This reverts commit f81a402aa1.
Reason for revert: golem benchmarks are failing to run
TEST=ci
Original change's description:
> "[SDK/VM] - Rename dart_precompiled_runtime to dartaotruntime, ensures we have a uniform name for the executable between the build directories and the SDK directory"
>
> Fixed golem breakage by temporarily copying dartaotruntime to dart_precompiled_runtime
>
> This reverts commit 75e6a748f7.
>
> TEST=ci
>
> Original change's description:
> > Revert "[SDK/VM] - Rename dart_precompiled_runtime to dartaotruntime, ensures we have a uniform name for the executable between the build directories and the SDK directory"
> >
> > This reverts commit 1b331d05c2.
> >
> > Reason for revert: golem builds are failing
> >
> > Original change's description:
> > > [SDK/VM] - Rename dart_precompiled_runtime to dartaotruntime, ensures we have a uniform name for the executable between the build directories and the SDK directory
> > >
> > > TEST=ci
> > >
>
> Change-Id: Id0f383eabb496c06c0acebc639c8e3b056ba82d0
> Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/393781
> Commit-Queue: Siva Annamalai <asiva@google.com>
> Reviewed-by: Ryan Macnak <rmacnak@google.com>
Change-Id: Iec494940412aa31dbefdc5280e35ae99e8cecb26
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/393764
Reviewed-by: Siva Annamalai <asiva@google.com>
Reviewed-by: Liam Appelbe <liama@google.com>
Bot-Commit: Rubber Stamper <rubber-stamper@appspot.gserviceaccount.com>
Commit-Queue: Siva Annamalai <asiva@google.com>
Fixed golem breakage by temporarily copying dartaotruntime to dart_precompiled_runtime
This reverts commit 75e6a748f7.
TEST=ci
Original change's description:
> Revert "[SDK/VM] - Rename dart_precompiled_runtime to dartaotruntime, ensures we have a uniform name for the executable between the build directories and the SDK directory"
>
> This reverts commit 1b331d05c2.
>
> Reason for revert: golem builds are failing
>
> Original change's description:
> > [SDK/VM] - Rename dart_precompiled_runtime to dartaotruntime, ensures we have a uniform name for the executable between the build directories and the SDK directory
> >
> > TEST=ci
> >
Change-Id: Id0f383eabb496c06c0acebc639c8e3b056ba82d0
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/393781
Commit-Queue: Siva Annamalai <asiva@google.com>
Reviewed-by: Ryan Macnak <rmacnak@google.com>
This CL adds basic infrastructure and tooling to perform allocation
profiling using uprobes.
See runtime/tools/profiling/README.md for more details.
TEST=manually tested, requires root access
Cq-Include-Trybots: luci.dart.try:vm-aot-mac-release-arm64-try
Change-Id: Id68d181740dbf227a12d8cdba84b11a9518e75a7
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/382405
Commit-Queue: Slava Egorov <vegorov@google.com>
Reviewed-by: Martin Kustermann <kustermann@google.com>
`Console.cursorPosition` can return null when you are typing too
fast because it is implemented by sending an escape code to
the terminal and reading what it responds. This response is
naturally intermingled with normal input and the implementation
does nothing to guard against that.
We do not really need to know the cursor position though
we are simply interested in keeping cursor on the same row.
But this can be achieved by sending _horizontal position absolute_
which only changes column rather than _cursor position_
which sets both row and column.
R=kustermann@google.com
TEST=manually tested
Change-Id: Ie1a064e37b90bc4529ac4f5c1259642ad5680ca6
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/388340
Commit-Queue: Slava Egorov <vegorov@google.com>
Reviewed-by: Martin Kustermann <kustermann@google.com>
Specifically apply prefer-inline to:
* All generated accessors. In general we almost always seem to
select them for inlining through small-accessor heuristic,
but sometimes it does not hit. Marking with pragma ensures
that we are not dependent on specific heuristics to produce
good code.
* Various helpers involved in creating struct instances from
typed data. These we *not* hit by normal heuristics and
consequently not inlined prior to this change. Inlining them
opens possibilities to fully eliminate allocations associated
with `Struct.create(...)` if Struct does not escape.
* Array creation helpers and accessors. They were also not inlined
which meant we were producing unnecessary temporary objects
when reading from inline arrays.
TEST=/vm/dart/ffi_structs_optimizations_il_test
CoreLibraryReviewExempt: no API or behavioral changes
Cq-Include-Trybots: luci.dart.try:vm-aot-linux-product-x64-try,vm-aot-linux-release-x64-try
Change-Id: I1650023200446bd5289118777365627ca19d3e5b
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/384261
Commit-Queue: Slava Egorov <vegorov@google.com>
Reviewed-by: Daco Harkes <dacoharkes@google.com>
No client of the VM uses this flag, only tests, and this flag was always
set to false in AOT mode. Thus, remove uses of this flag and instead
always lazily create dispatchers as needed when resolving method names
in JIT mode.
