1. 08 Mar, 2022 1 commit
    • Joyee Cheung's avatar
      [ic] name Set/Define/Store property operations more consistently · 0d1ffe30
      Joyee Cheung authored
      For background and reasoning, see
      https://docs.google.com/document/d/1jvSEvXFHRkxg4JX-j6ho3nRqAF8vZI2Ai7RI8AY54gM/edit
      This is the first step towards pulling the DefineNamedOwn operation out
      of StoreIC.
      
      Summary of the renamed identifiers:
      
      Bytecodes:
      
      - StaNamedProperty -> SetNamedProperty: calls StoreIC and emitted for
        normal named property sets like obj.x = 1.
      - StaNamedOwnProperty -> DefineNamedOwnProperty: calls
        DefineNamedOwnIC (previously StoreOwnIC), and emitted for
        initialization of named properties in object literals and named
        public class fields.
      - StaKeyedProperty -> SetKeyedProperty: calls KeyedStoreIC and emitted
        for keyed property sets like obj[x] = 1.
      - StaKeyedPropertyAsDefine -> DefineKeyedOwnProperty: calls
        DefineKeyedOwnIC (previously KeyedDefineOwnIC) and emitted for
        initialization of private class fields and computed public class
        fields.
      - StaDataPropertyInLiteral -> DefineKeyedOwnPropertyInLiteral: calls
        DefineKeyedOwnPropertyInLiteral runtime function (previously
        DefineDataPropertyInLiteral) and emitted for initialization of keyed
        properties in object literals and static class initializers. (note
        that previously the StoreDataPropertyInLiteral runtime function name
        was taken by object spreads and array literal creation instead)
      - LdaKeyedProperty -> GetKeyedProperty, LdaNamedProperty ->
        GetNamedProperty, LdaNamedPropertyFromSuper ->
        GetNamedPropertyFromSuper: we drop the Sta prefix for the property
        store operations since the accumulator use is implicit and to make
        the wording more natural, for symmetry the Lda prefix for the
        property load operations is also dropped.
      
      opcodes:
      
      - (JS)StoreNamed -> (JS)SetNamedProperty: implements set semantics for
        named properties, compiled from SetNamedProperty (previously
        StaNamedProperty) and lowers to StoreIC or Runtime::kSetNamedProperty
      - (JS)StoreNamedOwn -> (JS)DefineNamedOwnProperty: implements define
        semantics for initializing named own properties in object literal and
        public class fields, compiled from DefineNamedOwnProperty (previously
        StaNamedOwnProperty) and lowers to DefineNamedOwnIC
        (previously StoreOwnIC)
      - (JS)StoreProperty -> (JS)SetKeyedProperty: implements set semantics
        for keyed properties, only compiled from SetKeyedProperty(previously
        StaKeyedProperty) and lowers to KeyedStoreIC
      - (JS)DefineProperty -> (JS)DefineKeyedOwnProperty: implements define
        semantics for initialization of private class fields and computed
        public class fields, compiled from DefineKeyedOwnProperty (previously
        StaKeyedPropertyAsDefine) and calls DefineKeyedOwnIC (previously
        KeyedDefineOwnIC).
      - (JS)StoreDataPropertyInLiteral ->
        (JS)DefineKeyedOwnPropertyInLiteral: implements define semantics for
        initialization of keyed properties in object literals and static
        class initializers, compiled from DefineKeyedOwnPropertyInLiteral
        (previously StaDataPropertyInLiteral) and calls the
        DefineKeyedOwnPropertyInLiteral runtime function (previously
        DefineDataPropertyInLiteral).
      
      Runtime:
      - DefineDataPropertyInLiteral -> DefineKeyedOwnPropertyInLiteral:
        following the bytecode/opcodes change, this is used by
        DefineKeyedOwnPropertyInLiteral (previously StaDataPropertyInLiteral)
        for object and class literal initialization.
      - StoreDataPropertyInLiteral -> DefineKeyedOwnPropertyInLiteral_Simple:
        it's just a simplified version of DefineDataPropertyInLiteral that
        does not update feedback or perform function name configuration.
        This is used by object spread and array literal creation. Since we
        are renaming DefineDataPropertyInLiteral to
        DefineKeyedOwnPropertyInLiteral, rename this simplified version with
        a `_Simple` suffix. We can consider merging it into
        DefineKeyedOwnPropertyInLiteral in the future. See
        https://docs.google.com/document/d/1jvSEvXFHRkxg4JX-j6ho3nRqAF8vZI2Ai7RI8AY54gM/edit?disco=AAAAQQIz6mU
      - Other changes following the bytecode/IR changes
      
