Messages in this thread | | | Date | Tue, 8 Dec 1998 02:54:40 +0100 (CET) | From | Rik van Riel <> | Subject | [PATCH *] scheduler bigpatch 2.1.131-2 |
| |
Hi,
I just released a new version of the scheduler bigpatch which features two changes:
- don't recalculate priority when we were off of the runqueue for just one jiffie (3 on Alpha). This is done because often-sleeping but still CPU-hungry processes got a bit too much CPU time and you've missed pay time anyway when you just slept over a jiffie increment We now should be a bit fairer (read harsher) agaist those 10%-30% CPU using processes - after a process has been on the run queue for LOAD_AVG seconds, it is effectively put in a round robin scheduling mode so speed/throughput is increased slightly. This works because your CPU pigs are usually interrupted by a key press from your editor which pollutes (far) less than half of L2 and possibly not even all of L1 cache. It would be a real shame to switch to the next CPU pig before the timeslice is over -- yet the normal scheduler does this :)
You can get the patch from my home page. I am very interested in results versus the normal scheduler and versus my old scheduler patch. I am even more interested in results from larger multi- user machines with loads of interactive processes running -- how do response times change, what else is happening, is there a point where the scheduler 'breaks down', etc???
regards,
Rik -- the flu hits, the flu hits, the flu hits -- MORE +-------------------------------------------------------------------+ | Linux memory management tour guide. H.H.vanRiel@phys.uu.nl | | Scouting Vries cubscout leader. http://www.phys.uu.nl/~riel/ | +-------------------------------------------------------------------+
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/
| |