Remove the implicit value of `allow_add` for some Resolver
static methods. For callers that previously depended on the implicit
`true` value (which includes the AOT precompilier), pass `true` for
uses in the compiler and pass `!FLAG_precompiled_mode` for uses in the
runtime. Assert that `allow_add` is false when these methods are invoked
from the precompiled runtime.
Remove Resolver static methods that are no longer used.
TEST=ci
Change-Id: Ib6a7354f7a859e86743c381513a4129c14895753
Cq-Include-Trybots: luci.dart.try:vm-linux-debug-x64-try,vm-linux-release-x64-try,vm-aot-linux-debug-x64-try,vm-aot-linux-release-x64-try,vm-aot-mac-release-arm64-try,vm-mac-debug-arm64-try,vm-mac-release-arm64-try
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/366668
Reviewed-by: Ryan Macnak <rmacnak@google.com>
Reviewed-by: Martin Kustermann <kustermann@google.com>
Commit-Queue: Tess Strickland <sstrickl@google.com>
Print the number of pending task every 5 minutes to avoid getting killed by the 20 minute I/O timeout. We'll still get killed if we exceed the 50 minute step timeout.
Change-Id: I0bc12c5b78176303432d35fc6107b669ed7c6a82
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/365282
Commit-Queue: Ryan Macnak <rmacnak@google.com>
Reviewed-by: Alexander Markov <alexmarkov@google.com>
During FFI leaf calls, the Dart GC will not run. This means that we
can pass pointers into `TypedData` to FFI calls that take `Pointer`
arguments.
After this CL, we have three types of arguments that can flow into
`Pointer` argument in an FFI call:
* `Pointer`.
* `TypedData`: Any typed data including views.
* `_Compound`: A TypedData/Pointer and an offset in bytes.
The is only possible for `@Native external` functions, `asFunction`
does not support passing in `TypedData`. (See related GitHub issues
for discussion. TLDR: FFIgen should generate bindings without config.)
`.address` expressions on `TypedData` and `Array` elements do _not_
introduce bounds checks, even though `TypedData` and `Array` have
bounds information. E.g. `ffiNative(Uint8List(10)[20].address)` does
not throw.
Implementation details:
The CFE analyzes call-sites to `@Native external` functions. If the
arguments are `.address` expressions, it transforms the call site to
pass the compound or `TypedData`. If an additional offset needs to be
applied, the CFE constructs a new `_Compound` with the correct offset
in bytes.
The CFE then also creates a new `@Native external` function which have
`TypedData`s and `_Compound`s parameters. To avoid name clashes, these
functions are postfixed with `#` and `P`, `T`, or `C` for each Pointer
parameter.
TEST=pkg/vm/testcases/transformations/ffi/address_of_*
In the VM, `TypedData` arguments are passed as tagged values, and the
address is loaded inside the `FfiCallInstr`. `_Compound` arguments
turn into two IL definitions, one for the `TypedDataBase` (tagged),
and one for the offset in bytes (unboxed). The address is then loaded
inside the `FfiCallInstr` and the offset in bytes is applied.
Adding the offset in bytes required an extra temp register for ia32.
Also, it uncovered that the temp register in arm32 was conflicting
with the argument registers. However, TMP should suffice instead.
TEST=tests/ffi/address_of_array_generated_test.dart
TEST=tests/ffi/address_of_struct_generated_test.dart
TEST=tests/ffi/address_of_typeddata_generated_test.dart
Closes: https://github.com/dart-lang/sdk/issues/44589
Closes: https://github.com/dart-lang/sdk/issues/54771
CoreLibraryReviewExempt: VM only, unsupported in dart2wasm
Change-Id: I01fb428cfd6f9096a34689c2819c124a8003cb6b
Cq-Include-Trybots: dart/try:vm-aot-android-release-arm64c-try,vm-aot-android-release-arm_x64-try,vm-aot-linux-debug-x64-try,vm-aot-linux-debug-x64c-try,vm-aot-mac-release-arm64-try,vm-aot-mac-release-x64-try,vm-aot-obfuscate-linux-release-x64-try,vm-aot-optimization-level-linux-release-x64-try,vm-aot-win-debug-arm64-try,vm-aot-win-debug-x64c-try,vm-aot-win-release-x64-try,vm-appjit-linux-debug-x64-try,vm-asan-linux-release-x64-try,vm-checked-mac-release-arm64-try,vm-eager-optimization-linux-release-ia32-try,vm-eager-optimization-linux-release-x64-try,vm-ffi-android-debug-arm64c-try,vm-ffi-qemu-linux-release-arm-try,vm-ffi-qemu-linux-release-riscv64-try,vm-linux-debug-ia32-try,vm-linux-debug-x64-try,vm-linux-debug-x64c-try,vm-mac-debug-arm64-try,vm-mac-debug-x64-try,vm-msan-linux-release-x64-try,vm-reload-linux-debug-x64-try,vm-reload-rollback-linux-debug-x64-try,vm-ubsan-linux-release-x64-try,vm-win-debug-arm64-try,vm-win-debug-x64-try,vm-win-debug-x64c-try,vm-win-release-ia32-try
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/360882
Reviewed-by: Jens Johansen <jensj@google.com>
Reviewed-by: Brian Wilkerson <brianwilkerson@google.com>
Reviewed-by: Martin Kustermann <kustermann@google.com>
Commit-Queue: Daco Harkes <dacoharkes@google.com>
Reviewed-by: Tess Strickland <sstrickl@google.com>
Reviewed-by: Lasse Nielsen <lrn@google.com>
The clang-tidy check in PRESUBMIT.py is only run on Linux x64 hosts,
so CLs uploaded from other machines can introduce clang-tidy lint
violations.
