` check.
This means once the STC has grown over 100 entries, all STC checks
require a runtime call, but for checks cached in the STC, the answer is
found and provided without either doing the slower full runtime check
for assignability and without an attempt to respecialize the type
testing stub when applicable.
Our benchmarks generally see at most a ~2-5% performance regression for
cases where the STC remains linear (except for on the Raspberry Pi 4, where the hit ranges from ~25%-40%), but a performance improvement of
~500-1300% for cases where the STC is converted to a hash table.
TEST=vm/cc/TTS_STC_{Some,Many}Asserts
Change-Id: Ie8629708b5c12fa8464359a2f77c6da56ba46c0c
Cq-Include-Trybots: luci.dart.try:vm-aot-linux-debug-simriscv64-try,vm-aot-linux-debug-x64-try,vm-aot-linux-release-x64-try,vm-aot-linux-product-x64-try,vm-aot-linux-release-simarm64-try,vm-aot-linux-release-simarm_x64-try,vm-aot-linux-debug-x64c-try,vm-aot-mac-release-arm64-try,vm-kernel-precomp-linux-release-x64-try,vm-kernel-precomp-linux-debug-x64-try,vm-aot-dwarf-linux-product-x64-try,vm-linux-release-ia32-try,vm-linux-debug-x64c-try,vm-linux-debug-x64-try,vm-linux-debug-simriscv64-try,vm-linux-release-simarm64-try,vm-linux-release-simarm-try,vm-mac-release-arm64-try,vm-mac-release-x64-try,vm-tsan-linux-release-x64-try,vm-reload-rollback-linux-release-x64-try,vm-reload-linux-release-x64-try,vm-ffi-qemu-linux-release-arm-try
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/284682
Reviewed-by: Alexander Markov <alexmarkov@google.com>
Reviewed-by: Martin Kustermann <kustermann@google.com>
Make the free and allocated lists simply linked lists of `Metadata*`
(instead of `Metadata*` pointing to Trampoline, which - via some
logic - can be translated into next `Metadata*`).
Make the layout of the virtual address space mapping and
offsets `constexpr` functions instead of computing them at
runtime & caching in fields.
Use `uword` to represent `Trampoline` entrypoint (as we generally
use `uword` for `Code.EntryPoint()` / ...)
TEST=ci
Change-Id: If4ffa11712acc46c9295b609caff7576d2354fe4
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/305983
Commit-Queue: Martin Kustermann <kustermann@google.com>
Reviewed-by: Daco Harkes <dacoharkes@google.com>
This is the same metadata system that will support async callbacks. The
main thing this does is take the trampolines that used to be JIT only,
and use them in AOT too. This is partly a safety thing (there's some
extra checks that used to be skipped on AOT), but mostly just so that
the metadata system is unified between the sync and async callbacks.
More details about the design:
https://docs.google.com/document/d/1QDjyY_6wOTOgURwpeYMKU9qEz0gKxx2MUrdruC6Kp6c/edit?usp=sharing
I split this off the async CL. Some of the comments refer to async
stuff that doesn't exist yet, but it's coming immediately after this so
I didn't update them.
Change-Id: Icd5e86934ee9ae34c2c0e2ed2bbd1b928a7184ac
TEST=ffi_callback_metadata_test.cc and CI
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/302903
Reviewed-by: Ryan Macnak <rmacnak@google.com>
Reviewed-by: Martin Kustermann <kustermann@google.com>
Commit-Queue: Liam Appelbe <liama@google.com>
Reviewed-by: Daco Harkes <dacoharkes@google.com>
Reviewed-by: Alexander Markov <alexmarkov@google.com>
The ffi-callback related information on the [Thread] object is metadata
corresponding to ffi-callback-trampoline [Function] objects. There is
nothing thread or isolate specific about it.
Moving it away from [Thread] is needed because an [Isolate] can
have different [Thread] objects across its lifetime (see [0]): When
the stack of an isolate is empty, we reserve now the right to
re-cycle the [Thread]. If the isolate later runs again, it may
get a new [Thread] object.
