[LAU] difference between realtime-kernel and low-latency-kernel?

[Tracey Hytry]

Porl, I can understand where you're coming from on this.

I have a machine here that I play around with mostly to understand the debian system on.   Though most of the time I use fedora-ccrma on my my main machine, I've taken some care to get sound working well on this 2nd(debian based) machine.

The machine has both 64studio and ubuntu(not studio, but the ubuntu low-latency kernel).  For me, 64studio is a fantastic audio distribution with very low latency capabilities.  But no matter how hard I've tried over the past few years, ubuntu seems like a lost cause.  I've also tried the kernel from 64studio on ubuntu and got much better results running audio apps in ubuntu, but I don't want to mess up ubuntu too much by dragging a whole lot of extras over just to make up for ubuntu's audio problems.

I haven't been following the ubuntu/ubuntu-studio thing too well;  but I thought that the problems are that the studio folks know what they're doing(for audio work),  while the folks who run the distro have little interest in specifically fine tuning the system for audio work.  Now, it's not that just ubuntu has this problem;  if you want to look it up you can find Ingo's rants on how the redhat/fedora people insist on keeping the disk system's settings so conservative.

It seems until we see all of the low latency stuff into the mainstream kernels, and the redhat and ubuntus give us better tuning options(especially on the the desktop);  that audio distros such as 64studio or fedora(with Fernando/Ingo's patched kernels, tweaked pam, and other suggestions) are our best bet.  I'm not trying to say that those two are the only ones(there is a gentoo system here that does as well as 64studio and fedora/ccrmart (but then, if you run gentoo it's understood that you tune your system the way you want it(so ok, gentoo has it's problems also(like not updating the pam system to use rtlimits for close to a year and setting it far behind other distros)))).  And no, I haven't tried the suse audio stuff(mostly because I dislike the way yast won't let me change things on my own or the way I want to).

I could go on with this rant, but some troll from some mainstream computer magazine might think it's contents(taken out of perspective of course) could be a great anti-linux story.  The truth is that us audio folks just want the best, and a properly tuned linux system is fantastic for low latency audio work. 

The best I can say is use a custom_low_latency_kernel, tune up the pam system, use noatime on your disks, tell the window manager not to poll the i/o system, and use things like Rui's rtirq if needed.

Tracey.