QNX RTP Running on iPaq
An anonymous reader sends in: "iPaq just got new gracious looks. QNX microkernel and the gracious Photon micro GUI did wonders to iPaq. Get a sneak preview here. If you are in Boston next week, be sure to drop by Embedded Systems Boston to try your hands at the qPaq... ;)"
Your laughter is borne out of ignorance. Everybody publishes OS times for their OS when running in kernel mode only (which offers zero protection from processes run amok). But CE and EPOC don't run in that mode--you can't on these platforms since they're open and could be running malicious code.
To wit, look at QNX (http://www.qnx.com/products/os/qnxrtos.html#Perfo rmance) and On Time
(http://www.on-time.com/index.html?page=rtk45.htm) . Great numbers, but only
for kernel mode operation.
For protected systems using the MMU, it seems all the big players don't publish numbers. Why? Because this is a tough environment and the numbers look like shit. QNX offers nothing on their site about their Neutrino product performance. Neither do Mentor or Wind River.
Even the RT Linux folks are flaky here. http://www.zentropix.com/support/document/helpdox/ rtai.pdf claims they can
deliver a 4 uS average interrupt response time with 13 uS of jitter,
resulting in 17 uS worst case interrupt response time. This is on a 233 MHz
Pentium II.
Microsoft are claiming 7.5 uS worst case ISR latency on a 90 MHz Pentium II for CE (http://msdn.microsoft.com/library/techart/real_pe rf.htm). We're
seeing similar numbers on a StrongARM platform at a similar clock.
Clearly, CE is probably on par with the QNX/PSOS/VRTX crowd.
So, until Symbian actually publish some numbers on their interrupt performance, we can assume that, like code size, they are merely FUD'ing the industry.
As for the topic at hand, however, it's wonderful to see something like QNX running on iPac, maybe make one worth getting after all ;)