| 1 |
> Were I you, I would consider: |
| 2 |
> |
| 3 |
> - If keeping X, switching to the absolute most minimal wm possible |
| 4 |
> (twm, ratpoison, ion), to see what effect that had. |
| 5 |
> - If downstepping from X, investigating what programs run under |
| 6 |
> DirectFB and seeing what effect that had. |
| 7 |
> - If going cold-turkey off X, seeing how far you get with the |
| 8 |
> command-line and ncurses programs. |
| 9 |
|
| 10 |
I would also add the following: remoting X. X is a hog, as Holly said, but |
| 11 |
there's no reason the X server would need to run on the same box as the |
| 12 |
ongoing recording session. |
| 13 |
|
| 14 |
Running two systems, one running X and handling the gui operations, and one |
| 15 |
running your audio apps, might provide enough of the separation to reduce |
| 16 |
the latency on the audio box. Of course the two cards should probably be |
| 17 |
connected with at least a 100mb Ethernet connection (to eliminate the |
| 18 |
overhead of dealing with the network conversations for X). |
| 19 |
|
| 20 |
Another possibility might be your choice of filesystems (assuming the |
| 21 |
recordings are going to disk). Different filesystems have inherent latency |
| 22 |
based upon their design - journaling adds overhead, btree maintenance in |
| 23 |
reiser adds overhead, etc. Just using a simple ext2 filesystem for the |
| 24 |
initial recording followed by backups to a modern filesystem may have a |
| 25 |
measurable impact. |
| 26 |
|
| 27 |
Going back to X, it is a hog both in cpu cycles and in memory; you mention |
| 28 |
having an amd64 but no quotes on memory. My assumption is that such a |
| 29 |
system has a big chunk of memory, but I've learned what happens when one |
| 30 |
assumes. Obviously a lack of sufficient memory can cause you some swapping |
| 31 |
issues whether you were aware of it or not. |
| 32 |
|
| 33 |
|
| 34 |
-- |
| 35 |
gentoo-user@g.o mailing list |
Archives