Re: [patch 0/9] sched: use runnable load avg in balance

2013-06-10 Thread Alex Shi
On 06/10/2013 11:01 PM, Peter Zijlstra wrote: > On Fri, Jun 07, 2013 at 03:20:43PM +0800, Alex Shi wrote: >> Peter&Ingo: > > I'm tempted to apply 1-7; although I want to look them over in detail > once more; unfortunately (for you) I'm on PTO this week so it'll have to > be early next week. I am

Re: [patch 0/9] sched: use runnable load avg in balance

2013-06-10 Thread Peter Zijlstra
On Fri, Jun 07, 2013 at 03:20:43PM +0800, Alex Shi wrote: > Peter&Ingo: I'm tempted to apply 1-7; although I want to look them over in detail once more; unfortunately (for you) I'm on PTO this week so it'll have to be early next week. As to patch 8, I really need/want buy-in from Paul on that; it

Re: [patch 0/9] sched: use runnable load avg in balance

2013-06-09 Thread Alex Shi
Ingo & Peter: How are you? I know you are very very busy since services for wide key kernel parts. So hope this email don't disturb you. :) PJT's runnable load was introduced into kernel from 3.8 kernel. But besides increase longer code path, it does nothing help now. This patch try to enable th

Re: [patch 0/9] sched: use runnable load avg in balance

2013-06-07 Thread Alex Shi
On 06/07/2013 03:20 PM, Alex Shi wrote: > Peter&Ingo: > > v8: add a marco div64_ul and used it in task_h_load() > v7: rebasing on tip/sched/core tree. Peter & Ingo: IMHO, if the patch set missed in 3.11 kernel. We will lose the following widely benefit on many benchmarks, hackbench, pgbench, sys

[patch 0/9] sched: use runnable load avg in balance

2013-06-07 Thread Alex Shi
Peter&Ingo: v8: add a marco div64_ul and used it in task_h_load() v7: rebasing on tip/sched/core tree. I tested on Intel core2, NHM, SNB, IVB, 2 and 4 sockets machines with benchmark kbuild, aim7, dbench, tbench, hackbench, oltp, and netperf loopback etc. On SNB EP 4 sockets machine, the hackbe