Running `git cl presubmit --files="runtime/*.cc;runtime/*.h"` revealed
one violation in `runtime/vm/compiler/ffi/unit_test_custom_zone.cc`.
This file is not part of the normal build only of cc unit tests, so
it is built with different ifdefs which are not supplied from the
`run_clang_tidy.dart` script, so skip it.
Change-Id: Iddf0256625a092038f4a908ac5311a2dc9a8d62b
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/363561
Auto-Submit: Daco Harkes <dacoharkes@google.com>
Commit-Queue: Martin Kustermann <kustermann@google.com>
Reviewed-by: Martin Kustermann <kustermann@google.com>
This CL changes compounds (structs, unions, and arrays) to be backed
by a TypedDataBase and an int offset.
Before this CL, the compounds where only backed by a TypedDataBase.
This leads to the following issues:
1. Access to nested structs required code for allocating new typed data
views or pointers, which the optimizer then had to prevent from
being allocated after inlining.
2. Runtime branching on whether the TypedDataBase was a Pointer or
TypedData increased code size and prevented inlining.
https://github.com/dart-lang/sdk/issues/54892
This could not be properly optimized if in AOT both typed
data and pointer were flowing into the same compound.
3. Constructing TypedData views required calculating the length of the
view.
After this CL, accessing nested compounds will lead to accesses on the
original TypedDataBase with an extra offset.
This removes the polymorphism on TypedData vs Pointer, because the
final int/float/Pointer accesses in nested compounds operate on
TypedDataBase.
Also, it simplifies creating an `offsetBy` accessor, because it will
no longer have to be polymorphic in typed data vs pointer, nor will it
have to calculate the length of the field.
Implementation details:
* The changes in the CFE and patch files are straightforward.
* VM: Struct-by-value returns (or callback params) are initialized
with an offsetInBytes of 0.
* VM: Struct-by-value arguments (and callback return) need to read out
the offsetInBytes. Before this CL we were passing in the TypedData
as tagged value. With this CL we are passing the TypedData as tagged
value and the offset as unboxed int, from 1 IL input to 2 IL
inputs. (The alternative would have been to take the compound as
a tagged value, but that would have prevented optimizations from not
allocating the compound object in the optimizer.
The FfiCallInstr is updated to have two definitions for the case
where we were passing in the TypedData previously.
The NativeReturnInstr is refactored to be able to take two inputs
instead of 1. (Note that we don't have VariadicInstr only
VariadicDefinition in the code base. So the instruction is _not_
implemented as variadic, rather as having a fixed length of 2.)
* dart2wasm does no longer support nested compounds due to the
compound implementation only storing a pointer address.
https://github.com/dart-lang/sdk/issues/55083
Intending to land this after
https://dart-review.googlesource.com/c/sdk/+/353101.
TEST=test/ffi
CoreLibraryReviewExempt: VM and WASM-only implementation change.
Closes: https://github.com/dart-lang/sdk/issues/54892
Bug: https://github.com/dart-lang/sdk/issues/44589
Change-Id: I8749e21094bf8fa2d5ff1e48b6b002c375232eb5
Cq-Include-Trybots: dart-internal/g3.dart-internal.try:g3-cbuild-try
Cq-Include-Trybots: dart/try:vm-aot-android-release-arm64c-try,vm-aot-android-release-arm_x64-try,vm-aot-linux-debug-x64-try,vm-aot-linux-debug-x64c-try,vm-aot-mac-release-arm64-try,vm-aot-mac-release-x64-try,vm-aot-obfuscate-linux-release-x64-try,vm-aot-optimization-level-linux-release-x64-try,vm-aot-win-debug-arm64-try,vm-aot-win-debug-x64-try,vm-aot-win-debug-x64c-try,vm-appjit-linux-debug-x64-try,vm-asan-linux-release-x64-try,vm-checked-mac-release-arm64-try,vm-eager-optimization-linux-release-ia32-try,vm-eager-optimization-linux-release-x64-try,vm-ffi-android-debug-arm-try,vm-ffi-android-debug-arm64c-try,vm-ffi-qemu-linux-release-arm-try,vm-ffi-qemu-linux-release-riscv64-try,vm-fuchsia-release-x64-try,vm-linux-debug-ia32-try,vm-linux-debug-x64-try,vm-linux-debug-x64c-try,vm-mac-debug-arm64-try,vm-mac-debug-x64-try,vm-msan-linux-release-x64-try,vm-reload-linux-debug-x64-try,vm-reload-rollback-linux-debug-x64-try,vm-ubsan-linux-release-x64-try,vm-win-debug-arm64-try,vm-win-debug-x64-try,vm-win-debug-x64c-try,vm-win-release-ia32-try
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/354226
Reviewed-by: Tess Strickland <sstrickl@google.com>
FFI loads and stores via structs can have a TypedData as receiver,
so this CL updates those loads to `kMayBeInnerPointer`.
