tag:blogger.com,1999:blog-6469704299235308349.post7930416397144157874..comments2024-03-28T15:50:13.644+11:00Comments on -ck hacking: Towards Transparent CPU Schedulingckhttp://www.blogger.com/profile/02904761195451530213noreply@blogger.comBlogger9125tag:blogger.com,1999:blog-6469704299235308349.post-81717284818240888442012-01-28T13:25:12.195+11:002012-01-28T13:25:12.195+11:00I am currently in the midst of developing a very l...I am currently in the midst of developing a very low level python implementation of an adaptive predictive scheduling model. I'm importing a bunch of existing PHP modules. The problem is, JavaScript is necessary for the Python shell. Is there any way you guys know of to switch the parameter to accept said PHP modularity?Ryan Bastersennoreply@blogger.comtag:blogger.com,1999:blog-6469704299235308349.post-91792008330433923142012-01-22T10:15:07.387+11:002012-01-22T10:15:07.387+11:00the tendency of the predictive scheduling models t...the tendency of the predictive scheduling models to superimpose data overlay onto the existing user space is key. i agree with the methodology as such, but as with all cs projects i of course have a lot of input.mtnoreply@blogger.comtag:blogger.com,1999:blog-6469704299235308349.post-25659455699906824542012-01-14T14:05:17.791+11:002012-01-14T14:05:17.791+11:00One thing I am glad of with CFS and BFS over the O...One thing I am glad of with CFS and BFS over the O(1) is the nice 19 not rescheduling too much. Even for just that the newer schedulers are worth it.Samuel Orrhttps://www.blogger.com/profile/10739116231128995436noreply@blogger.comtag:blogger.com,1999:blog-6469704299235308349.post-2058390694528887792012-01-12T07:50:16.076+11:002012-01-12T07:50:16.076+11:00The docs EXTENSIVELY describe how it chooses a CPU...The docs EXTENSIVELY describe how it chooses a CPU to schedule a task on. When a task wakes up, it always tries to schedule on the same CPU it last ran on. Bouncing occurs when the last CPU it ran on is not idle, and another CPU is idle.ckhttps://www.blogger.com/profile/02904761195451530213noreply@blogger.comtag:blogger.com,1999:blog-6469704299235308349.post-51339209121216020702012-01-12T02:15:32.726+11:002012-01-12T02:15:32.726+11:00Quoting from the docs:
"trying that [last] C...Quoting from the docs:<br /><br />"trying that [last] CPU first when [did you mean then?] looking for an idle CPU to use the next time it's scheduled."<br /><br />What if the last CPU goes idle as a result? How do you protect from bouncing tasks? Can you add a few words about the bouncing to the docs please?Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-6469704299235308349.post-25827151913799298852012-01-10T01:11:29.765+11:002012-01-10T01:11:29.765+11:00That's pretty much correct from whatever angle...That's pretty much correct from whatever angle you look at it. BFS barely has even the concept of process migration since no CPU has ownership of any process. Everything belongs to the global queue.ckhttps://www.blogger.com/profile/02904761195451530213noreply@blogger.comtag:blogger.com,1999:blog-6469704299235308349.post-46097309117154691042012-01-10T01:00:38.325+11:002012-01-10T01:00:38.325+11:00I think this guy correctly found the reason why BF...I think this guy correctly found the reason why BFS outperforms CFS: more aggressive process migrations permits faster balancing and better CPU loading. <br /><br />Is this not true with whichever version?Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-6469704299235308349.post-14636961594076136472012-01-09T17:33:05.567+11:002012-01-09T17:33:05.567+11:00Indeed they did use quite an old BFS by today'...Indeed they did use quite an old BFS by today's standards. However, unlike the mainline scheduler, BFS has not changed that drastically since then. It has had nanosecond acounting added throughout, and the soft throughput CPU affinity code has been simplified even further than it was back then - unlike mainline which has gotten more and more complex. So it is likely going to behave virtually the same as the older version.ckhttps://www.blogger.com/profile/02904761195451530213noreply@blogger.comtag:blogger.com,1999:blog-6469704299235308349.post-48913974863187243842012-01-09T12:50:54.367+11:002012-01-09T12:50:54.367+11:00The versions they chose are outdated. So much work...The versions they chose are outdated. So much work into this thesis, but they're slacking on using a recent kernel? Seriously, that's messed up. It's a thesis, so they couldn't have started this more than a year ago, but they used 2.6.32 from late 2009? Wtf?Anonymoushttps://www.blogger.com/profile/11469174621439712081noreply@blogger.com