      IC:
      
      - StoreOwn -> DefineNamedOwn: used for initialization of named
        properties in object literals and named public class fields.
        - StoreOwnIC -> DefineNamedOwnIC
        - StoreMode::kStoreOwn -> StoreMode::kDefineNamedOwn
        - StoreICMode::kStoreOwn -> StoreICMode::kDefineNamedOwn
        - IsStoreOwn() -> IsDefineNamedOwn()
      - DefineOwn -> DefineKeyedOwn: IsDefineOwnIC() was already just
        IsDefineKeyedOwnIC(), and IsAnyDefineOwn() includes both named and
        keyed defines so we don't need an extra generic predicate.
        - StoreMode::kDefineOwn -> StoreMode::kDefineKeyedOwn
        - StoreICMode::kDefineOwn -> StoreICMode::kDefineKeyedOwn
        - IsDefineOwn() -> IsDefineKeyedOwn()
        - IsDefineOwnIC() -> IsDefineKeyedOwnIC()
        - Removing IsKeyedDefineOwnIC() as its now a duplicate of
          IsDefineKeyedOwnIC()
      - KeyedDefineOwnIC -> DefineKeyedOwnIC,
        KeyedDefineOwnGenericGenerator() -> DefineKeyedOwnGenericGenerator:
        make the ordering of terms more consistent
      - IsAnyStoreOwn() -> IsAnyDefineOwn(): this includes the renamed and
        DefineNamedOwn and DefineKeyedOwn. Also is_any_store_own() is
        removed since it's just a duplicate of this.
      - IsKeyedStoreOwn() -> IsDefineNamedOwn(): it's unclear where the
        "keyed" part came from, but it's only used when DefineNamedOwnIC
        (previously StoreOwnIC) reuses KeyedStoreIC, so rename it accordingly
      
      Interpreter & compiler:
      - BytecodeArrayBuilder: following bytecode changes
          - StoreNamedProperty -> SetNamedProperty
        - StoreNamedOwnProperty -> DefineNamedOwnProperty
        - StoreKeyedProperty -> SetKeyedProperty
        - DefineKeyedProperty -> DefineKeyedOwnProperty
        - StoreDataPropertyInLiteral -> DefineKeyedOwnPropertyInLiteral
      - FeedbackSlotKind:
        - kDefineOwnKeyed -> kDefineKeyedOwn: make the ordering of terms more
          consistent
        - kStoreOwnNamed -> kDefineNamedOwn: following the IC change
        - kStoreNamed{Sloppy|Strict} -> kSetNamed{Sloppy|Strict}: only
          used in StoreIC for set semantics
        - kStoreKeyed{Sloppy|Strict} -> kSetKeyed{Sloppy|Strict}: only used
          in KeyedStoreIC for set semantics
        - kStoreDataPropertyInLiteral -> kDefineKeyedOwnPropertyInLiteral:
          following the IC change
      - BytecodeGraphBuilder
        - StoreMode::kNormal, kOwn -> NamedStoreMode::kSet, kDefineOwn: this
          is only used by BytecodeGraphBuilder::BuildNamedStore() to tell the
          difference between SetNamedProperty and DefineNamedOwnProperty
          operations.
      
      Not changed:
      
      - StoreIC and KeyedStoreIC currently contain mixed logic for both Set
        and Define operations, and the paths are controlled by feedback. The
        plan is to refactor the hierarchy like this:
        ```
        - StoreIC
          - DefineNamedOwnIC
          - SetNamedIC (there could also be a NamedStoreIC if that's helpful)
          - KeyedStoreIC
            - SetKeyedIC
            - DefineKeyedOwnIC
            - DefineKeyedOwnICLiteral (could be merged into DefineKeyedOwnIC)
            - StoreInArrayLiteralIC
          - ...
        ```
        StoreIC and KeyedStoreIC would then contain helpers shared by their
        subclasses, therefore it still makes sense to keep the word "Store"
        in their names since they would be generic base classes for both set
        and define operations.
      - The Lda and Sta prefixes of bytecodes not involving object properties
        (e.g. Ldar, Star, LdaZero) are kept, since this patch focuses on
        property operations, and distinction between Set and Define might be
        less relevant or nonexistent for bytecodes not involving object
        properties. We could consider rename some of them in future patches
        if that's helpful though.
      