This CL adds an IL test to verify that for `Pointer` loads the untagged
value is treated correctly as `kCannotBeInnerPointer`.
(And adds some prefer-inline pragmas to make some common operations
be inlined to avoid allocating `Pointer` objects.)
This CL updates the load in the FFI closures to use a load-field.
This can also potentially enable not allocating a pointer object when
this closure is inlined.
TEST=tests/ffi/unwrap_typeddata_generated_test.dart
TEST=tests/ffi
CoreLibraryReviewExempt: Only adding some pragmas.
Change-Id: If687e54c676f275cc849b3fed526a13766ab331a
Cq-Include-Trybots: luci.dart.try:vm-aot-android-release-arm64c-try,vm-aot-android-release-arm_x64-try,vm-aot-linux-debug-x64-try,vm-aot-linux-debug-x64c-try,vm-aot-mac-release-arm64-try,vm-aot-mac-release-x64-try,vm-aot-obfuscate-linux-release-x64-try,vm-aot-optimization-level-linux-release-x64-try,vm-aot-win-debug-arm64-try,vm-aot-win-debug-x64-try,vm-appjit-linux-debug-x64-try,vm-asan-linux-release-x64-try,vm-checked-mac-release-arm64-try,vm-eager-optimization-linux-release-ia32-try,vm-eager-optimization-linux-release-x64-try,vm-ffi-android-debug-arm-try,vm-ffi-android-debug-arm64c-try,vm-ffi-qemu-linux-release-arm-try,vm-ffi-qemu-linux-release-riscv64-try,vm-fuchsia-release-x64-try,vm-kernel-linux-debug-x64-try,vm-kernel-precomp-linux-release-x64-try,vm-linux-debug-ia32-try,vm-linux-debug-x64-try,vm-linux-debug-x64c-try,vm-mac-debug-arm64-try,vm-mac-debug-x64-try,vm-msan-linux-release-x64-try,vm-reload-linux-debug-x64-try,vm-reload-rollback-linux-debug-x64-try,vm-ubsan-linux-release-x64-try,vm-win-debug-arm64-try,vm-win-debug-x64-try,vm-win-release-ia32-try
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/349241
Reviewed-by: Alexander Markov <alexmarkov@google.com>
Upstreamed changes from cl/579854752.
The cross-compiler checks some things that are check on Windows.
* Correct capitalization of filenames in includes.
* Field initialization order in constructors.
In the cross compilation process, some binaries are run on the host.
This unveiled missing `UnwindingRecords::GenerateRecordsInto`.
Finally, when emitting blob data with a symbol in assembly, the
`.type` directive is not supported but the format should still be
unix style assemble. So this CL introduces "win_gnu". This tool is
directly invoked from the build-rules in cl/579854752, and will not be
used until we address https://github.com/dart-lang/sdk/issues/28617.
TEST=windows bots
Change-Id: I94256589e8c231b45b8e14a63727c782416c2e98
Cq-Include-Trybots: luci.dart.try:vm-aot-win-debug-arm64-try,vm-aot-win-debug-x64c-try,pkg-win-release-try,pkg-win-release-arm64-try,vm-win-debug-arm64-try,vm-win-debug-x64c-try,vm-win-debug-x64-try,vm-msvc-windows-try
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/335520
Commit-Queue: Daco Harkes <dacoharkes@google.com>
Reviewed-by: Martin Kustermann <kustermann@google.com>
This can be controlled via codesigning_identity GN arg.
For example, setting codesigning_identity="-"
would produce ad-hoc signed binaries.
This CL also includes changes in vm/cc tests which are needed
for tests to be green when running with hardened runtime.
Issue https://github.com/dart-lang/sdk/issues/53928
Tested: enabled ad-hoc signing and tested on bots.
