2019 SL User Groups 23/1: Simulator User Group

Whimberly; Inara Pey, April 2019, on FlickrWhimberlyblog post

Server Deployments

At the time of writing, a deployment thread had yet to be posted (in fact, it’s not entirely clear what is happening with the weekly release threads). However:

  • There was no SLS (Main) channel deployment on Tuesday, June 4th, leaving it on server maintenance package #19 19.05.17.527341.
  • On Wednesday, June 5th, the Magnum and LeTigre RC channels should be updated to server maintenance package #19 19.05.24.527547, bringing them to parity with BlueSteel.

SL Viewer

There have been no LL viewer updates at the start of the week, leaving the pipelines as follows:

  • Current Release version 6.2.2.527338, formerly the Teranino RC viewer, promoted May 22 – No Change.
  • Release channel cohorts (please see my notes on manually installing RC viewer versions if you wish to install any release candidate(s) yourself):
  • Project viewers:
  • Linux Spur viewer, version 5.0.9.329906, dated November 17, 2017 and promoted to release status 29 November – offered pending a Linux version of the Alex Ivy viewer code.
  • Obsolete platform viewer, version 3.7.28.300847, May 8, 2015 – provided for users on Windows XP and OS X versions below 10.7.

In Brief

Script Processing Issues

As I’ve recently reported, there have been numerous reports of script run time issues, with some reporting that problems only started occurring following the roll-back on April 18th, 2019. BUG-226851 (and more recently BUG-227099) outlines some of the problems. Some regions experiencing the issue have been cloned to Aditi (the beta grid) where they are running fine, and the Lab is currently looking at script scheduling on Agni (the Main grid).

There is a forum thread on the impact of idle scripts (i.e. scripts that are scheduled, but waiting on an event) on full regions (primarily Mainland). In particular, the finding suggest that limiting in-world scripts to 1 per 10LI (2250 script for a Full region) might help pending improvements – although this falls woefully short of shopping-related regions.

Commenting on the situation, Simon Linden states:

We’re actively looking at script scheduling now and these odd cases are on the list to sort out. There hasn’t been a big magic fix yet but it’s getting attention.

Rider Linden added:

And, yes. Idle scripts are not really idle… they can drag down sim performance even if they are not actually doing anything. Right now llListen() is a major culprit (but I’m hoping we can clean that up in the near future) … [and] all idling is not equal. Some events are more expensive than others to wait on.

EEP Issues

As per my May 30th CCUG update, it’s been noted that a recent deployment has resulted in rendering issues for the Environment Enhancement Project (EEP) – see BUG-227100 as an example.

The EEP RC viewer is due for an update, and it is hoped that these will include shader fixes that should correct the problems in that viewer. However, in terms of the simulator problems that are causing issues for those on non-EEP viewers, Rider Linden commented:

We have had some confusion on our end. When we were chasing the teleport bug a couple fixes were accidentally dropped from the simulator. We’ve been sorting those out as we find them.

Other Performance Impacts

  • It’s long been recognised that teleports – particularly teleports involving mesh-heavy avatars into a region can adverse affect performance (see BUG-8496).
    • Beq Janus has been looking at this, and can show – via a a comment posted to the above Jira, that the effect “ripples through” simulator stats, including hitting script stats.
    • She estimates that overall, the arrival of a mesh avatar in a region can take around 6 seconds to ripple through the region (and impacting users within the region).
    • It’s currently estimate by both Beq and Lou Netizen that female avatars carry on average 70 scripts, and mail avatars an average of 68 (see here for more).
  • Beq has also been investigate another performance impact – that of the viewer flooding the server with GroupProfileRequest at log-in due to persistent notifications see BUG-227094). She is currently working on a patch for contribution to LL to resolve this issue.

2 thoughts on “2019 SL User Groups 23/1: Simulator User Group

  1. Around 70 scripts for the average avatar?

    There are different sorts of average, mean, mode, or median, and it feels astonishingly high, but then I remembered how, in the past, for some features one-script-per-prim was common to avoid deliberate pauses, imposed by Linden Lab in the coding, on one-script-for-everything methods. Prim-based hair could be use a lot of scripts to do a colour change or adjust size.

    I can’t remember the details. I know it got changed. Older content, still in the marketplace, didn’t get updated. I tried to find the functions that had changed, but, but references look to have vanished, and I have always had problems finding the right search term for info on the SL website. There does seem to be a lack of info on the Marketplace about such things as script counts, streaming weights, and display weights, which can hammer a teleport.

    Like

    1. If you take into consideration the scripting in the mesh body, the potential for mesh clothing items to be scripted (Auto-alpha systems, etc), scripted attachments – particularly scripted attachments, and 70 as an average

      Like

Comments are closed.