780185 - Turn on frame pointers on Nightly builds (--enable-profiling)

Status: RESOLVED FIXED

(SeaMonkey :: Build Config, defect)

Description

[Philip Chee]

We should turn on frame pointers on Nightly builds (--enable-profiling) so we can do profiling with the built-in profiler.

See:
https://groups.google.com/group/mozilla.dev.platform/browse_thread/thread/504366c143850a45#
https://developer.mozilla.org/en/Performance/Profiling_with_the_Built-in_Profiler

Comment 1

[Philip Chee]

And Firefox Bug 764216 - Turn on frame pointers on Nightly desktop builds

Comment 2

[Edmund Wong (:ewong)]

Attachment: Turn on frame pointers on Nightly Builds. (--enable-profiling) (v1)

Comment 3

[Justin Wood (:Callek)]

Comment on attachment 648932 [details] [diff] [review]
Turn on frame pointers on Nightly Builds. (--enable-profiling) (v1)

Review of attachment 648932 [details] [diff] [review]:
-----------------------------------------------------------------

So based on reading https://groups.google.com/forum/#!topic/mozilla.dev.platform/UENmwUOFCkU and the stuff in Bug 764216 I'm going to WONTFIX this, however I am entirely *willing* to turn it on if someone is wanting/willing to do profiling/perf testing and investigation for SeaMonkey.

Comment 4

[Justin Wood (:Callek)]

per c#3 WONTFIX, if someone wants to commit to doing said perf testing/investigation, I'll be happy to take this though

Comment 5

[Phoenix]

Finished performance and build time investigation, no negative affection detected.
Test system: Intel Pentium E2160 (1.8 GHz), 1 GB RAM
---------------------------
Build time
Profiling: 02:59:20
Regular: 03:00:47
---------------------------
SunSpider
Profiling:
Total:                  501.5ms +/- 14.8%
Total:                  449.8ms +/- 1.9%
Total:                  442.2ms +/- 1.1%
Regular:
Total:                  458.9ms +/- 5.4%
Total:                  438.4ms +/- 0.7%
Total:                  443.9ms +/- 1.4%
---------------------------
Google
Profiling:
Score: 4203
Score: 4187
Score: 4157
Regular:
Score: 4257
Score: 4116
Score: 4103
---------------------------
Futuremark Peacekeeper
Profiling:
854
Regular:
856

Comment 6

[Justin Wood (:Callek)]

(In reply to Justin Wood (:Callek) from comment #4)
> per c#3 WONTFIX, if someone wants to commit to doing said perf
> testing/investigation, I'll be happy to take this though

(In reply to Phoenix from comment #5)
> Finished performance and build time investigation, no negative affection
> detected.

For the record, the people at Mozilla who designed this flag claim it does affect performance negatively, and visibly in most cases. So forgive me if I discard the base claims here [for now.

That said, are you committing to improving SeaMonkey specific performance issues in a way that would be benefited by adding this flag? [since it also has overhead for when we migrate branches, do a release, etc. in making sure we remove the flag at the appropriate time] If you are can you give me an idea on what is helped specifically by us adding it?

Comment 7

[Phoenix]

(In reply to Justin Wood (:Callek) from comment #6)
> If you are can you give me an
> idea on what is helped specifically by us adding it?
The main goal is to make Gecko Profiling Addon work in SeaMonkey
https://github.com/bgirard/Gecko-Profiler-Addon
http://benoitgirard.wordpress.com/2012/07/27/javascript-profiling-with-the-gecko-profiler-and-js-anti-pattern/
It can help users to determine the source of performance problem faster, than in wild guessing case. Fresh example of such thread in Mozillazine:
http://forums.mozillazine.org/viewtopic.php?f=40&t=2543177

Comment 8

[Philip Chee]

http://logs.glob.uno/?c=mozilla%23seamonkey&a=date&s=Yesterday&e=Today&h=#c621328

19:18	* Px also sticks a pin to Callek's doll for bug 780185
19:21	RattyAway	"For the record, the people at Mozilla who designed this flag claim it does affect performance negatively, and visibly in most cases. So forgive me if I discard the base claims here [for now."
19:21	RattyAway	I understand that the performance regressions have all been fixed.
19:21	RattyAway	Callek: could we have your r+ now?
19:21	Callek	RattyAway: stamp+
19:22	RattyAway	Callek: thanks

Comment 9

[Edmund Wong (:ewong)]

Attachment: unbitrotted patch

Comment 10

[Edmund Wong (:ewong)]

Pushed to comm-central:
https://hg.mozilla.org/comm-central/rev/32b8c18a0abf