Messages in this thread | | | Date | Thu, 23 Jun 2005 18:46:30 -0700 (PDT) | From | William Weston <> | Subject | Re: [patch] Real-Time Preemption, -RT-2.6.12-rc6-V0.7.48-00 |
| |
On Thu, 23 Jun 2005, William Weston wrote:
> Running -50-17 on the Xeon/HT box, now with the ICH5 USB controller > disabled in BIOS. > > The strangely regular ~200us idle jumps for both CPUs went away. Now I'm > seeing one CPU at a time disappear (often completely absent) in the > traces, with a the maximum wakeup generally hitting between 150us and > 300us. While compiling a kernel and running vlc (which would skip frames > like mad, even with RR scheduling) I was able to get one trace of 2556us. > > There's also BUG warnings for update_out_trace(). I'm not quite sure if > the trace behavior I'm seeing is related to this bug, but judging by the > behavior of vlc, I wouldn't bet on it. > > Since the 2556us trace is quite large all the info is posted at: > > http://sysex.net/testing/2.6.12-RT-V0.7.50-17/xeonht/
OK... Running on -50-17 with all RT debug except CONFIG_WAKEUP_TIMING and CONFIG_LATENCY_TIMING disabled. Max wakeup latency is 59us now, with a couple erroneous values of 2533412143us. VLC performance (streaming from a multicast UDP source) was great with one instance of burnP6 running. Running a second burnP6 along with VLC locks up the box (reproduceable) in about a minute. Running two or more instances of burnP6 without VLC does not result in any odd behavior or noticeably impact GUI performance. I rebuilt with CONFIG_DETECT_SOFTLOCKUP enabled, but couldn't catch a trace.
FWIW, --ww - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
| |