LiftoffRegister is an abstraction over Register and DoubleRegister.
Many functions in Liftoff do not need to distinguish GP and FP
registers. LiftoffRegister allows to implement most functions
generically. Accessors allow to convert them back to Register or
DoubleRegister.
Both register types are represented in a unified index space, which
allows to extend this concept to more register types and implement
aliasing in a straight-forward manner.
Many functions currently only implement the GP case. FP will be added
later.
R=titzer@chromium.org
Bug: v8:6600
Change-Id: I043b787bc09dd1a06506ad515eb325b8ea19746d
Reviewed-on: https://chromium-review.googlesource.com/793390
Reviewed-by: Ben Titzer <titzer@chromium.org>
Commit-Queue: Clemens Hammacher <clemensh@chromium.org>
Cr-Commit-Position: refs/heads/master@{#49780}
In the special case of KeyedLoadIC, where the key is always undefined,
LowerCheckEqualsInternalizedString fails and triggers deoptimization
loops. It's because {exp}, which is undefined_string, does not match
{val}, which is undefined_value.
This patch drops the convertion from undefined_value to undefined_string
for undefined key in KeyedLoadIC. It makes undefined key be able to pass
LowerCheckEqualsInternalizedString and avoids the deoptimization loops.
Bug:
Change-Id: Ie15358e705490fb2e4907f4d37a3ef7921f74844
Reviewed-on: https://chromium-review.googlesource.com/789613
Commit-Queue: Igor Sheludko <ishell@chromium.org>
Reviewed-by: Igor Sheludko <ishell@chromium.org>
Cr-Commit-Position: refs/heads/master@{#49779}
This CL introduces a Context::kInvalidContext sentinel value to make clear that
no context is active. We silently accept smi 0 (= nullptr) as a non-set context
which usually was the default value making it hard to ensure whether this
happened on purpose or not.
Change-Id: I5c35616f26b0b64c1cd976563a6eeb0ce474927d
Reviewed-on: https://chromium-review.googlesource.com/790291
Reviewed-by: Igor Sheludko <ishell@chromium.org>
Commit-Queue: Camillo Bruni <cbruni@chromium.org>
Cr-Commit-Position: refs/heads/master@{#49778}
It reflects the semantics of ToBoolean, so it must be adapted for
bigints.
Bug: v8:6791
Change-Id: I18931df21528463dacf5ad50fa8264b1c968c6b5
Reviewed-on: https://chromium-review.googlesource.com/799831
Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
Reviewed-by: Jakob Kummerow <jkummerow@chromium.org>
Commit-Queue: Georg Neis <neis@chromium.org>
Cr-Commit-Position: refs/heads/master@{#49776}
This reverts commit e42e7fc800.
Reason for revert: Speculative revert for:
https://crbug.com/v8/7149
Original change's description:
> [objects] No longer create short external strings.
>
> This fixes String::MakeExternal() to bail out if the subject string
> doesn't fit a regular ExternalString, instead of creating a short
> external string. The observation here is that for short external strings
> the overhead of having to have the StringResource plus going to the
> runtime/C++ for each and every character access from JavaScript land
> is probably bigger than the anticipated benefits.
>
> If this turns out to be wrong and there's a real benefit, we should make
> use of ThinStrings instead of having a separate way to represent
> external strings.
>
> Bug: v8:6621, v8:7109, v8:7145
> Change-Id: I4b75da08b82a72027c782a69de9c8eaf3cca1d4d
> Reviewed-on: https://chromium-review.googlesource.com/799750
> Reviewed-by: Yang Guo <yangguo@chromium.org>
> Commit-Queue: Benedikt Meurer <bmeurer@chromium.org>
> Cr-Commit-Position: refs/heads/master@{#49735}
TBR=yangguo@chromium.org,bmeurer@chromium.org
Change-Id: I3f5cfa9ab5c99ddce1d61ede9ed9515cb3936cdd
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Bug: v8:6621, v8:7109, v8:7145, v8:7149
Reviewed-on: https://chromium-review.googlesource.com/801675
Reviewed-by: Michael Achenbach <machenbach@chromium.org>
Commit-Queue: Michael Achenbach <machenbach@chromium.org>
Cr-Commit-Position: refs/heads/master@{#49770}
This save ~40KB of code size in release mode.
Bug:
Change-Id: Ibdf7e64bb2592f53a4edf07665016dd5a65714a0
Reviewed-on: https://chromium-review.googlesource.com/801515
Reviewed-by: Michael Lippautz <mlippautz@chromium.org>
Reviewed-by: Clemens Hammacher <clemensh@chromium.org>
Commit-Queue: Ulan Degenbaev <ulan@chromium.org>
Cr-Commit-Position: refs/heads/master@{#49769}
In my previous typer changes I had forgotten to remove one line.
