1665348 - Warp: Transpile LoadArgumentsObjectArgResult, LoadArgumentsObjectLengthResult, and LoadFunctionLengthResult

Status: RESOLVED FIXED

(Core :: JavaScript Engine: JIT, task, P2)

Comment 1

[André Bargull [:anba]]

Attachment: Bug 1665348 - Part 1: Don't attach LoadArgumentsObjectArgResult when it will fail. r=jandem!

Don't attach LoadArgumentsObjectArgResult when it will fail anyway. The checks mirror
the dynamic checks from CacheIRCompiler.

Comment 2

[André Bargull [:anba]]

Attachment: Bug 1665348 - Part 2: Add MacroAssembler::loadArgumentsObjectElement. r=jandem!

Add MacroAssembler::loadArgumentsObjectElement in preparation for the next
part.

Comment 3

[André Bargull [:anba]]

Attachment: Bug 1665348 - Part 3: Transpile LoadArgumentsObjectArgResult. r=jandem!

Uses the same alias set as MGetArgumentsObjectArg. A more narrow alias set
may be possible, but we need to be sure not to run into any issues with mapped
arguments.

Depends on D90403

Comment 4

[André Bargull [:anba]]

Attachment: Bug 1665348 - Part 4: Add MacroAssembler::loadArgumentsObjectLength. r=jandem!

Add in preparation for the next part.

Depends on D90404

Comment 5

[André Bargull [:anba]]

Attachment: Bug 1665348 - Part 5: Transpile LoadArgumentsObjectLengthResult. r=jandem!

Depends on D90405

Comment 6

[André Bargull [:anba]]

Attachment: Bug 1665348 - Part 6: Check property name before the lookup when trying to attach FunctionLength. r=jandem!

This avoids calling LookupPropertyPure unless the property name is actually "length".

Depends on D90406

Comment 7

[André Bargull [:anba]]

Attachment: Bug 1665348 - Part 7: Move bound-function length slot constant into FunctionExtended. r=jandem!

No self-hosted code reads BOUND_FUN_LENGTH_SLOT, so we might as well move it
to the other constants in FunctionExtended.

Drive-by fix:

• Use fallibleUnboxInt32 in MacroAssembler::loadFunctionLength().

Depends on D90407

Comment 8

[André Bargull [:anba]]

Attachment: Bug 1665348 - Part 8: Transpile LoadFunctionLengthResult. r=jandem!

Also change CacheIR to not attach the stub if it will fail.

Depends on D90408

Comment 9

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/58aa6d7e066e
Part 1: Don't attach LoadArgumentsObjectArgResult when it will fail. r=jandem
https://hg.mozilla.org/integration/autoland/rev/7f927ef4d046
Part 2: Add MacroAssembler::loadArgumentsObjectElement. r=jandem
https://hg.mozilla.org/integration/autoland/rev/d418a39177ad
Part 3: Transpile LoadArgumentsObjectArgResult. r=jandem
https://hg.mozilla.org/integration/autoland/rev/b04893b59aba
Part 4: Add MacroAssembler::loadArgumentsObjectLength. r=jandem
https://hg.mozilla.org/integration/autoland/rev/2fd7fe53abd9
Part 5: Transpile LoadArgumentsObjectLengthResult. r=jandem
https://hg.mozilla.org/integration/autoland/rev/37353059ea5a
Part 6: Check property name before the lookup when trying to attach FunctionLength. r=jandem
https://hg.mozilla.org/integration/autoland/rev/237faf2a86b3
Part 7: Move bound-function length slot constant into FunctionExtended. r=jandem
https://hg.mozilla.org/integration/autoland/rev/ad05bb3bcd45
Part 8: Transpile LoadFunctionLengthResult. r=jandem

Comment 10

[Bogdan Tara[:bogdan_tara | bogdant]]

https://hg.mozilla.org/mozilla-central/rev/58aa6d7e066e
https://hg.mozilla.org/mozilla-central/rev/7f927ef4d046
https://hg.mozilla.org/mozilla-central/rev/d418a39177ad
https://hg.mozilla.org/mozilla-central/rev/b04893b59aba
https://hg.mozilla.org/mozilla-central/rev/2fd7fe53abd9
https://hg.mozilla.org/mozilla-central/rev/37353059ea5a
https://hg.mozilla.org/mozilla-central/rev/237faf2a86b3
https://hg.mozilla.org/mozilla-central/rev/ad05bb3bcd45