Re: Windows 2000 is The Borg!

On Tue, 16 Sep 2003 17:44:21 -0700, "Joel Kolstad" wrote: [snippage]

There are virus writers out there targeting Linux and Unix systems. It just doesn't make the headlines as much as the Windoze stuff. I just got another notice from Red Hat about a potentially nasty hole in OpenSSH. Second in about a year or two period for OpenSSH. I'm getting used to updating my Linux box every month due to various vulnerabilities. Mark

In programming, you use a NOP to allow time for a device to settle before sampling it. You have to ask yourself how much time is consumed by programming a 0 cycle NOP, and how many times will I have to loop it before, say, 31 nanoseconds is consumed?

Harry C.

Actually, if you coded a 'NOP', it did take a cycle. The problem was the instruction 'pre-fetch' on the K6, bypassed the 'implicit' NOP, in a decrement and loop instruction.

Best Wishes

You can't use instruction timings for delays on any modern CPU; they just aren't that predictable. This used to be recommended on '286 machines where JMP $+2 was used after I/O instructions to allow the slow ISA bus to settle. When '386s came along, you used to see strings of JMPs (in BIOS hard drive code, for example) until people finally wised up and used hardware timing. I've used it on '486s in special cases where I calibrated the delay at the start of the program, but it's a real hassle. Hopeless on a Pentium or later.

Having said all that, instruction delays are still useful in lots of embedded processor applications, where they use old-style CPU cores that have deterministic instruction timings. And there are a _lot_ of embedded systems out there!

Bob Masta dqatechATdaqartaDOTcom D A Q A R T A Data AcQuisition And Real-Time Analysis Shareware from Interstellar Research

Aligning code on cache line boundaries, perhaps. Filling to the end of code pages. There was a time when NOPs were put in as place-holders for self-modified code (another bad idea ;-).

OTOH, why waste the energy dispatching, executing, and completing a NOP. It's wasting system resources (and heat) that could be used for real instructions. Architecturally it doesn't do anything (obviously). Any inference one makes on the function of a NOP (or any other parameter not specified in the architecture) is simply asking for trouble. Add in the OoO nature of modern processors; counting on the execution time of any instruction/loop gets even more dubious.

Unless

one-cycle

On the face of it, the 31nS loop time would seem to need ... 31nS divided by the Planc time of 5.38e-44 seconds or roughly 5.76e35 no-cycle loops. regards john

Please send me the design of your counter. We can make some serious money! ;-)

Nope. It was dropped from the pipe (never saw an execution unit). It took zero clocks, as long as there wasn't an unusual branch around.

Nope. It wasn't dispatched. This isn't unusual these days. Anyone using NOPS, or any other instructions for that matter, for timing loops should be taken out and *shot*. There is a reason we have timers these days.

I was reading an article just yesterday that suggested the way to perform a delay in C was to code up an empty for loop. I suppose someone using that approach also wouldn't know to turn on the compiler's optimizer... which would undoubtedly just remove the loop entirely! (Although it might issue a warning message if the user were lucky...)

All your cursing reveals you're a total moron, not a "hacker". If you're too stupid to get your computer to work with Windows, try Linux. Although I can hardly believe Linux is something you ever heard of.

What the hell is wrong with you people?

Joe

Allow me to take exception to your assertion.

Is it stupidity when the vendor deliberately withholds SECRET information that's NECESSARY for proper operation of equipment?

Is it my stupidity when the computer WON'T EVEN RUN USING FACTORY DEFAULTS?

Sounds more like collusion on the part of the vendors, Microsquat, and who knows who-all else.

And I don't even want to go into how WinXP gets "registered."

Thanks for shar> >