The v2 Rotary Interactivity Favor Scheduler

[3766691]

What? What's "noop" and "deadline"?

These are io scheduler actually.

[3766691]

What? What's "noop" and "deadline"?

Please add
p->policy = SCHED_IDLE;
in the init_idle() function.
This is a bug. :-(

[3766691]

Now RIFS v3 will have two RC version first.
RC-1 is now avaliable.
RC-2 is still being written.

http://code.google.com/p/rifs-schedu...q=#makechanges

Chen

[elmariachi]

using it thus far.
the most intensive task I do in a normal day is compiling the kernel -j2 and watching a 720p/1080p movie. all works great with this. Maybe better than BFS, but this stuff is so subjective... BFS had hiccups sometimes.

Eagerly waiting for a cleaner patch

[darkbasic]

It would be interesting to test it with a 16 tracks recording with the realtime kernel. Maybe I will do it in the future, I'm quite busy right now.

[3766691]

RC2 have been released. :-)
http://code.google.com/p/rifs-schedu...q=#makechanges

[3766691]

using it thus far.
the most intensive task I do in a normal day is compiling the kernel -j2 and watching a 720p/1080p movie. all works great with this. Maybe better than BFS, but this stuff is so subjective... BFS had hiccups sometimes.

Eagerly waiting for a cleaner patch

Yeah, you have realised that there are so many back up files in the patch right?

[elmariachi]

Yes, but why would we need them? We're all big boys, we now how to revert things

[3766691]

Yes, but why would we need them? We're all big boys, we now how to revert things

I want a piece of spare time to make it clean. Unfortunately I am going to have examination at the end of June so I 'm not going to use my lovely box, I use my mobilephone instead. Also I have an idea on load-balancing but I don't want to implement it now.

Could you help me to try these two RC versions of RIFS V3 and tell me which one works better or they are worse than RIFS V2 ? Thanks for your help very much!


Algorithm of RC1: MLFQ(pure, all pick next case)+Bounding Rotary : O(1) Time complexity

Algorithm of RC2: MLFQ(static timeslice+half timeslice priority raise, preempting pick next case)+EEVDF(timeslice ending pick next case[Actually for preventing starvation]) : O(1) Time complexity

[elmariachi]

Sorry.. why is your newest file RC1? is that different from the one released earlier? or is it mislabeled and it's in fact RC2?