R=jarin@chromium.org
Bug:
Change-Id: If9e9984c23273707548c2d7329f9fa715a170a90
Reviewed-on: https://chromium-review.googlesource.com/800270
Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
Commit-Queue: Georg Neis <neis@chromium.org>
Cr-Commit-Position: refs/heads/master@{#49761}
Even though a previous patch made the number of slots pushed/claimed on
the stack aligned, the boundary between frames was not a multiple of
two slots as well. We were pushing the number of arguments (which belongs
in the ArgumentAdaptor frame) together with the arguments to pass to the
callee (which belong to the frame of the callee). Those need to be
separated so we can drop the arguments without messing up the alignment.
Bug: v8:6644
Change-Id: I259c58db33a7c2726e5a3c74bcd67496f607d1d0
Reviewed-on: https://chromium-review.googlesource.com/793047
Reviewed-by: Ross McIlroy <rmcilroy@chromium.org>
Commit-Queue: Georgia Kouveli <georgia.kouveli@arm.com>
Cr-Commit-Position: refs/heads/master@{#49759}
This is always equal to isolate_ (so getting it from the isolate is totally
free), and this makes the code slightly smaller while avoiding a few loads
and stores here and there.
Change-Id: Ib648cd0a1f6ea7cbb73a0188ec92cd091ce34792
Reviewed-on: https://chromium-review.googlesource.com/794046
Reviewed-by: Marja Hölttä <marja@chromium.org>
Commit-Queue: Jeremy Roman <jbroman@chromium.org>
Cr-Commit-Position: refs/heads/master@{#49758}
This is a reland of 98cc9e862f
Original change's description:
> [test] Creating command before execution phase.
>
> Immutable command class with shell, flags and
> environment.
>
> Command creation moved from worker to the main
> process. Because of that there is no need to send
> test cases beyond process boundaries and load test
> suites in worker processes.
>
> Bug: v8:6917
> Change-Id: Ib6a44278095b4f7141eb9b96802fe3e8117678a6
> Reviewed-on: https://chromium-review.googlesource.com/791710
> Commit-Queue: Michał Majewski <majeski@google.com>
> Reviewed-by: Michael Achenbach <machenbach@chromium.org>
> Cr-Commit-Position: refs/heads/master@{#49746}
Bug: v8:6917
Change-Id: I49c29a8db813c47909f2cc45070ac7721a447c7a
Reviewed-on: https://chromium-review.googlesource.com/800370
Reviewed-by: Michael Achenbach <machenbach@chromium.org>
Commit-Queue: Michał Majewski <majeski@google.com>
Cr-Commit-Position: refs/heads/master@{#49756}
Using the "jco" gdb macro on a WasmCode address would currently crash.
With this change, you can also use it to print code objects off the gc
heap.
R=titzer@chromium.orgCC=wasm-team@google.com
Bug: v8:7140, v8:6876
Change-Id: I21daf58b6b4734e3d052b66284b95938b4f57d40
Reviewed-on: https://chromium-review.googlesource.com/796416
Reviewed-by: Ben Titzer <titzer@chromium.org>
Commit-Queue: Clemens Hammacher <clemensh@chromium.org>
Cr-Commit-Position: refs/heads/master@{#49753}
This reverts commit 98cc9e862f.
Reason for revert: Breaks test isolation:
https://build.chromium.org/p/client.v8/builders/V8%20Linux%20-%20builder/builds/29746
Original change's description:
> [test] Creating command before execution phase.
>
> Immutable command class with shell, flags and
> environment.
>
> Command creation moved from worker to the main
> process. Because of that there is no need to send
> test cases beyond process boundaries and load test
> suites in worker processes.
>
> Bug: v8:6917
> Change-Id: Ib6a44278095b4f7141eb9b96802fe3e8117678a6
> Reviewed-on: https://chromium-review.googlesource.com/791710
> Commit-Queue: Michał Majewski <majeski@google.com>
> Reviewed-by: Michael Achenbach <machenbach@chromium.org>
> Cr-Commit-Position: refs/heads/master@{#49746}
TBR=machenbach@chromium.org,sergiyb@chromium.org,majeski@google.com
Change-Id: I44b99468d18fd093833f4185dad067a9eeaf2bc1
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Bug: v8:6917
Reviewed-on: https://chromium-review.googlesource.com/800292
Reviewed-by: Michael Achenbach <machenbach@chromium.org>
Commit-Queue: Michael Achenbach <machenbach@chromium.org>
Cr-Commit-Position: refs/heads/master@{#49747}
Immutable command class with shell, flags and
environment.
