| --- Log opened Fri Aug 28 00:00:19 2015 | ||
| maxpaln | howdy all - been a while | 07:56 |
|---|---|---|
| maxpaln | Unfortunately, I won't be able to make orconf this year - and CERN would have been a cool place to visit. Our new baby is consuming far too much time already - have a good one though. | 07:58 |
| maxpaln | I didn't jump online just to tell you that though :-) I have a question about context switching - one of my customers using the MOR1KX is seeing occassional instances where context switching takes a long time, this is resulting in a timeout further down their sw chain. The version of MOR1KX they are using is probably around 1 year old (the last firmware update I gave them). I remember a bunch | 08:01 |
| maxpaln | of work was done last year on areas around this (to support multi-cores IIRC). I am wondering if we could expect to see any difference in the way context switching occurs if I updated to the latest MOR1KX code? | 08:01 |
| maxpaln | We were also wondering if increasing the size of the instruction cache would help - we have spare memory in the device but if there is no advantage to increasing the instruction cache depth we won't bother investigating. | 08:02 |
| juliusb | maxpaln: sorry to hear you can't make it along this year | 09:13 |
| juliusb | there's always next year, though! | 09:13 |
| juliusb | and regarding mor1kx context switching, can you quantify how long in terms of clocks or relative to the amount of time it normally takes (ie. it's taking 4 times as long) | 09:15 |
| juliusb | and what is the context switch code doing? | 09:15 |
| juliusb | purely switching state? not polling anything that could make it stall for a while? | 09:16 |
| maxpaln | both excellent questions - I will try and get more details but this has arrived to me through a few intermediate engineers so the detail has been lost. The engineer who reported it should have some good detail though - let me see if I can get more information... | 09:17 |
| juliusb | cool. I don't know how much help I can be, btw, having not inspected the code for a while, but being able to pinpoint the source of the latency might give you an idea of where you might benefit from any changes that have subsequently gone in | 09:28 |
| -!- Netsplit *.net <-> *.split quits: rokka, NCommander, heroux | 10:27 | |
| -!- Netsplit *.net <-> *.split quits: p1oooop, olofk, trevorman | 10:27 | |
| -!- Netsplit over, joins: trevorman, p1oooop, rokka | 10:29 | |
| -!- Netsplit over, joins: olofk, heroux, NCommander | 10:30 | |
| -!- Netsplit *.net <-> *.split quits: olofk, NCommander, heroux | 10:30 | |
| -!- Netsplit over, joins: olofk, heroux, NCommander | 10:32 | |
| --- Log closed Sat Aug 29 00:00:20 2015 | ||
Generated by irclog2html.py 2.15.2 by Marius Gedminas - find it at mg.pov.lt!