Sunspider-regexp-dna takes 107% more time on windows

RESOLVED FIXED in mozilla34

Status

()

Core
JavaScript Engine
RESOLVED FIXED
4 years ago
4 years ago

People

(Reporter: h4writer, Assigned: h4writer)

Tracking

(Blocks: 1 bug)

unspecified
mozilla34
x86
Windows 7
Points:
---

Firefox Tracking Flags

(Not tracked)

Details

Attachments

(1 attachment)

(Assignee)

Description

4 years ago
When running ss-regexp-dna with --no-ion I get the following difference on Windows/Linux (on the same computer):

Windows: 29ms
Linux: 14ms

I can only assume it can be related to two things.
1) Concat of strings is slower on windows?
2) The regexp could get transformed into string matching on which linux has a superiour memchr.
(Assignee)

Updated

4 years ago
Blocks: 1028242
(Assignee)

Comment 1

4 years ago
Ok, this is a windows-shell only problem. Apparently the "load()" js function is very unoptimized on windows. In the browser this isn't a problem, since we don't have/use "load". Though in the shell builds we do.

This also explains the discrepancy viewed on AWFY:
http://arewefastyet.com/#machine=17&view=single&suite=ss&subtest=dna&start=1407832202&end=1407915344
(Assignee)

Comment 2

4 years ago
Created attachment 8477615 [details] [diff] [review]
Check if having _getc_nolock

Apparently we aren't using the fast call _getc_nolock anymore, since we are skipping compiler and library checks.

This fixes the seen issue.
Attachment #8477615 - Flags: review?(jorendorff)
(Assignee)

Updated

4 years ago
Assignee: nobody → hv1989
Component: JavaScript Engine: JIT → JavaScript Engine
Summary: [BaselineCompiler] Sunspider-regexp-dna takes 107% more time on windows → Sunspider-regexp-dna takes 107% more time on windows
Comment on attachment 8477615 [details] [diff] [review]
Check if having _getc_nolock

Review of attachment 8477615 [details] [diff] [review]:
-----------------------------------------------------------------

Thanks for the patch.

Incidentally, in English, "seen" cannot modify a noun attributively. I don't know why. There seems to be no general rule.

    a/the known [noun]      -  756 hits in http://corpus.byu.edu/coca/
    a/the broken [noun]     - 2811
    a/the written [noun]    - 2036
    a/the observed [noun]   -  622
    a/the seen [noun]       -   16, all weird

What's really crazy is that "unseen" is used this way:

    a/the unseen [noun]     -  199 (!)
Attachment #8477615 - Flags: review?(jorendorff) → review+
(Assignee)

Comment 4

4 years ago
Oh I had no idea. Thanks for the explanation.

https://hg.mozilla.org/integration/mozilla-inbound/rev/e2eda589ea8d
https://hg.mozilla.org/mozilla-central/rev/e2eda589ea8d
Status: NEW → RESOLVED
Last Resolved: 4 years ago
Resolution: --- → FIXED
Target Milestone: --- → mozilla34
You need to log in before you can comment on or make changes to this bug.