Previously we would only fuse this pattern for Smi operations
which (almost) never happens in AOT. This CL enables fusion
for Int64 operations as well.
The implementation is limited to X64 and ARM64 for now
because implementing it on 32-bit platforms is somewhat
cumbersome.
Issue https://github.com/dart-lang/sdk/issues/55522R=alexmarkov@google.com
TEST=vm/cc/IL_TestIntInstr,vm/dart/test_int_pattern_il_test
Cq-Include-Trybots: luci.dart.try:vm-linux-release-simarm-try,vm-aot-linux-release-simarm_x64-try,vm-aot-linux-debug-simarm_x64-try,dart-sdk-linux-riscv64-try,vm-aot-linux-debug-simriscv64-try,vm-ffi-qemu-linux-release-riscv64-try,vm-linux-debug-simriscv64-try,vm-linux-release-ia32-try
Change-Id: I62a482640db45befac6b0b78850f23a8cc624c75
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/365463
Reviewed-by: Alexander Markov <alexmarkov@google.com>
Commit-Queue: Slava Egorov <vegorov@google.com>
Try to keep loop blocks together, previously AOT compiler would produce
bad block order for loops like these:
while (true) {
if (smth) {
// (*)
return;
}
}
And put `(*)` blocks in between loop header and other loop blocks.
TEST=ci
Cq-Include-Trybots: luci.dart.try:vm-aot-linux-debug-x64-try,vm-aot-mac-release-arm64-try,vm-aot-linux-release-x64-try
Change-Id: I52304810b61ff9288fe9df648b21c0258299d61e
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/358447
Reviewed-by: Alexander Markov <alexmarkov@google.com>
Commit-Queue: Slava Egorov <vegorov@google.com>
If we have constant type arguments include them
into the type.
This CL also expands IL tests framework with the
capability to match CompileTypes.
TEST=vm/dart/regress_53817_il_test
Bug: https://github.com/dart-lang/sdk/issues/53817
Cq-Include-Trybots: luci.dart.try:vm-aot-linux-product-x64-try,vm-aot-linux-release-arm64-try,vm-aot-linux-release-simarm_x64-try,vm-aot-obfuscate-linux-release-x64-try
Change-Id: I48722b002d9cd436dbc8ca3c9afbefa124f18996
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/331900
Commit-Queue: Slava Egorov <vegorov@google.com>
Reviewed-by: Daco Harkes <dacoharkes@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>
Using the preprocessor string literal operator # on __VA_ARGS__
in a variadic macro does not separately stringize each argument,
but instead creates a single string of the expanded arguments.
Thus, rename DECLARE_ATTRIBUTES to DECLARE_ATTRIBUTE and have it take a
single expression. For multiple attributes, DECLARE_ATTRIBUTES_NAMED
must be used instead.
Also, use a comma fold expression in AttributesSerializer::WriteTuple to
separately call WriteAttribute with each element of the tuple in turn,
instead of calling WriteAttribute once with all the tuple contents as
arguments.
Move the demangling of attribute names that are directly stringized
from expressions from the JSON consumer to the JSON producer.
To illustrate how to add multiple attributes to an instruction, this CL
adds two attributes to the StoreField instruction: slot and
is_initialization.
When printing an instruction in the IL matcher, if the instruction has
any boolean attributes, don't print the attribute if the value is false
and only print the name if the attribute is true to reduce clutter.
TEST=ci
Change-Id: Ie541364374171c3dcbee4b609942d415a1c464e8
Cq-Include-Trybots: luci.dart.try:vm-aot-linux-debug-x64-try,vm-aot-linux-release-x64-try
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/327864
Reviewed-by: Alexander Markov <alexmarkov@google.com>
Commit-Queue: Tess Strickland <sstrickl@google.com>
This is a partial reland of 6673f84d that just fixes the inlining of
_TypedList._getX in AOT without the changes to the base implementation
of the _TypedList._getX and _TypedList._setX functions.
TEST=vm/dart/inline_TypedList_getUint32
Fixes: https://github.com/dart-lang/sdk/issues/53513
Cq-Include-Trybots: luci.dart.try:vm-aot-linux-debug-x64-try,vm-aot-linux-release-x64-try,vm-aot-mac-release-arm64-try,dart-sdk-linux-riscv64-try,vm-aot-linux-debug-simriscv64-try
Change-Id: I2ef5244d50b2551a2326d6d511b9b720792346a5
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/327840
Reviewed-by: Alexander Markov <alexmarkov@google.com>
Commit-Queue: Tess Strickland <sstrickl@google.com>
Reviewed-by: Martin Kustermann <kustermann@google.com>
This reverts commit 6673f84d59.