This CL moves this information from [Thread] to the [ObjectStore]. In
addition we make the compiler be responsible for populating this
metadata - instead of doing this per-call site of
`Pointer.fromFunction()`. It will be preserved across snapshot writing
& snapshot reading (for AppJIT as well as AppAOT).
Similarly the JIT trampolines that are on Isolate aren't isolate
specific and can go to [IsolateGroup]. This simplifies doing the above
as the compiler can allocate those as well.
The effect is that [Thread] object gets smaller, GC doesn't have to
visit the 2 slots per-thread. It comes at expense of 2 more loads
when invoking the callback.
[0] https://dart-review.googlesource.com/c/sdk/+/297920
TEST=Regression test is vm/ffi{,_2}/invoke_callback_after_suspension_test
Change-Id: Ifde46a9f6e79819b5c0e359c3d3998d1d93b9b1e
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/303700
Reviewed-by: Daco Harkes <dacoharkes@google.com>
Reviewed-by: Liam Appelbe <liama@google.com>
Reviewed-by: Ryan Macnak <rmacnak@google.com>
Commit-Queue: Martin Kustermann <kustermann@google.com>
This field was supposed[1] to replace Stack.asyncCausalFrames but IDEs
and other service protocol users never adopted it.
The mechanism for collecting awaiterFrames and asyncCausalFrames
was originally considerably different, but we have since unified
both[2] after we switched to lazy async stack traces[3].
Today the only differences between the two are:
- asyncCausalFrames represens async gaps explicitly, while
awaiterFrames does not;
- when asynchronous function is running its synchronous part
before the first suspension awaiterFrames will still represent
the corresponding frame as an asynchronous activation, while
asyncCausalFrames will treat the same frame as regular frame.
Consequently having this field does not add any value.
This CL removes it from the response and updates all tests to
test against asyncCausalFrames.
[1]: https://codereview.chromium.org/2692803006/
[2]: https://dart-review.googlesource.com/c/sdk/+/167561
[3]: https://github.com/dart-lang/sdk/issues/37668
Tested: pkg/vm_service, service, service_2
Bug: https://github.com/dart-lang/sdk/issues/51763
Change-Id: If2b035a8840047423b8ed8ce3b5d81155e9f38d0
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/301062
Commit-Queue: Slava Egorov <vegorov@google.com>
Reviewed-by: Ben Konyi <bkonyi@google.com>
output unless there is an assertion or Fatal error.
(This is needed to ensure that we are able to run the flutter tests with
all these verifications turned on and not disturbing the output from the
tests with these extraneous trace messages)
TEST=ci
Change-Id: Ifb7211e340f1cf29f6d2bff412eb4107f9cb64a2
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/299440
Commit-Queue: Siva Annamalai <asiva@google.com>
Reviewed-by: Ryan Macnak <rmacnak@google.com>
TypeRef type wraps around another type and it was used to represent
a graph of recursive types. After [0], the only use of TypeRef is
for TypeParameter.bound which may indirectly reference the same
TypeParameter.
This change replaces TypeParameter.bound with TypeParameter.owner and
removes TypeRef entirely. Various parts of the VM no longer need to
handle and support TypeRefs.
TypeParameter.owner can reference a FunctionType, Class,
or, as an optimization, it can be set to null in order to share
class type parameters among different classes.
With the exception of the 'TypeParameter.owner' back pointer,
VM types are now not recursive and can be visited without
additional tracking.
Caveats:
* Generic FunctionType cannot be cloned in a shallow way:
when copying a FunctionType, type parameters should be cloned too
and their owners should be updated. For that reason,
a mapping between 'from' and 'to' function types
(FunctionTypeMapping) is maintained during type transformations
such as InstantiateFrom.
FunctionType::Clone is used instead of Object::Clone where
appropriate.
