Javascript Performance Thread
-
- Posts: 24
- Joined: March 23rd, 2012, 12:48 am
Re: Javascript Performance Thread
I'm not a C++ expert but I'm wondering how this change ( http://hg.mozilla.org/projects/ionmonkey/rev/2a18a9755806 ), can affect performance almost 869 points O:
namespace change maybe?
this much!
namespace change maybe?
this much!
-
- Posts: 63
- Joined: December 3rd, 2010, 9:28 am
Re: Javascript Performance Thread
Probably there were multiple changesets and the revisions you are looking at is the most recent. Unfortunatly, the links in awfy dont give a list of all the changesets between versions. usually awfy runs are done immediately after an interesting change, but not in this case it seems (perhaps the merge bustage mentioned in the changeset had something to do with it).
Try clicking parent a few times to see other changes made recently.
Try clicking parent a few times to see other changes made recently.
-
- Posts: 93
- Joined: September 18th, 2008, 2:57 pm
Re: Javascript Performance Thread
AWFY updates every hour or so, that's why there may be multiple changesets between two runs.
This regression was definitely caused by the merge. It's one of the reasons why we want to land Ion soon: merges pull in many changesets and regressions are hard to track down.
This regression was definitely caused by the merge. It's one of the reasons why we want to land Ion soon: merges pull in many changesets and regressions are hard to track down.
-
- Posts: 24
- Joined: March 23rd, 2012, 12:48 am
Re: Javascript Performance Thread
Thanks for the info
The list here http://hg.mozilla.org/projects/ionmonkey/summary leads to the correct change set, it seems the cause is the merge:
http://hg.mozilla.org/projects/ionmonke ... 3feda8a0b3
The list here http://hg.mozilla.org/projects/ionmonkey/summary leads to the correct change set, it seems the cause is the merge:
http://hg.mozilla.org/projects/ionmonke ... 3feda8a0b3
-
- Posts: 1030
- Joined: June 28th, 2008, 4:57 am
Re: Javascript Performance Thread
Since there has been little talk of actual bugs here recently, the work in bug 774253 and related bugs looks pretty exciting! Putting everything together, it should make compiling everything with IonMonkey after the initial JM+TI compilation essentially free as long as you have a core available to do the work.
- faulty_fox
- Posts: 421
- Joined: April 20th, 2010, 10:25 am
Re: Javascript Performance Thread
662702 – [meta] Land type inference branch to Tracemonkey Resolved fixed. Are we any closer to seeing this landed in firefox?
- ferongr
- Posts: 537
- Joined: February 16th, 2011, 9:51 am
Re: Javascript Performance Thread
Release versions have had TI for a very long time, since the Fx 9 release.
What Falken giveth, the tōge taketh away.
- faulty_fox
- Posts: 421
- Joined: April 20th, 2010, 10:25 am
Re: Javascript Performance Thread
I see my mistake then sorry. I remember reading about it now. My memory fails me again, nothing new.
-
- Posts: 8829
- Joined: May 7th, 2006, 10:29 pm
- Location: California
Re: Javascript Performance Thread
We are getting pretty close to having Ionmonkey landed in the normal Nightlies, though...
-
- Posts: 1033
- Joined: July 18th, 2003, 10:09 pm
Re: Javascript Performance Thread
This next bit of Javascript performance would have to come from GC. IGC is not fully working yet and we have GGC to work. Hopefully more attention will be put on to these two projects.
-
- Posts: 169
- Joined: July 23rd, 2004, 11:48 am
- Location: London, UK
Re: Javascript Performance Thread
In 64-bit mac-mini the JM+TI+Ion hasn't been showing up for quite some time. Now it seems the same is happening to the JM+TI is also failing to show up. Anyone knows who to message to check?
http://www.arewefastyet.com/?runs=650&v ... &machine=8
In 32-bit all seems to be working fine
http://arewefastyet.com/?runs=650&view= ... &machine=9
http://www.arewefastyet.com/?runs=650&v ... &machine=8
In 32-bit all seems to be working fine
http://arewefastyet.com/?runs=650&view= ... &machine=9
-
- Posts: 42
- Joined: April 1st, 2011, 10:03 am
Re: Javascript Performance Thread
iamhere wrote:In 64-bit mac-mini the JM+TI+Ion hasn't been showing up for quite some time. Now it seems the same is happening to the JM+TI is also failing to show up. Anyone knows who to message to check?
http://www.arewefastyet.com/?runs=650&v ... &machine=8
In 32-bit all seems to be working fine
http://arewefastyet.com/?runs=650&view= ... &machine=9
Actually, I kind of doubt they've been using those machines much at all recently, due to age. I think 11 and 12 were added with the intention of being replacements for those two.
11 is a 32-bit Mac Pro: http://www.arewefastyet.com/awfy2.php?m ... ew=regress
12 is a 64-bit Mac Pro: http://www.arewefastyet.com/awfy2.php?m ... ew=regress
Also, using 650 runs seems rather excessive and unnecessary, especially when it autoupdates. You shouldn't need more than a few days to see most regressions that matter. I suggest not being so unkind to AWFY.
-
- Posts: 169
- Joined: July 23rd, 2004, 11:48 am
- Location: London, UK
Re: Javascript Performance Thread
@tamemoto Thanks for the update. I kind of wish they weren't dropping support on those machines for several reasons (still hope that's really not the case). Anyway, thank you for your comment about the runs. I know about the 650 being excessive. It was there to highlight the disappearance of such metrics in the referenced machines. Thanks again.
-
- Posts: 469
- Joined: August 30th, 2012, 4:27 am
Re: Javascript Performance Thread
For me IonMonkey x64 seemed to run a lot slower than regular nightly x64 and bench-marked about 20% slower on multiple JS benchmarks...
- Omega X
- Posts: 8225
- Joined: October 18th, 2007, 2:38 pm
- Location: A Parallel Dimension...
Re: Javascript Performance Thread
Ion was supposed to land last week, anyone know what's the hold up?