Fillod Stephane wrote:
Philippe Gerum wrote:
[..]
The last significant change between -00 and -01 is actually the one related to
the fork pressure (others are cosmetic ones aimed at better sharing stuff with
the blackfin port). The patch below against -02 removes it.
Here is the result of tests with version 1.0-02+ (entry.S patch) on e500:
load: ~1 minute ping -f, one run of calibrator chewing 64MiB.
SWITCH without load:
RTH| lat min| lat avg| lat max| lost
RTD| 3660| 3690| 8070| 0 1.0-00
RTD| 4620| 4740| 8730| 0 1.0-01
RTD| 4620| 4740| 8190| 0 1.0-02
RTD| 4650| 4740| 7980| 0 1.0-02+
KLATENCY with load:
RTH|-----lat min|-----lat avg|-----lat max|-overrun|
RTS| -7350| -5715| 6420| 0| 00:03:17 1.0-00
RTS| -6150| -4384| 12180| 0| 00:03:13 1.0-01
RTS| -6150| -4183| 12480| 0| 00:03:38 1.0-02
RTS| -6120| -4447| 8370| 0| 00:03:47 1.0-02+
LATENCY with load:
== Sampling period: 100 us
RTH|-----lat min|-----lat avg|-----lat max|-overrun|
RTS| -6930| -4260| 8700| 0| 00:08:06 1.0-00
RTS| -5670| -4620| 12930| 0| 00:12:39 1.0-01
RTS| -5700| -3750| 11280| 0| 00:06:05 1.0-02
RTS| -5640| -2250| 11310| 0| 00:10:31 1.0-02+
Load is still generated manually, so the pattern may vary: lat avg
is not meaningful. The amplitude would be a much more interesting value.
Anyway, are the generated load comparable?
The klatency is in better shape, however the latency test did not improve.
The 1.0-00 patch should be rerun as a double check.
I'm afraid we're already entering performance fine tuning, while we have
Actually no, such a high variation in latency is the sign of a serious
regression, not just a fine tuning issue. Let's stop there, I'll remove the fork
patch for -03 since it had no positive impact anyway. Thanks.
IMO other higher priority tasks for now. At least, I do :-(
--
Philippe.
_______________________________________________
Xenomai-core mailing list
Xenomai-core@gna.org
https://mail.gna.org/listinfo/xenomai-core