Cq-Include-Trybots: luci.dart.try:vm-mac-debug-arm64-try,vm-mac-release-arm64-try,vm-mac-release-x64-try,vm-mac-debug-x64-try
Change-Id: I3c3a6265c62b2904e43a326b7d8223bcfd393577
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/333401
Reviewed-by: Ryan Macnak <rmacnak@google.com>
Commit-Queue: Slava Egorov <vegorov@google.com>
This is a reland of commit 06d7a2352e
This version fixes an issue when a phi node has multiple inputs with
different unboxed integer representations. The original CL made a
change where only the representations were considered, not the range
of values for the Phi calculated by range analysis. The reland goes
back to the old behavior for this case.
Also fixes the new tests on 32-bit architectures.
Original change's description:
> [vm/compiler] Change MemoryCopy to also take untagged addresses.
>
> This CL adds the ability to pass the payload address of the source
> and destination directly to the MemoryCopy instruction as an untagged
> value.
>
> The new translation of the _TypedListBase._memMoveN methods use the new
> MemoryCopy constructor, retrieving the untagged value of the data field
> of both the source and destination. This way, if inlining exposes the
> allocation of the object from which the data field is being retrieved,
> then allocation sinking can remove the intermediate allocation if there
> are no escaping uses of the object.
>
> Since Pointer.asTypedList allocates such ExternalTypedData objects,
> this CL makes that method inlined if at all possible, which removes
> the intermediate allocation if the only use of the TypedData object
> is to call setRange for memory copying purposes.
>
> This CL also separates unboxed native slots into two groups: those
> that contain untagged addresses and those that do not. The former
> group now have the kUntagged representation, which mimics the old
> use of LoadUntagged for the PointerBase data field and also ensures
> that any arithmetic operations on untagged addresses must first be
> explicitly converted to an unboxed integer and then explicitly converted
> back to untagged before being stored in a slot that contains untagged
> addresses.
>
> When a unboxed native slot that contains untagged addresses is defined,
> the definition also includes a boolean which represents whether
> addresses that may be moved by the GC can be stored in this slot or not.
> The redundancy eliminator uses this to decide whether it is safe to
> eliminate a duplicate load, replace a load with the value originally
> stored in the slot, or lift a load out of a loop.
>
> In particular, the PointerBase data field may contain GC-moveable
> addresses, but only for internal TypedData objects and views, not
> for external TypedData objects or Pointers. To allow load optimizations
> involving the latter, the LoadField and StoreField instructions now
> take boolean flags for whether loads or stores from the slot are
> guaranteed to not be GC-moveable, to override the information from
> the slot argument.
>
> Notable benchmark changes on x64 (similar for other archs unless noted):
>
> JIT:
> * FfiMemory.PointerPointer: 250.7%
> * FfiStructCopy.Copy1Bytes: -26.73% (only x64)
> * FfiStructCopy.Copy32Bytes: -25.18% (only x64)
> * MemoryCopy.64.setRange.Pointer.Uint8: 19.36%
> * MemoryCopy.64.setRange.Pointer.Double: 18.96%
> * MemoryCopy.8.setRange.Pointer.Double: 17.59%
> * MemoryCopy.8.setRange.Pointer.Uint8: 19.46%
>
> AOT:
> * FfiMemory.PointerPointer: 323.5%
> * FfiStruct.FieldLoadStore: 483.3%
> * FileIO_readwrite_64kb: 15.39%
> * FileIO_readwrite_512kb (Intel Xeon): 46.22%
> * MemoryCopy.512.setRange.Pointer.Uint8: 35.20%
> * MemoryCopy.64.setRange.Pointer.Uint8: 55.40%
> * MemoryCopy.512.setRange.Pointer.Double: 29.45%
> * MemoryCopy.64.setRange.Pointer.Double: 60.37%
> * MemoryCopy.8.setRange.Pointer.Double: 59.54%
> * MemoryCopy.8.setRange.Pointer.Uint8: 55.40%
> * FfiStructCopy.Copy32Bytes: 398.3%
> * FfiStructCopy.Copy1Bytes: 1233%
>
> TEST=vm/dart/address_local_pointer, vm/dart/pointer_as_typed_list
>
> Issue: https://github.com/dart-lang/sdk/issues/42072
> Fixes: https://github.com/dart-lang/sdk/issues/53124
>
> Cq-Include-Trybots: luci.dart.try:vm-ffi-qemu-linux-release-arm-try,vm-eager-optimization-linux-release-x64-try,vm-linux-release-x64-try,vm-linux-debug-x64-try,vm-aot-linux-release-x64-try,vm-aot-linux-debug-x64-try
> Change-Id: I563e0bfac5b1ac6cf1111649934067c12891b631
> Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/324820
> Reviewed-by: Alexander Markov <alexmarkov@google.com>
> Commit-Queue: Tess Strickland <sstrickl@google.com>
> Reviewed-by: Martin Kustermann <kustermann@google.com>
TEST=vm/dart/address_local_pointer, vm/dart/pointer_as_typed_list
Issue: https://github.com/dart-lang/sdk/issues/42072
Fixes: https://github.com/dart-lang/sdk/issues/53124
Change-Id: Iabb0e910f12636d0ff51e711c8c9c98ad40e5811
Cq-Include-Trybots: luci.dart.try:vm-ffi-qemu-linux-release-arm-try,vm-eager-optimization-linux-release-x64-try,vm-linux-release-x64-try,vm-linux-debug-x64-try,vm-aot-linux-release-x64-try,vm-aot-linux-debug-x64-try,vm-aot-linux-release-simarm_x64-try,vm-aot-linux-debug-simarm_x64-try
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/330600
Reviewed-by: Alexander Markov <alexmarkov@google.com>
Commit-Queue: Tess Strickland <sstrickl@google.com>
Reviewed-by: Martin Kustermann <kustermann@google.com>
This reverts commit 06d7a2352e.