      Bug: v8:12548
      Change-Id: Ia36997b02f59a87da3247f20e0560a7eb13077f3
      Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3481475Reviewed-by: 's avatarLeszek Swirski <leszeks@chromium.org>
      Reviewed-by: 's avatarTobias Tebbi <tebbi@chromium.org>
      Reviewed-by: 's avatarIgor Sheludko <ishell@chromium.org>
      Reviewed-by: 's avatarDominik Inführ <dinfuehr@chromium.org>
      Reviewed-by: 's avatarShu-yu Guo <syg@chromium.org>
      Reviewed-by: 's avatarJakob Gruber <jgruber@chromium.org>
      Reviewed-by: 's avatarToon Verwaest <verwaest@chromium.org>
      Commit-Queue: Joyee Cheung <joyee@igalia.com>
      Cr-Commit-Position: refs/heads/main@{#79409}
      0d1ffe30
  2. 02 Nov, 2021 1 commit
  3. 21 Oct, 2021 1 commit
  4. 13 Oct, 2021 1 commit
    • Joyee Cheung's avatar
      [class] Add IC support for defining class fields to replace runtime call · 713ebae3
      Joyee Cheung authored
      Introduces several new runtime mechanics for defining private fields,
      including:
        - Bytecode StaKeyedPropertyAsDefine
        - Builtins StoreOwnIC{Trampoline|Baseline|_NoFeedback}
        - Builtins KeyedDefineOwnIC{Trampoline|Baseline|_Megamorphic}
        - TurboFan IR opcode JSDefineProperty
      
      These new operations can reduce a runtime call per class field into a
      more traditional Store equivalent. In the microbenchmarks, this
      results in a substantial win over the status quo (~8x benchmark score
      for single fields with the changes, ~20x with multiple fields).
      
      The TurboFan JSDefineProperty op is lowered in
      JSNativeContextSpecialization, however this required some hacks.
      Because private fields are defined as DONT_ENUM when added to the
      object, we can't find a suitable transition using the typical data
      property (NONE) flags. I've added a mechanism to specify the required
      PropertyAttributes for the transition we want to look up.
      
      Details:
      
      New bytecodes:
        - StaKeyedPropertyAsDefine, which is essentially StaKeyedProperty
          but with a different IC builtin (KeyedDefineOwnIC). This is a
          bytecode rather than a flag for the existing StaKeyedProperty in
          order to avoid impacting typical keyed stores in any way due to
          additional branching and testing.
      
      New builtins:
        - StoreOwnIC{TTrampoline|Baseline|_NoFeedback} is now used for
          StaNamedOwnProperty. Unlike the regular StoreIC, this variant will
          no longer look up the property name in the prototype.
          In adddition, this CL changes an assumption that
          StoreNamedOwnProperty can't result in a map transition, as we
          can't rely on the property already being present in the Map due
          to an object literal boilerplate.
      
          In the context of class features, this replaces the runtime
          function %CreateDataProperty().
      
        - KeyedDefineOwnIC{Trampoline|Baseline|_Megamorphic} is used by the
          new StaKeyedPropertyAsDefine bytecode. This is similar to an
          ordinary KeyedStoreIC, but will not check the prototype for
          setters, and for private fields, will take the slow path if the
          field already exists.
      
          In the context of class features, this replaces the runtime
          function %AddPrivateField().
      
      TurboFan IR:
        - JSDefineProperty is introduced to represent a situation where we
          need to use "Define" semantics, in particular, it codifies that we
          do not consult the prototype chain, and the semantics relating to
          private fields are implied as well.
      
      R=leszeks@chromium.org, syg@chromium.org, rmcilroy@chromium.org
      
      Bug: v8:9888
      Change-Id: Idcc947585c0e612f9e8533aa4e2e0f8f0df8875d
      Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2795831Reviewed-by: 's avatarIgor Sheludko <ishell@chromium.org>
      Reviewed-by: 's avatarMichael Stanton <mvstanton@chromium.org>
      Reviewed-by: 's avatarLeszek Swirski <leszeks@chromium.org>
      Reviewed-by: 's avatarGeorg Neis <neis@chromium.org>
      Reviewed-by: 's avatarShu-yu Guo <syg@chromium.org>
      Commit-Queue: Joyee Cheung <joyee@igalia.com>
      Cr-Commit-Position: refs/heads/main@{#77377}
      713ebae3