[Webkit-unassigned] [Bug 156404] New: TypedArray.subarray slowdown by 5x

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=156404

            Bug ID: 156404
           Summary: TypedArray.subarray slowdown by 5x
    Classification: Unclassified
           Product: WebKit
           Version: Safari Technology Preview
          Hardware: Macintosh
                OS: OS X 10.11
            Status: NEW
          Severity: Normal
          Priority: P2
         Component: JavaScriptCore
          Assignee: webkit-unassigned at lists.webkit.org
          Reporter: puurtuur at me.com

In STP (and nightlies) TypedArray.subarray has slowed down by about 5x compared to Safari 9.1.

See the following test page (runs tests on page load):

http://zenmumbler.net/bugs/array_creation.html

On my 2013 iMac, I get the following (representative run, reproducible every time):

SAFARI 9.1
Iterations: 25, Items: 200000
F32: Min: 3.030, Max: 6.795, Avg: 3.951
Sub: Min: 10.705, Max: 14.020, Avg: 12.179
Poo: Min: 33.810, Max: 35.875, Avg: 34.790
Arr: Min: 2.115, Max: 5.265, Avg: 2.761

SAFARI TECHNOLOGY PREVIEW
Iterations: 25, Items: 200000
F32: Min: 2.225, Max: 4.155, Avg: 2.721
Sub: Min: 55.300, Max: 62.155, Avg: 58.000
Poo: Min: 48.705, Max: 49.750, Avg: 49.127
Arr: Min: 1.975, Max: 3.640, Avg: 2.390

So, while creating basic Float32Arrays has improved, using .subarray() has slowed down tremendously, which I feel directly in my 3D engine which deals a lot with binary data.
In general, TypedArrays are already much slower than normal arrays (.set() is also really slow), but this is a remarkable slowdown from one version to the next.

-- 
You are receiving this mail because:
You are the assignee for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.webkit.org/pipermail/webkit-unassigned/attachments/20160408/f44379d4/attachment.html>