Cool VL Viewer forum

When entering new regions with Cool VL Viewer 1.25, I experience very high CPU load, with the viewer taking up an average of 90-95% CPU usage on both cores and a framerate reduction of one half. After everything has loaded, the CPU usage levels off and drops to what it was in 1.23/1.24

Please, provide the info as listed in the rules for this forum. I can't investigate if I don't even know what OS and OS version you are using...

Sorry, forgot about that:

I'm myself using Linux and never encountered such an issue...

You didn't provide the whole copy of the about screen (viewer and server version), but the only case I know of that would result in such an issue would be the use of a viewer complied by yourself and using a libcurl version that is not using c-ares (which provides non-blocking DNS calls). The libcurl version provided together with the viewer does use c-ares and doesn't suffer from such blocking calls during DNS lookups (which occur much more often with v1.25 than with v1.23 if you are using the "HTTP texture loading" advanced render setting in the former).

Here's the full one:

I see nothing wrong with your setting.

The only change between v1.24 and v1.25 is that v1.25 (and the latest v1.23 release too) got the objects caching fixed, which results in a much faster rezzing speed for region you already visited once in the past and therefore in a much higher number of textures decoding per second on login and TP in these regions (since the viewer doesn't have to wait for the server to send the objects parameters among which are the texture ids).

If your viewer is configured to "Run Multiple Thread" (Advanced -> Rendering menu), then it will spawn a new thread for each texture decoding, and Linux will spread the load over the free cores of your CPU. Here, with a quad core (Q6600 @ 3.4GHz), I don't see any slow down in the frame rate, and the CPU load stays low (around 30-40%), but with a slower dual core, you might indeed get a 100% CPU load during initial texture decoding.

The efficiency of the Linux scheduler and the response time of the system to events during high CPU loads may also vary from one distro to another, depending on what preemptive mode is used (mine is configured for "Voluntary Kernel Preemption"), and the timer frequency (mine is set for 300Hz).

I'll have to have a look at that, then. My CPU is a 2.6ghz dual-core, so it probably will have higher CPU usage than yours. Do you know how I can check which preemptive mode my kernel is using? And does setting it require a kernel recompile?

You should be able to find the config file for your kernel in the /boot directory. Look for the following settings:
CONFIG_PREEMPT_NONE ( = No Forced Preemption)
CONFIG_PREEMPT_VOLUNTARY ( = Voluntary Kernel Preemption)
CONFIG_PREEMPT ( = Preemptible Kernel)

CONFIG_PREEMPT_VOLUNTARY is a good compromise.

Yes.