Command creation moved from worker to the main
process. Because of that there is no need to send
test cases beyond process boundaries and load test
suites in worker processes.
Bug: v8:6917
Change-Id: Ib6a44278095b4f7141eb9b96802fe3e8117678a6
Reviewed-on: https://chromium-review.googlesource.com/791710
Commit-Queue: Michał Majewski <majeski@google.com>
Reviewed-by: Michael Achenbach <machenbach@chromium.org>
Cr-Commit-Position: refs/heads/master@{#49746}
Finally address that long-standing TODO where ConsString allocation in
TurboFan would always go for the two byte map instead of choosing the
one byte map if the inputs are one byte strings.
Bug: v8:5269, v8:7109
Change-Id: Ibcfceaf499ceebef0ef928ebc5f204bcacf29bc0
Reviewed-on: https://chromium-review.googlesource.com/799700
Reviewed-by: Jakob Gruber <jgruber@chromium.org>
Commit-Queue: Benedikt Meurer <bmeurer@chromium.org>
Cr-Commit-Position: refs/heads/master@{#49744}
Currently RuntimeCallStats stores CounterIds as inner pointers.
This patch replaces them with enums and removes static table.
Bug: chromium:758183
Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng
Change-Id: Icb4030fc3ad3dd02e9c2648ce7c43b6f2d47fa9d
Reviewed-on: https://chromium-review.googlesource.com/796477
Reviewed-by: Ross McIlroy <rmcilroy@chromium.org>
Reviewed-by: Camillo Bruni <cbruni@chromium.org>
Reviewed-by: Adam Klein <adamk@chromium.org>
Commit-Queue: Ulan Degenbaev <ulan@chromium.org>
Cr-Commit-Position: refs/heads/master@{#49743}
This CL fixes an issue with --enable-tracing which was introduced
recently where the tracing file was closed too early. In addition it
adds a test for --enable-tracing to avoid such an issue in the future.
R=machenbach@chromium.org, rmcilroy@chromium.org
Change-Id: I1b3699a4dfbe27230ac1d21d5f6bc7b2ee9ed435
Reviewed-on: https://chromium-review.googlesource.com/796214
Reviewed-by: Michael Achenbach <machenbach@chromium.org>
Reviewed-by: Ross McIlroy <rmcilroy@chromium.org>
Commit-Queue: Andreas Haas <ahaas@chromium.org>
Cr-Commit-Position: refs/heads/master@{#49742}
This derives two global variables to ease adding new variants.
Bug: chromium:788104
Change-Id: Ibe933dc0999f6dfa0ae8bf2afa0a5de8e44aefca
Reviewed-on: https://chromium-review.googlesource.com/799790
Commit-Queue: Michael Achenbach <machenbach@chromium.org>
Reviewed-by: Sergiy Byelozyorov <sergiyb@chromium.org>
Cr-Commit-Position: refs/heads/master@{#49741}
This prefixes the flag to be more uniform with existing flags and makes
it available as a gn argument.
The switch feature is enabled on a subset of bots, the linux64 release
trybot and all sanitizers. Note, this doesn't switch to slow paths -
it just allows to use the runtime flag on those bots.
Bug: v8:7120
Change-Id: I706a7a9bd503c13d5432cac9a13a909f4f32a0a7
Reviewed-on: https://chromium-review.googlesource.com/798412
Reviewed-by: Jakob Gruber <jgruber@chromium.org>
Commit-Queue: Michael Achenbach <machenbach@chromium.org>
Cr-Commit-Position: refs/heads/master@{#49739}
When we try to get a function literal with an ID beyond the
last known ID we easily create out-of-bound read bugs. It's preferable
to crash in this situation.
BUG=chromium:789764
R=marja@chromium.org
Change-Id: I4f35e9231ef6af18204bbac96df3652c3d30c29f
Reviewed-on: https://chromium-review.googlesource.com/798411
Reviewed-by: Marja Hölttä <marja@chromium.org>
Commit-Queue: Jochen Eisinger <jochen@chromium.org>
Cr-Commit-Position: refs/heads/master@{#49738}
Handles created inside each loop iteration should be freed as the
current iteration goes out of scope.
