View previous topic :: View next topic |
Author |
Message |
schnake n00b
Joined: 03 Dec 2003 Posts: 49 Location: Siegburg / Germany
|
Posted: Fri Apr 15, 2005 2:25 pm Post subject: |
|
|
hyp0r wrote: | Well, the length of an instruction or OP is always a matter of design and importance, think of atomic instructions or instructions, that should be executed at once to prevent raceconditions or hazards (you know about RAW hazards in cpus).
Low latency is always desired, no doubt, but as soon as tasks are highly preemptible the designer should make sure this does not end in a non-determinstic behaviour or even thrashing. So tasks must be designed for preemption. |
Yes. And judging from articles like http://lwn.net/Articles/128951/ I get the impression that for RTP this is taken care of seriously by serious people.
hyp0r wrote: | Going for low latency and highly interruptable tasks are proven as an advantage as we can see in solutions like qnx, rtai, ibm aix etc... But this gain must be achieved, not used. |
Well said.
Hey, hyp0r, albeit your first response was not sounding particularly inviting I really enjoyed this discussion. Thanks. |
|
Back to top |
|
|
fallow Bodhisattva
Joined: 08 Jan 2004 Posts: 2208 Location: Poland
|
Posted: Fri Apr 15, 2005 2:36 pm Post subject: |
|
|
imho Ingo's RTP in Full RT mode isnt required for a overall/regular destkop using at all . This even can make a lags for _overall_ using. Low-latency destkop mode is much more accurate in this case. If Somebody want to reduce latencies for really low-latencies audio/video working/processing then Full RT mode will be good in this case - but what is number of that people here ? . Of coruse Im talking just about Ingo's RTP. Many of people are interested also in just SCHED_ISO policy for the Staircase as good one for the jack* and etc.
Very often people are too impressed in some new ideas , and without knowledge are beginning a rice to catch the world ( read. patchsets ricing often ) . and often they forgeting that in most cases is good to have balance beetwen everything also beetwen interactivity and performance. many many of examples about this are on this forum
Easy example , You can have better performance of some task(s) with interactivity loss. try to compile something with -19 priority. There is no full of magic in the world
cheers _________________ "Time is a companion that goes with us on a journey. It reminds us to cherish each moment, because it will never come again. What we leave behind is not as important as how we have lived" J-L. Picard |
|
Back to top |
|
|
schnake n00b
Joined: 03 Dec 2003 Posts: 49 Location: Siegburg / Germany
|
Posted: Fri Apr 15, 2005 3:57 pm Post subject: |
|
|
fallow wrote: | imho Ingo's RTP in Full RT mode isnt required for a overall/regular destkop using at all. This even can make a lags for _overall_ using |
Agreed. But I'm running RTP kernels most of the time now, and doing (at least) 8 hrs * 5 days a week "desktop" work on my gentoo box (not using any MS OS for 3 years now). And frankly, I never noticed RTP (or the former "mingoll") affecting "general" performance (neither good nor bad). I just additionally can do low latency audio. But mileage may vary, of course.
fallow wrote: | If Somebody want to reduce latencies for really low-latencies audio/video working/processing then Full RT mode will be good in this case - but what is number of that people here ? |
I do not know, but there are some who want.
fallow wrote: | Very often people are too impressed in some new ideas , and without knowledge are beginning a rice to catch the world ( read. patchsets ricing often ) . and often they forgeting that in most cases is good to have balance beetwen everything also beetwen interactivity and performance. many many of examples about this are on this forum |
Yes. But at least they also learn on their way (although it takes guts to start learning effords that publicly). I think that "ricer-phobia" got sometimes a little hysterical lately. Hey, you are "bodhisattva", your third perfection is patience, isn't it?
But for JadeX: Tiger683 already "got a lot of stick" here and has done remarkably well in responding to that and staying open minded. If the next JadeX indeed will be explicitly advertised as "audio kernel" (and maybe a "second in bootloader" one) I think it should be clear that this is not about posing or "ricing", but really is a win for gentoo instead. |
|
Back to top |
|
|
hyp0r Tux's lil' helper
Joined: 11 Oct 2003 Posts: 139
|
Posted: Fri Apr 15, 2005 4:54 pm Post subject: |
|
|
Quote: | but really is a win for gentoo instead. |
Actually. And I've never seen a distro, that deals with some many system-features as gentoo does, before. Just make it right. :p |
|
Back to top |
|
|
fallow Bodhisattva
Joined: 08 Jan 2004 Posts: 2208 Location: Poland
|
Posted: Fri Apr 15, 2005 10:45 pm Post subject: |
|
|
schnake wrote: |
Yes. But at least they also learn on their way (although it takes guts to start learning effords that publicly). I think that "ricer-phobia" got sometimes a little hysterical lately. Hey, you are "bodhisattva", your third perfection is patience, isn't it?
|
Yeah . Im learned a lot from this . This was also my way for some time
But not all people can draw a conclusions. I hope most of they can.
And Yes , I have patience , I can regulate my anger hehe
cheers. _________________ "Time is a companion that goes with us on a journey. It reminds us to cherish each moment, because it will never come again. What we leave behind is not as important as how we have lived" J-L. Picard |
|
Back to top |
|
|
^PsYcHo n00b
Joined: 27 Feb 2005 Posts: 25
|
Posted: Sun Apr 17, 2005 7:10 am Post subject: |
|
|
Great work Tiger683!!
Using this patchset now. _________________ -m3h s3x11 sCr33n1e- |
|
Back to top |
|
|
|
|
You cannot post new topics in this forum You cannot reply to topics in this forum You cannot edit your posts in this forum You cannot delete your posts in this forum You cannot vote in polls in this forum
|
|