961977 - IonMonkey: Consider Inlining Array.prototype.slice

Status: NEW

(Core :: JavaScript Engine: JIT, defect, P3)

Description

[Masaya Iseki[:isk](UTC+9)]

Attachment: array_slice_bench.js

I'm attaching a shell test case.

d8: 631ms
SM: 2145ms

SM is slower than d8.
I think this is because Array#slice is not inlined.

Comment 1

[Boris Zbarsky [:bzbarsky]]

Do you think that based on profiling, or something else?

Comment 2

[Nicolas B. Pierron [:nbp]]

Just wondering, what are the results if you replace arr.slice() by res.slice(), and do res = arr.slice() before the loop?

Comment 3

[Masaya Iseki[:isk](UTC+9)]

(In reply to Boris Zbarsky [:bz] from comment #1)
> Do you think that based on profiling, or something else?
Sorry I don't have exactly basis.
I think inlining Array#slice can be fast because It is similar to MStringSplit. 
But I don't have any profiling which indicate the cause is it.

Comment 4

[Masaya Iseki[:isk](UTC+9)]

(In reply to Nicolas B. Pierron [:nbp] from comment #2)
> Just wondering, what are the results if you replace arr.slice() by
> res.slice(), and do res = arr.slice() before the loop?

I measured it.
The both result is the same.

Comment 5

[Nicolas B. Pierron [:nbp]]

SM: 900ms
node: 490ms

Comment 6

[Jim Michaels]

don't forget the ability to do Array==Array.slice()

Comment 7

[Mayank Bansal]

Nightly: https://share.firefox.dev/3QH7zN3 (335ms)
Chrome: https://share.firefox.dev/3OFPKLN (110ms)