* When testing types for subtyping and equivalence, mapping
between function types is passed to make sure
type parameters belong to the equivalent function types.
* IL serializer needs to serialize function types as a whole before
serializing any types potentially pointing into the middle of a
function type (such as return type 'List<Y0>' pointing into
the middle of a function type 'List<Y0> Function<Y0>()').
[0] https://dart-review.googlesource.com/c/sdk/+/296300
TEST=ci
Change-Id: I67c2fd0117c6183a45e183919a7847fd1af70b3e
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/294165
Reviewed-by: Ryan Macnak <rmacnak@google.com>
Commit-Queue: Alexander Markov <alexmarkov@google.com>
Some cleanups factored out of a larger CL (which refactors enter/exit of threads):
* remove unused `#include "vm/thread_registry.h"`
* remove unused/unnecessary fields from [Thread] object
* rename IsMutator() -> IsDartMutator()
* make tests using setjmp() drain the sticky error
=> to ensure there's no sticky error on isolate shutdown
TEST=ci
Change-Id: I53935e8bd0628ab3768627d6d5e01c3f0e3a57ad
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/296582
Commit-Queue: Martin Kustermann <kustermann@google.com>
Reviewed-by: Alexander Aprelev <aam@google.com>
The representation of record shape in record instances and record
types is changed from a pair
int num_fields
Array field_names
to a single integer - packed bitfield
int num_fields(16)
int field_names_index(kSmiBits-16)
where field names index is an index in the array available from
ObjectStore.
With the new representation of record shapes:
1) Size of record instances is reduced.
2) Number of comparisons for a shape test reduced from 2 to 1
(shape test is used during type checks).
3) A few operations removed from Record.hashCode.
4) Type testing stubs (TTS) are now supported for record types with
named fields. Previously it was not possible to check shape of records
with named fields in TTS as TTS cannot access object pool and cannot
load field names array).
TEST=existing
Issue: https://github.com/dart-lang/sdk/issues/49719
Change-Id: I7cdcbb53938aba5d561cd24dc99530395dbbea7e
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/276201
Reviewed-by: Ryan Macnak <rmacnak@google.com>
Commit-Queue: Alexander Markov <alexmarkov@google.com>
'is' tests against record types are split into series of 'is' tests of
record fields. This is more efficient as record instances cannot be
added to subtype test caches (because type of a record instance
depends on types of its fields).
This change also adds canonicalization and constant evaluation of
LoadFieldInstr for record fields.
Performance on a trivial micro-benchmark (on x64):
JIT (RunTime): 4519104.5 -> 20031.5
AOT (RunTime): 4352583.0 -> 26281.6
TEST=ci
Issue: https://github.com/dart-lang/sdk/issues/49719
Change-Id: I2ed464cd3b31f365b17805f4e7debe1d6d1051fa
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/268080
Reviewed-by: Ryan Macnak <rmacnak@google.com>
Reviewed-by: Slava Egorov <vegorov@google.com>
Commit-Queue: Alexander Markov <alexmarkov@google.com>
This change introduces specialized stubs and IL instruction for
allocating records with 2 or 3 fields. This makes allocation of
small records slightly faster compared to a construction of similar
class instances and makes code size of record allocation smaller.
Benchmark:
MultipleReturns.NotInlined.Record(RunTime) 77150 -> 66222
MultipleReturns.NotInlined.RecordNamed(RunTime) 78073 -> 67044
MultipleReturns.Forwarded.Record(RunTime) 97130 -> 77635
MultipleReturns.Forwarded.RecordNamed(RunTime) 96495 -> 77904
TEST=ci
Issue: https://github.com/dart-lang/sdk/issues/49719
Change-Id: I8ed7add06b39ba79dfd78bbe2afaefe606cc505b
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/266420
Commit-Queue: Alexander Markov <alexmarkov@google.com>
Reviewed-by: Slava Egorov <vegorov@google.com>
Ryan Macnak