Welcome to the Second Life Forums Archive

These forums are CLOSED. Please visit the new forums HERE

Listen VS LinkMessage, the Truth

Rhombur Volos
King of Scripture & Kebab
Join date: 6 Oct 2004
Posts: 102
02-08-2005 06:50
1. Afaik, there's a numerical limit to how many llListens an object or script can use. I've heard many not-so-well scripted objects scream errors about this.

2. Quote from the wiki: "Note that llMessageLinked() has been shown through practical experiments to be much faster than standard llWhisper() or llSay() calls."
Francis Chung
This sentence no verb.
Join date: 22 Sep 2003
Posts: 918
02-08-2005 07:16
From: Jesrad Seraph
Sim FPS with 0 listeners: 996
Sim FPS with 6400 listeners (64 listeners in 100 cubes): 906

Test done in Eclipse.


The Lindens seem to have improved the drain on listens, my tests say it's *roughly* two-fold since I last ran a study.

/54/c8/25865/1.html

I'm here in Meins, and the SIM FPS is roughly 6000. Rezzing 512 listeners (like I did last time) drops the sim fps to roughly 2000.

Maybe someone wants to re-run the whole suite of experiments?

But, FYI, 64 listeners in 1 cube is not the same thing as 64 listeners in 64 cubes.

I ran a test earlier in Aleph. 4096 listeners in 64 cubes had negligible effect on the the sim performance, but 4096 listens in 4096 cubes dropped the sim FPS from 1400 to 140. For reference, 4096 default scripts had no measurable performance impact.
_____________________
--
~If you lived here, you would be home by now~
Sitting Lightcloud
Registered User
Join date: 13 May 2004
Posts: 109
02-08-2005 09:27
From: Jesrad Seraph
Sim FPS with 0 listeners: 996
Sim FPS with 6400 listeners (64 listeners in 100 cubes): 906

Test done in Eclipse.



Would be interesting to the the result for this test with a listener events that checks the msg beeing sent to the event looking for 'LOL' then if found say 'hehe' and doing it on channel 0 with 5-10 avatars around it, talking...oh and maybe as francis suggested place 6400 cubes out instead of 100 :D
and take a snapshot of the Diagnostics Screen...

THEN

Do the same on a higer channel.... and take a snapshot of the Diagnostics Screen...

Anyone who has nothing to do today? :)
_____________________

Nexus Nash
Undercover Linden
Join date: 18 Dec 2002
Posts: 1,084
02-08-2005 14:23
From: Francis Chung
The Lindens seem to have improved the drain on listens, my tests say it's *roughly* two-fold since I last ran a study.

/54/c8/25865/1.html

I'm here in Meins, and the SIM FPS is roughly 6000. Rezzing 512 listeners (like I did last time) drops the sim fps to roughly 2000.

Maybe someone wants to re-run the whole suite of experiments?

But, FYI, 64 listeners in 1 cube is not the same thing as 64 listeners in 64 cubes.

I ran a test earlier in Aleph. 4096 listeners in 64 cubes had negligible effect on the the sim performance, but 4096 listens in 4096 cubes dropped the sim FPS from 1400 to 140. For reference, 4096 default scripts had no measurable performance impact.


Fran hit it right on the money there!

To take this a little further, put the same amount of link_messages in boxes. You'll notice that there is little to no drop in Dilation or FPS in the sim. I believe this is because link_mesage is an event (a push, only triggered on message, it's not activly looking for messages, it just waits) while llListen is more like a rapid active timer (it is always looking for something)

So when you want to do is use link_message whenever possible, not listens. Listens are lsl whores, it's the most "degrading" function\event in LSL. (IE Listens hurt SIMs more then anything else found in scripting) So for the love of god... when ever possible DO NOT USE LISTENS!

EDIT: To add on Tcoz's post with using listens on large data processing. This is good ONLY in a high FPS, low Dilation sim. If you start using listens in high populated areas, you'll find that all that info is waiting in the chat queue.
_____________________
Eloise Pasteur
Curious Individual
Join date: 14 Jul 2004
Posts: 1,952
02-09-2005 02:39
From: Tcoz Bach
What can I say /shrug. I'll use listens when I need lots of data moving fast in my games, because I have proven countless times that they are faster for what I need. All this guessing at the mechanics of SL is pointless. As is apparently two years of coding LSL. What do I know.


Whilst I respect your experience, and you might be entirely correct for your particular application(s) the body of evidence that I am seeing is building up to favour link messages rather than listens where possible.

In test situations they seem to have tested as faster and having less impact on the server... That is not to say they are ideal, and there are certainly things that can't be done any other way.

Without wishing to guess the mechanics of SL, I know that they have changed LSL at least once in a fairly major way whilst I've been here, maybe they have introduced other changes that affect the relative speeds since you ran your tests?
Sitting Lightcloud
Registered User
Join date: 13 May 2004
Posts: 109
02-09-2005 06:53
From: Eloise Pasteur
Whilst I respect your experience, and you might be entirely correct for your particular application(s) the body of evidence that I am seeing is building up to favour link messages rather than listens where possible.


YAY! Doing the Sitting Victory Dance!!! :D


Good choice Eloise Pasteur
_____________________

1 2