I have run the Sculpt operating system once, and in a separate attempt, I successfully ran the Falkon browser on hardware with a Celeron processor. However, when using graphical tools with fade effects or similar features, the system becomes extremely slow. I have experimented with affinity and priority settings to improve performance, but there has been no significant improvement, and the system remains noticeably sluggish and frustrating to use.
Can anyone explain the performance optimization tools available in the Genode framework in detail and provide practical suggestions for improving performance? Or should I simply accept the overhead of the microkernel structure and live with this issue?
In my experience, Falkon’s rendering in Genode is quite fast, to such an extent that I’d rather use that instead of the Debian/FireFox combination in some cases (.e.g. TG feeds).
If you’ve run into a different problem and Falkon somehow renders slowly on recent hardware, I’d say it’s worth filing a ticket or investigating, yes
I can’t locate it right now, but there was a set of changes to Sculpt OS a few weeks back, where Norman tweaked the priority settings for nitpicker, leitzentrale, et al. Not sure it’s relevant to your issue though. In my experience the clean separation of components in Genode does not impede performance in a significant way for common use cases (with the possible exception of the fs stack, which Norman explained might well get optimized in the coming year or two). So if that’s not the case for you it’s worth digging I think.
EDIT: on second reading, it seems your troubles are not related to Falkon at all ? As before, you’re gonna have to be more specific
One thing you could do is leave nano3d running in the background, together with whatever application you’re experiencing sluggishness with… If nano3d keeps running without skipping a beat, you’ll know the problem is specific to the problematic application, and not to Genode and nitpicker…
I don’t have any issues with Genode’s architecture itself. The problem arises when I try to play high-bitrate videos or handle VDI output in HTML5 format in Falkon—it becomes noticeably slow. However, Sculpt OS’s overall performance is satisfactory.
That could well be – I cannot say anything to high-bandwidth playback since I have (very) low bandwidth internet here ^^. That could be one area where ‘comparative benchmarks’ (broadly speaking) could be useful: testing the Linux port of Falkon vs. its Genode port, and see where we lag behind, which area could be prioritized for optimizations, when found to perform much better in Linux… and which ones cannot be easily optimized, when found to not perform well in Linux either.
P.S. make sure to use the “-jemalloc” variant of the Falkon port (I believe there are two variants in the Genode depot). I seem to recall the team mentionning that variant is way more efficient at memory management, it might make a difference for video playback.