Reason for revert: everything crashes on vm-aot-linux-debug-simarm_x64
Original change's description:
> [vm/compiler] Change MemoryCopy to also take untagged addresses.
>
> This CL adds the ability to pass the payload address of the source
> and destination directly to the MemoryCopy instruction as an untagged
> value.
>
> The new translation of the _TypedListBase._memMoveN methods use the new
> MemoryCopy constructor, retrieving the untagged value of the data field
> of both the source and destination. This way, if inlining exposes the
> allocation of the object from which the data field is being retrieved,
> then allocation sinking can remove the intermediate allocation if there
> are no escaping uses of the object.
>
> Since Pointer.asTypedList allocates such ExternalTypedData objects,
> this CL makes that method inlined if at all possible, which removes
> the intermediate allocation if the only use of the TypedData object
> is to call setRange for memory copying purposes.
>
> This CL also separates unboxed native slots into two groups: those
> that contain untagged addresses and those that do not. The former
> group now have the kUntagged representation, which mimics the old
> use of LoadUntagged for the PointerBase data field and also ensures
> that any arithmetic operations on untagged addresses must first be
> explicitly converted to an unboxed integer and then explicitly converted
> back to untagged before being stored in a slot that contains untagged
> addresses.
>
> When a unboxed native slot that contains untagged addresses is defined,
> the definition also includes a boolean which represents whether
> addresses that may be moved by the GC can be stored in this slot or not.
> The redundancy eliminator uses this to decide whether it is safe to
> eliminate a duplicate load, replace a load with the value originally
> stored in the slot, or lift a load out of a loop.
>
> In particular, the PointerBase data field may contain GC-moveable
> addresses, but only for internal TypedData objects and views, not
> for external TypedData objects or Pointers. To allow load optimizations
> involving the latter, the LoadField and StoreField instructions now
> take boolean flags for whether loads or stores from the slot are
> guaranteed to not be GC-moveable, to override the information from
> the slot argument.
>
> Notable benchmark changes on x64 (similar for other archs unless noted):
>
> JIT:
> * FfiMemory.PointerPointer: 250.7%
> * FfiStructCopy.Copy1Bytes: -26.73% (only x64)
> * FfiStructCopy.Copy32Bytes: -25.18% (only x64)
> * MemoryCopy.64.setRange.Pointer.Uint8: 19.36%
> * MemoryCopy.64.setRange.Pointer.Double: 18.96%
> * MemoryCopy.8.setRange.Pointer.Double: 17.59%
> * MemoryCopy.8.setRange.Pointer.Uint8: 19.46%
>
> AOT:
> * FfiMemory.PointerPointer: 323.5%
> * FfiStruct.FieldLoadStore: 483.3%
> * FileIO_readwrite_64kb: 15.39%
> * FileIO_readwrite_512kb (Intel Xeon): 46.22%
> * MemoryCopy.512.setRange.Pointer.Uint8: 35.20%
> * MemoryCopy.64.setRange.Pointer.Uint8: 55.40%
> * MemoryCopy.512.setRange.Pointer.Double: 29.45%
> * MemoryCopy.64.setRange.Pointer.Double: 60.37%
> * MemoryCopy.8.setRange.Pointer.Double: 59.54%
> * MemoryCopy.8.setRange.Pointer.Uint8: 55.40%
> * FfiStructCopy.Copy32Bytes: 398.3%
> * FfiStructCopy.Copy1Bytes: 1233%
>
> TEST=vm/dart/address_local_pointer, vm/dart/pointer_as_typed_list
>
> Issue: https://github.com/dart-lang/sdk/issues/42072
> Fixes: https://github.com/dart-lang/sdk/issues/53124
>
> Cq-Include-Trybots: luci.dart.try:vm-ffi-qemu-linux-release-arm-try,vm-eager-optimization-linux-release-x64-try,vm-linux-release-x64-try,vm-linux-debug-x64-try,vm-aot-linux-release-x64-try,vm-aot-linux-debug-x64-try
> Change-Id: I563e0bfac5b1ac6cf1111649934067c12891b631
> Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/324820
> Reviewed-by: Alexander Markov <alexmarkov@google.com>
> Commit-Queue: Tess Strickland <sstrickl@google.com>
> Reviewed-by: Martin Kustermann <kustermann@google.com>
Issue: https://github.com/dart-lang/sdk/issues/42072
Change-Id: I7c31434e01108487de69a32154bbefd1538c6f0f
Cq-Include-Trybots: luci.dart.try:vm-ffi-qemu-linux-release-arm-try,vm-eager-optimization-linux-release-x64-try,vm-linux-release-x64-try,vm-linux-debug-x64-try,vm-aot-linux-release-x64-try,vm-aot-linux-debug-x64-try
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/330523
Bot-Commit: Rubber Stamper <rubber-stamper@appspot.gserviceaccount.com>
Commit-Queue: Alexander Aprelev <aam@google.com>
Auto-Submit: Alexander Markov <alexmarkov@google.com>
Reviewed-by: Alexander Aprelev <aam@google.com>
This CL adds the ability to pass the payload address of the source
and destination directly to the MemoryCopy instruction as an untagged
value.