Reason for revert: does not honor SupportsUnboxedSimd128(), breaking RISC-V
Original change's description:
> [vm/compiler] Perform inlining of _TypedList._getX in AOT.
>
> Before, the inliner only replaced calls to the _TypedList._getX methods
> with specialized IL if speculation was allowed. This means that the
> inlining would not happen in AOT mode, even though the generated IL
> does not require speculation.
>
> In addition, this CL replaces the native functions used for the
> base definition of _TypedList._getX and _TypedList._setX with
> versions built in the FlowGraphBuilder. With this, the VM avoids
> the overhead of going to the runtime for a native call when these
> methods are not inlined, which should also reduce the impact of
> a failure to inline.
>
> TEST=vm/dart/inline_TypedList_getUint32
>
> Issue: https://github.com/dart-lang/sdk/issues/53513
> 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-linux-release-x64-try,vm-mac-release-arm64-try,vm-kernel-precomp-linux-release-x64-try
> Change-Id: I66b6b8634b2b9b413fb745f02433eb58f2ff913e
> Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/325703
> Reviewed-by: Martin Kustermann <kustermann@google.com>
> Commit-Queue: Tess Strickland <sstrickl@google.com>
> Reviewed-by: Alexander Markov <alexmarkov@google.com>
Issue: https://github.com/dart-lang/sdk/issues/53513
Change-Id: If3a224e184f084fbe5d059cf036b2c2fb72cd57b
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-linux-release-x64-try,vm-mac-release-arm64-try,vm-kernel-precomp-linux-release-x64-try
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/327802
Auto-Submit: Ryan Macnak <rmacnak@google.com>
Reviewed-by: Alexander Markov <alexmarkov@google.com>
Commit-Queue: Alexander Markov <alexmarkov@google.com>
Bot-Commit: Rubber Stamper <rubber-stamper@appspot.gserviceaccount.com>
Before, the inliner only replaced calls to the _TypedList._getX methods
with specialized IL if speculation was allowed. This means that the
inlining would not happen in AOT mode, even though the generated IL
does not require speculation.
In addition, this CL replaces the native functions used for the
base definition of _TypedList._getX and _TypedList._setX with
versions built in the FlowGraphBuilder. With this, the VM avoids
the overhead of going to the runtime for a native call when these
methods are not inlined, which should also reduce the impact of
a failure to inline.
TEST=vm/dart/inline_TypedList_getUint32
Issue: https://github.com/dart-lang/sdk/issues/53513
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-linux-release-x64-try,vm-mac-release-arm64-try,vm-kernel-precomp-linux-release-x64-try
Change-Id: I66b6b8634b2b9b413fb745f02433eb58f2ff913e
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/325703
Reviewed-by: Martin Kustermann <kustermann@google.com>
Commit-Queue: Tess Strickland <sstrickl@google.com>
Reviewed-by: Alexander Markov <alexmarkov@google.com>
Before running CP find all branches of form `Branch(v eq C)`,
where `eq` is one of `===`, `==`, `!==` or `!=` and `C` is
a constant and insert a redefinition of `v` in the true
successor which captures the information that `v` is equal
to `C` guaranteed by the branch.
SCCP algorithm can then use this information when propagating
constants. After SCCP all redefinitions inserted are removed
again.
We don't actually replace `v` with `C` in the dominated code
because this might lead to issues later, for example
it might result in redundant phis `phi(C, v)` (`C` flowing
from a predecessor where `v == C`) which are complicated
to eliminate again.
TEST=vm/dart/comparison_canonicalization_il_test
Cq-Include-Trybots: luci.dart.try:vm-aot-linux-release-x64-try,vm-aot-linux-debug-simarm_x64-try,vm-aot-linux-debug-x64-try,vm-aot-android-release-arm64c-try
Change-Id: I388dddb97b5e35f09d9904000c585864f27400f6
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/324980
Reviewed-by: Alexander Markov <alexmarkov@google.com>
Commit-Queue: Slava Egorov <vegorov@google.com>
`IsTrivial` is supposed to return true iff `ToCompileType` would
produce a trivial (dynamic) type. However `IsTrivial` was taking
some of the irrevant parts of the `InferredTypeMetadata` into
account - meaning that it returned `false` where it should have
returned `true`.
