RTOS

Which RTOS? OS-9?

Be advised that for an 8088 thru Pentium type CPU, a truly RTOS is impossible; the memory refresh and timer tick are tied to a *non* maskable interrupt. That means that timing is ranDUMB because that "burp" tick is ranDUMB relative to any program that can be written. Oh yes, the more modern variants have a rather nice added instruction FRDTSC that gives cpu clock cycles since startup. So use FRDTSC to get the count and save it, then CALL some routine, then use FRDTSC again to get the new count and use the difference. Convert to seconds if you know the "exact" frequency the CPU is running at. Yawn. RanDUMBly the "burp" tick will mess with that time difference; the longer the routine takes, the more probable that the "burp" will be caught, until if it is really long, many "burps" will happen and perhaps if the routine is long enough, one may know (within one "burp") how many happen. One cannot directly generate music as syncopated as a person can, due to that ranDUMB re-direction of critical timing routines.

Moral? Use *any* other CPU than the X86/X88 family!

That's the PC architecture not the 808x. Systems have certainly been built w/o that particular feature. Similar features have been implemented on other micros. Also as I recall the early PCs at least the timer wasn't on the NMI although from (I belive the AT) there is a watchdog timer on the NMI. Any conversion of that 'watchdog' occurred after I stopped caring.

Robert

messagenews: snipped-for-privacy@s34g2000cwa.googlegroups.com...

RTOS? OS-9?

On a sunny day (Mon, 1 Jan 2007 00:24:12 -0500) it happened Robert Adsett wrote in :

yes correct. Also talkin gabot real time without ,mentionin ga latency makes little sense. As the ultimate minimum time would be a processors clock cycle or two, so nothing is immediate. For the x86 platform several solutions are availabe:

bullshit neither are processor features, both are maskable , even controllable.

or RTFM.

Bye. Jasen

You recall incorrectly; the timer was a NMI and i could scan the schematics from the IBM Technical Reference manual AKA Personal Computer Hardware Reference Library, First Edition (August 1981). And it has been an NMI from then to now and into the future.

That 30 microseconds is meaningless and can be considered incorrect, as the CPU clock speed was not mentioned. Using my Pentium MMX-233 computer, i can do a multiply in 42uSec, and the estimated call overhead is 3uSec.

Refer to the IBM Technical Reference manual, First Edition (August 1981).

On a sunny day (Tue, 02 Jan 2007 05:28:39 GMT) it happened Robert Baer wrote in :

Where do you buy a < 1GHz PC these days... Also it does clearly not depend so much on clock speed, but on the software system (kernel) itself, one clock at 1GHz = 1ns. So the PC has at least 30 000 clock cycles to do things.

But does it run a webserver, mail server, name server, graphical user interface, webbrowser, editor, mp3player, TV recording, USB interfaces, etc at the same time?

On a sunny day (Tue, 02 Jan 2007 05:30:08 GMT) it happened Robert Baer wrote in :

Some things changed in the last 26 years :-)

Hmm, I thought it went to the PIC. I yield to your reference though, I no longer have access to a copy to check.

Still leaves the basic point that hooking anything in particular to the NMI is a PC architecture decision not an 808x feature.

Robert

interface,

time?

Do not act stupid.

The NMI hardware/software structure has been faithfully copied to present PCs.

Yes, if anyone has designed and built different hardware - hopefully to eliminate that drawback.

Good grief, Gert! The 80186/8 were used in the embedded industry for decades! ...as has been the 8086 and its derivatives. The IBM 5150 PC x86.

That multiply timing is not believable, are you sure you don't mean 0.42 uS? I have seen 1970 class computers that could 32 bit by 32 bit multiply in 22 clock ticks.

In the "PC" platform MNI is tied to keyboard controller 8048 for CTRL-ALT-DEL, not timer tick it is on IRQ 2.

On a sunny day (Thu, 04 Jan 2007 02:43:05 GMT) it happened Robert Baer wrote in :

interface,

time?

You are defeated :-) That is what the Novell real times Linux can do and more. Good reason to leave MS DOOZ

On a sunny day (Thu, 04 Jan 2007 02:44:01 GMT) it happened Robert Baer wrote in :

I am not so sure. First there is a totally different memory structure, and refresh is likely handled by the chipset, along with many other things. So your argument NMI is used for refresh is no likely longer valid.

Linux has a nice utility to watch the interrupts happening: ftp://panteltje.com/pub/xosview.gif My mouse is on 12