The new translation of the _TypedListBase._memMoveN methods use the new
MemoryCopy constructor, retrieving the untagged value of the data field
of both the source and destination. This way, if inlining exposes the
allocation of the object from which the data field is being retrieved,
then allocation sinking can remove the intermediate allocation if there
are no escaping uses of the object.
Since Pointer.asTypedList allocates such ExternalTypedData objects,
this CL makes that method inlined if at all possible, which removes
the intermediate allocation if the only use of the TypedData object
is to call setRange for memory copying purposes.
This CL also separates unboxed native slots into two groups: those
that contain untagged addresses and those that do not. The former
group now have the kUntagged representation, which mimics the old
use of LoadUntagged for the PointerBase data field and also ensures
that any arithmetic operations on untagged addresses must first be
explicitly converted to an unboxed integer and then explicitly converted
back to untagged before being stored in a slot that contains untagged
addresses.
When a unboxed native slot that contains untagged addresses is defined,
the definition also includes a boolean which represents whether
addresses that may be moved by the GC can be stored in this slot or not.
The redundancy eliminator uses this to decide whether it is safe to
eliminate a duplicate load, replace a load with the value originally
stored in the slot, or lift a load out of a loop.
In particular, the PointerBase data field may contain GC-moveable
addresses, but only for internal TypedData objects and views, not
for external TypedData objects or Pointers. To allow load optimizations
involving the latter, the LoadField and StoreField instructions now
take boolean flags for whether loads or stores from the slot are
guaranteed to not be GC-moveable, to override the information from
the slot argument.
Notable benchmark changes on x64 (similar for other archs unless noted):
JIT:
* FfiMemory.PointerPointer: 250.7%
* FfiStructCopy.Copy1Bytes: -26.73% (only x64)
* FfiStructCopy.Copy32Bytes: -25.18% (only x64)
* MemoryCopy.64.setRange.Pointer.Uint8: 19.36%
* MemoryCopy.64.setRange.Pointer.Double: 18.96%
* MemoryCopy.8.setRange.Pointer.Double: 17.59%
* MemoryCopy.8.setRange.Pointer.Uint8: 19.46%
AOT:
* FfiMemory.PointerPointer: 323.5%
* FfiStruct.FieldLoadStore: 483.3%
* FileIO_readwrite_64kb: 15.39%
* FileIO_readwrite_512kb (Intel Xeon): 46.22%
* MemoryCopy.512.setRange.Pointer.Uint8: 35.20%
* MemoryCopy.64.setRange.Pointer.Uint8: 55.40%
* MemoryCopy.512.setRange.Pointer.Double: 29.45%
* MemoryCopy.64.setRange.Pointer.Double: 60.37%
* MemoryCopy.8.setRange.Pointer.Double: 59.54%
* MemoryCopy.8.setRange.Pointer.Uint8: 55.40%
* FfiStructCopy.Copy32Bytes: 398.3%
* FfiStructCopy.Copy1Bytes: 1233%
TEST=vm/dart/address_local_pointer, vm/dart/pointer_as_typed_list
Issue: https://github.com/dart-lang/sdk/issues/42072
Fixes: https://github.com/dart-lang/sdk/issues/53124
Cq-Include-Trybots: luci.dart.try:vm-ffi-qemu-linux-release-arm-try,vm-eager-optimization-linux-release-x64-try,vm-linux-release-x64-try,vm-linux-debug-x64-try,vm-aot-linux-release-x64-try,vm-aot-linux-debug-x64-try
Change-Id: I563e0bfac5b1ac6cf1111649934067c12891b631
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/324820
Reviewed-by: Alexander Markov <alexmarkov@google.com>
Commit-Queue: Tess Strickland <sstrickl@google.com>
Reviewed-by: Martin Kustermann <kustermann@google.com>
Will warm up the memory cache and then measure how much live memory the analyzer consumes after visiting all elements with the warmed up cache.
