Recent Features
| |||||||||||||
Interview with Andrew Tanenbaum (LinuxFr.org)Interview with Andrew Tanenbaum (LinuxFr.org)Posted Nov 18, 2011 0:06 UTC (Fri) by marcH (subscriber, #57642)In reply to: Interview with Andrew Tanenbaum (LinuxFr.org) by cmccabe Parent article: Interview with Andrew Tanenbaum (LinuxFr.org)
In most REALLY embedded systems (NOT phones and tablets), performance is not a concern any more.
(Log in to post comments)
Interview with Andrew Tanenbaum (LinuxFr.org) Posted Nov 18, 2011 0:31 UTC (Fri) by cmccabe (guest, #60281) [Link]
Sure, you could put a 3 Ghz CPU in your microwave oven, or vacuum cleaner, or whatever. But now, your power consumption is significantly higher than your competitor's product. And you may now have problems with heat dissipation. And your BOM cost is a few dollars more. So yeah, in some kind of theoretical world with unlimited resources, performance is no longer a concern. But in the real world, it very much is.
The other dimension to performance is realtime guarantees, which Minix can't provide. So overall, I think AT is just whistling in the dark here.
Interview with Andrew Tanenbaum (LinuxFr.org) Posted Nov 18, 2011 8:45 UTC (Fri) by marcH (subscriber, #57642) [Link]
> Sure, you could put a 3 Ghz CPU in your microwave oven or vacuum cleaner
To do what? Un-compress and display a video recipe from Youtube? Please be serious. I'm talking "Mega"/cheap and you answer "Giga"; you are off-topic.
Do you have any idea what is the cost of 1 MIPS in the low range nowadays? Answer: totally negligible compared to the rest of the SoC. And not just in dollars but in terms of power and real estate just as well.
Backtracking: yes, the extra performance cost of a micro-kernel would not make any difference to your microwave oven.
Interview with Andrew Tanenbaum (LinuxFr.org) Posted Nov 19, 2011 3:30 UTC (Sat) by k8to (subscriber, #15413) [Link]
Do people really run microkernels on 8 or 16 bit microcontrollers?
Interview with Andrew Tanenbaum (LinuxFr.org) Posted Dec 1, 2011 10:42 UTC (Thu) by wookey (subscriber, #5501) [Link]
I don't know, but they are all rapidly in the process of getting 32-bit microcontrollers (which is pretty remarkable but seems to be the way the world is headed).
Interview with Andrew Tanenbaum (LinuxFr.org) Posted Nov 18, 2011 14:10 UTC (Fri) by jonth (subscriber, #4008) [Link]
Wrong. Every embedded platform I've ever worked on (for close to two decades) has performance concerns. If it didn't then the platform was, by definition, too big and expensive for its use case and was replaced with something cheaper until we did have performance concerns.
And what's a REALLY embedded system and why don't phones or tablets count as one?
Interview with Andrew Tanenbaum (LinuxFr.org) Posted Nov 18, 2011 15:04 UTC (Fri) by marcH (subscriber, #57642) [Link]
> And what's a REALLY embedded system and why don't phones or tablets count as one?
Thanks for demonstrating what you are missing from the picture.
Performance wise smartphones and tablets are much much closer to PCs than to: microwave ovens, vacuum cleaners, washing machines, traffic lights, alarms, vending machines, most industrial machines, etc.
Interview with Andrew Tanenbaum (LinuxFr.org) Posted Nov 18, 2011 15:17 UTC (Fri) by jonth (subscriber, #4008) [Link]
I'm not missing anything from the picture: I believe you are. An embedded system is a computer system designed for specific control functions within a larger system.
A phone is just as embedded as a washing machine. Mobile devices come in many flavours, not just smartphones. Even within the smartphone segment there is more to the system than just the apps processor. Trust me, the modem processor in any phone is resource limited. And them some.
For _any_ embedded product there should be a processor sized/priced correctly, from a PIC up to an ARM Cortex. If you're using an ARM7 to power a washing machine then fine, you're not resource limited. But you're also using an over-designed processor for the job.
Interview with Andrew Tanenbaum (LinuxFr.org) Posted Nov 22, 2011 21:49 UTC (Tue) by JanC_ (guest, #34940) [Link]
Once an ARM7 SoC costs not significantly more than a 16-bit PIC (that could also do the job), the flexibility & familiarity of the former for average developers (and thus the cost of development) might be more important for a manufacturer...
Interview with Andrew Tanenbaum (LinuxFr.org) Posted Nov 23, 2011 10:42 UTC (Wed) by mpr22 (subscriber, #60784) [Link] How many pins on a typical ARM7 SoC, and how are they laid out? Because the 16-bit microcontroller has under 50, and they're all at the edge of the part.
Interview with Andrew Tanenbaum (LinuxFr.org) Posted Nov 23, 2011 18:03 UTC (Wed) by andreasb (subscriber, #80258) [Link] Why specifically ARM7? ARM7 is pretty old at this point. Cortex-M is what you're looking for when you want ARM microcontrollers and aren't bound by legacy ARM7 software you would have to port. Well okay, they don't have the ARM instruction set as such and understand only Thumb2, but that is in the end mostly just a different encoding of the same instruction set.If you need low pin count packages then boy did NXP announce some interesting chips for you. How about Cortex-M SoCs in SO20 and TTSOP20? I bet you're just going to love the one in a DIP28. Seriously, if all you want is under 50 edge pins (something like QFP48) then there's lots of choice in the 32 bit ARM world.
Interview with Andrew Tanenbaum (LinuxFr.org) Posted Nov 23, 2011 19:19 UTC (Wed) by mpr22 (subscriber, #60784) [Link] Well, I said ARM7 because the person I was replying to did. Looks like some nice parts there, I'm pleasantly impressed.
|
|||||||||||||