TBR=yangguo@chromium.org
Bug: v8:7122, chromium:763839
Change-Id: I70a95457c773b26792d87734dab62bc8f4f5fbd0
Reviewed-on: https://chromium-review.googlesource.com/796794
Reviewed-by: Jakob Gruber <jgruber@chromium.org>
Commit-Queue: Jakob Gruber <jgruber@chromium.org>
Cr-Commit-Position: refs/heads/master@{#49737}
This test expects certain functions to be optimized, which does not
happen if we force array objects onto the slow path.
Bug: v8:7122
Change-Id: I716954fff564f1c4f0782b3452557ec89a3b4307
Reviewed-on: https://chromium-review.googlesource.com/796860
Commit-Queue: Jakob Gruber <jgruber@chromium.org>
Reviewed-by: Michael Achenbach <machenbach@chromium.org>
Cr-Commit-Position: refs/heads/master@{#49736}
This fixes String::MakeExternal() to bail out if the subject string
doesn't fit a regular ExternalString, instead of creating a short
external string. The observation here is that for short external strings
the overhead of having to have the StringResource plus going to the
runtime/C++ for each and every character access from JavaScript land
is probably bigger than the anticipated benefits.
If this turns out to be wrong and there's a real benefit, we should make
use of ThinStrings instead of having a separate way to represent
external strings.
Bug: v8:6621, v8:7109, v8:7145
Change-Id: I4b75da08b82a72027c782a69de9c8eaf3cca1d4d
Reviewed-on: https://chromium-review.googlesource.com/799750
Reviewed-by: Yang Guo <yangguo@chromium.org>
Commit-Queue: Benedikt Meurer <bmeurer@chromium.org>
Cr-Commit-Position: refs/heads/master@{#49735}
In order to test that we don't repeatedly go through the
WasmCompileLazy runtime function, add a flag to the
LazyCompilationOrchestrator to "freeze" it, i.e. disallow any further
lazy compilation.
In tests, use this flag to first call a method, then freeze lazy
compilation, then call the method again to assert that no further lazy
compilation is triggered.
This test currently fails with --wasm-jit-to-native, so disable it for
that variant.
R=titzer@chromium.orgCC=mtrofin@chromium.org
Bug: v8:7140, chromium:788441, v8:5991
Change-Id: I18a40d302c24041740d8a54351d06ed968f4beec
Reviewed-on: https://chromium-review.googlesource.com/796430
Reviewed-by: Mircea Trofin <mtrofin@chromium.org>
Commit-Queue: Clemens Hammacher <clemensh@chromium.org>
Cr-Commit-Position: refs/heads/master@{#49734}
According to the style guide, enum names they must either be
capitalized, or start with a "k". I prefer the kFoo syntax.
R=mtrofin@chromium.org
Bug: v8:7109
Change-Id: I9c06c4cb05b05ec50de8d68d118f1a0807938426
Reviewed-on: https://chromium-review.googlesource.com/796856
Reviewed-by: Mircea Trofin <mtrofin@chromium.org>
Commit-Queue: Clemens Hammacher <clemensh@chromium.org>
Cr-Commit-Position: refs/heads/master@{#49733}
Makes it easier to modify JSFunction; the failing asserts will tell which places
need to be modified.
Change-Id: I70c7cd0b89cf05964b4730990fb3d7f23927437a
Reviewed-on: https://chromium-review.googlesource.com/781689
Reviewed-by: Igor Sheludko <ishell@chromium.org>
Reviewed-by: Benedikt Meurer <bmeurer@chromium.org>
Commit-Queue: Marja Hölttä <marja@chromium.org>
Cr-Commit-Position: refs/heads/master@{#49732}
Strings are immutable in JavaScript land (contrast with the runtime,
where we can truncate strings that haven't escaped to JavaScript yet),
so the length of a String is immutable. Thus loading the length of a
String is a pure operation and should be expressed as such (i.e. doesn't
depend on control or effect). The StringLength operator does exactly
this and is hooked up to the effect chain in the EffectControlLinearizer.
This will eventually allow us to simplify the optimization of string
concatention and other operations that are a bit cumbersome in TurboFan
currently, and it will also allow us to optimize string operations
across effectful operations, for example combining multiple invocations
to String#slice with the same inputs.
Bug: v8:5269, v8:6936, v8:7109, v8:7137
Change-Id: Iffcccbb0c7fc4cfe1281c10e7af24b40eba4c987
Reviewed-on: https://chromium-review.googlesource.com/799690
Reviewed-by: Yang Guo <yangguo@chromium.org>
Reviewed-by: Benedikt Meurer <bmeurer@chromium.org>
Commit-Queue: Benedikt Meurer <bmeurer@chromium.org>
Cr-Commit-Position: refs/heads/master@{#49731}
