Hi Sebastian,
nice to see you joining the discussion!
I agree that the audio topic certainly deserves more attention. But before re-approaching the topic again, I'd like to stress the need for proper measurement tools that help us to interpret the observed behavior correctly instead relying on assumptions as an aid.
I am thinking of a graphical front end for Genode's tracing feature that
1. can be readily deployed in Sculpt, 2. lists all components and their threads, 3. allows the user to select one or multiple threads from the list, 4. records a trace (by pressing a button), and 5. shows the timeline visually.
I'd love to have such an application at hand, also for many other performance-related developments. And I think that implementing such a tool using Qt seems quite doable.
Sebastian, would you support prioritizing the creation of such a tool over the actual audio work?
Cheers Norman
On 28.12.18 16:02, Sebastian Sumpf wrote:
Hi folks,
here are my three cents: I am very proud of what our team achieved last year. Sculpt is up and running. I did setup the long running version and it has not crashed so far and even if it did, I would see it as a chance to mature Genode. We are humans after all and we do make mistakes, but thanks to the various kernels, and thus different runtime behavior, we are able to trigger bugs very fast - enabling us to fix them quickly. There is more robustness now than I ever imagined.
Otherwise, I see two concurrent paths: 1. POSIX compliance and 2. The Genode way. For porting software (1) is very convenient. But when writing new applications I would like to see more of (2). Most likely we will have to traverse both paths.
Besides doing Java and getting it ready to work properly, I would like to rework the whole audio stack. That would be the third iteration, but sometimes things emerge very slowly. So, if anyone here has experiences with low latency audio, a good book or homepage to read, or just knows how to design this stuff ... let me know.
Even tough Norman's talk got rejected from the FOSDEM organizers, we do enjoy a very successful run at the Microkernel devroom [1], we will organize it this year, where we actually had to reject more talks than I am comfortable with. But there are only eight hours ...
I think by now we have a very small but fine community,
Sebastian
Genode users mailing list users@lists.genode.org https://lists.genode.org/listinfo/users