All results
--------------------------------
flutter_elements
reachableObjects
count: 9071480
size: 799672175 = 780929 KB = 762 MB
_SimpleUri
count: 3112
size(shallow): 248960 = 243 KB
duplicateCount: 0
Change-Id: I75209f88f6f615172127cd439afe9d0cd83b1c4f
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/329682
Reviewed-by: Martin Kustermann <kustermann@google.com>
Commit-Queue: Konstantin Shcheglov <scheglov@google.com>
Currently the heap snapshot tool uses a regular Set<int> for storing
ints between 0 and the number of objects (n).
This CL instead introduces a specialized int set backed by a Uint8List
where each bit represents a number between 0 and n.
This specialized set is then created to fit the number of objects, so
even an empty set takes the same amount of space as a full set: about
125 kb per million possible values (or ~6 mb per 50 million).
For small sets thats a lot worse than before, but for big sets it's a
lot better. Runtime is also a lot better.
"Benchmarks" on a snapshot that's ~1.5GB:
Runtimes:
Before:
closure roots (or all after the first one): ~30s, ~19s, ~20s
eval and: ~5.5s, ~6.3, ~5.6
eval or: ~7.3s, ~6.5s, ~6.4s
eval minus: ~4.5s, ~6.4s, ~6.5s
With CL:
closure roots (or all after the first one): ~7.7s, ~4.3s, ~4.1s
eval and: ~0.2s, ~0.3s, ~0.2s
eval or: ~0.6s, ~0.6s, ~0.6s
eval minus: ~0.3s, ~0.2s, ~0.3s
Memory usage:
Before:
after loaded:
$ grep -E "VmPeak|VmSize|VmRSS" /proc/588700/status
VmPeak: 10293036 kB
VmSize: 9896880 kB
VmRSS: 9260604 kB
after all = closure roots
VmPeak: 12354840 kB
VmSize: 9529612 kB
VmRSS: 8898088 kB
after 2 x closure all
VmPeak: 13068580 kB
VmSize: 12478236 kB
VmRSS: 11891620 kB
after 3 x eval and
VmPeak: 18377540 kB
VmSize: 18376520 kB
VmRSS: 17803032 kB
after 3 x eval or
VmPeak: 19621712 kB
VmSize: 18114376 kB
VmRSS: 17491808 kB
after 3 x eval minus
VmPeak: 21522780 kB
VmSize: 21522272 kB
VmRSS: 20907724 kB
With CL:
after loaded:
$ grep -E "VmPeak|VmSize|VmRSS" /proc/594718/status
VmPeak: 10293560 kB
VmSize: 9967036 kB
VmRSS: 9323404 kB
after all = closure roots
VmPeak: 10293560 kB
VmSize: 9805272 kB
VmRSS: 9137364 kB
after 2 x closure all
VmPeak: 10868740 kB
VmSize: 10867208 kB
VmRSS: 10199784 kB
after 3 x eval and
VmPeak: 10906236 kB
VmSize: 10905728 kB
VmRSS: 10238424 kB
after 3 x eval or
VmPeak: 10925496 kB
VmSize: 10924988 kB
VmRSS: 10257388 kB
after 3 x eval minus
VmPeak: 10944756 kB
VmSize: 10944248 kB
VmRSS: 10276672 kB
Change-Id: I8d0b65fa51ac2bd4696c4d2782c4423966a7fbae
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/324682
Reviewed-by: Martin Kustermann <kustermann@google.com>
Commit-Queue: Jens Johansen <jensj@google.com>
We were exporting all symbols to the dynamic table so that they could be looked up using `dladdr` for the profiler and backtracer. The symbols include our statically-linked libcxx, which can create trouble when another DSO has a different version of libcxx. Now we export only the VM embedding API functions (`Dart_*`) and use a specially produced table to do the symbolization.
TEST=runtime/tests/vm/dart/exported_symbols_test.dart
TEST=runtime/tests/vm/dart/symbolized_crash_test.dart
Bug: https://github.com/dart-lang/sdk/issues/53267
Change-Id: I2ee494fba86f67127ba0f6f402622f01a4662207
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/323702
Reviewed-by: Daco Harkes <dacoharkes@google.com>
Commit-Queue: Ryan Macnak <rmacnak@google.com>
`DartType.VOID` and `DartType.VOID_NULLABLE` are the same constant,
therefore only one of them should be listed in
`DartLib.typeToLibraryMethods`, to avoid an `equal_keys_in_map`
warning.
Change-Id: I65beef01f8a810af96be5e458eeecef7a7acff01
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/321820
Reviewed-by: Ben Konyi <bkonyi@google.com>
Commit-Queue: Ben Konyi <bkonyi@google.com>
Auto-Submit: Paul Berry <paulberry@google.com>
Siva Annamalai