[Re: [PATCH RT 0/6] Linux 3.14.48-rt49-rc1] On 08/08/2015 (Sat 19:23) Steven Rostedt wrote:
> On Sat, 8 Aug 2015 19:10:17 -0400 > Paul Gortmaker <[email protected]> wrote: > > > > So we took 20 IRQs in 5s, or 4/s ; not quite the 1/s minimum, but > > definitely not > > the HZ/s we'd get w/o NOHZ_FULL. Re-running it got consistently 18-20 IRQ > > / 5s. > > OK, so NO_HZ_FULL still isn't great on -rt, but this shows that it > keeps disturbances much lower than without it. Here is several runs on vanilla ("make oldconfig" of the -rt .config) and all else (bootargs, isolation, rcu offload) the same: -------------------------------------------------- root@testbox:/home/paul# ./t LOC: 86341 740 724 708 694 679 663 649 633 619 545 526 512 494 481 463 448 434 420 404 Local timer interrupts LOC: 91349 742 727 708 694 679 663 649 633 619 545 526 512 496 483 465 450 436 423 406 Local timer interrupts LOC: 96353 742 727 710 697 709 665 651 635 622 547 528 514 496 483 465 450 436 423 406 Local timer interrupts root@testbox:/home/paul# ./t LOC: 106188 744 729 712 699 711 667 653 638 624 549 530 514 498 485 467 452 438 425 408 Local timer interrupts LOC: 111194 746 729 712 699 711 667 653 638 624 549 530 516 500 487 469 454 440 427 410 Local timer interrupts LOC: 116199 746 731 714 701 729 669 655 640 626 551 532 516 500 487 469 454 440 427 410 Local timer interrupts root@testbox:/home/paul# ./t LOC: 122860 748 733 716 703 731 669 655 640 626 551 532 518 502 489 471 456 442 429 412 Local timer interrupts LOC: 127866 748 733 716 703 731 671 657 642 628 553 534 520 504 492 473 456 442 429 412 Local timer interrupts LOC: 132872 750 735 718 706 748 671 657 642 628 553 534 520 504 492 473 458 444 431 415 Local timer interrupts root@testbox:/home/paul# cat /proc/version Linux version 3.14.48 (paul@yow-dellw-pg2) (gcc version 4.8.2 (Ubuntu 4.8.2-19ubuntu1) ) #2 SMP PREEMPT Sat Aug 8 19:52:12 EDT 2015 root@testbox:/home/paul# -------------------------------------------------- So 30, 18, and 17 per 5s interval. I'd say that is statisitcally close enough to -rt on this simple test to call -rt on par with vanilla. Things might show a difference with more complicated tests, more cores doing real work etc etc. -- this really is just a simple sanity test though... but at least it is a test that we now can pass! Paul. -- > > Thanks for the report! > > -- Steve > > > > > I specifically did NOT unplug and replug cores to "clean" them of tasks; the > > hotplug code just seems to unstable for that from what I've seen in the > > past, > > and by the looks of the irq counts above, there was no need to either. > > > > The rootfs was basically a ubu 14.10 server install (no X11/gfx) -- not that > > it should matter - so long as other tasks weren't running on the nohz cores. > > > > Paul. > > -- > > > > > > > > Enjoy, > > > > > > -- Steve > > > > -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