This lead us to assign useless `dynamic` result type to various `StaticCall`, some of these calls can later be lowered to
`LoadIndexed` instructions, which will then also get the `dynamic`
result type. This useless type will then take priority over more
precise element type which could be inferred by
`LoadIndexedInstr::ComputeType`
TEST=vm/dart/load_indexed_trivial_type_il_test
Cq-Include-Trybots: luci.dart.try:vm-aot-linux-release-x64-try,vm-aot-linux-debug-simarm_x64-try,vm-aot-linux-debug-x64-try,vm-aot-android-release-arm64c-try
Change-Id: I69ee965ba0d0ac85ede3989680371f0dda19d8a9
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/325302
Reviewed-by: Alexander Markov <alexmarkov@google.com>
Commit-Queue: Slava Egorov <vegorov@google.com>
Inlined functions might contain call-sites which can be devirtualized
in the context of the caller. To achieve that we add a pass which
calls canonicalization on all dynamic calls (and their data
dependencies) between rounds of inlining.
More specifically this helps cases which require interleaving
some redundancy elimination and inlining to remove the cost of
abstractions. The primary example is for-in loop with an array of
a known type which usually requires inlining `get:iterator` and
propagating iterable value through interator's fields to specialize
and inline `get:length` and `operator[]` on the `Iterable`.
See https://github.com/dart-lang/sdk/issues/48433 for example.
TEST=vm/dart/devirtualization_during_inlining_il_test
Cq-Include-Trybots: luci.dart.try:vm-kernel-precomp-linux-debug-x64-try,vm-kernel-precomp-linux-release-simarm-try,vm-kernel-precomp-linux-product-x64-try
Change-Id: I1fae50484a111f8c21c81bcf0c3d6a63a856338a
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/279517
Reviewed-by: Alexander Markov <alexmarkov@google.com>
Commit-Queue: Slava Egorov <vegorov@google.com>
* Add Canonicalize pass after CSE to erase graph differences between
64-bit and 32-bit platforms. These differences originate from different
treatment of integers (we unbox more eagerly on 64-bit platform) and
consequently affect load forwarding - which generates some redundant[1]
phis on 64-bit platform, but not on 32-bit platform. Adding this pass
also addresses other cases of unpredictable performance when IL changes
can cause CSE not to trigger a Canonicalize inside itself and regress
performance, by inhibiting subsequent optimizations.
* Allow IL tests to check whether the graph was compiled in sound
null safety mode or not - because graphs can have minor differences.
* Use `"$i"` instead of `i.toString()` in the test to allow TFA to
infer non-nullability of fields in non-sound-nullsafety mode.
[1] The definition of redundant phi differs between canonicalization
pass and load forwarding pass. The former unwraps redefinitions and
the latter does not. We are going to address this difference in a
more systematic way when we split input dependencies from control
dependencies.
TEST=ci
Cq-Include-Trybots: luci.dart.try:vm-kernel-precomp-linux-debug-x64-try,vm-kernel-precomp-linux-product-x64-try,vm-kernel-precomp-linux-release-x64-try,vm-kernel-precomp-linux-release-simarm-try,vm-kernel-precomp-nnbd-linux-release-x64-try,vm-kernel-precomp-nnbd-linux-release-simarm_x64-try
Change-Id: Ifbd68fbb7f3b374ecee8ada1b3a4a2249139a108
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/220541
Commit-Queue: Slava Egorov <vegorov@google.com>
Reviewed-by: Daco Harkes <dacoharkes@google.com>
* Add support for dumping flow graphs in JSON (to make it machine
readable);
* Add support for controlling which functions are dumped through
a pragma annotation:
`@pragma('vm:testing:print-flow-graph'[, "passes filter"])`
* Replace simple matching DSL with programmatic matching
* Support obfuscated builds
Fixes https://github.com/dart-lang/sdk/issues/47340
TEST=vm/dart{,_2}/aot_prefer_equality_comparison_il_test
Cq-Include-Trybots: luci.dart.try:vm-kernel-precomp-linux-debug-x64-try,vm-kernel-precomp-linux-product-x64-try,vm-kernel-precomp-linux-release-x64-try
Change-Id: Ie067ba451d311e6019a8c3a88c012231b0c50eb5
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/219240
Commit-Queue: Slava Egorov <vegorov@google.com>
Reviewed-by: Alexander Markov <alexmarkov@google.com>
Vyacheslav Egorov