-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 On Monday, 2009-11-02 at 10:25 -0000, G T Smith wrote:
Carlos E. R. wrote:
On Saturday, 2009-10-31 at 11:12 -0000, G T Smith wrote:
The CMOS clock was not part of the original PC, it was an add-on card invented later. I think they still use the same chip, or an equivalent. And read access to it was (is?) slow, so it can not be used for timing operations.
The IBM PC (1981) did not have a CMOS hardware clock, but from the PC/AT (1984) onwards it was standard. The situation with intel based non-IBM kit (it did exist in 1981) is a bit more complex (but I do remember powering up such early machines without needing to enter the date). The IBM machines were technically solid (in more ways than one :-) ), but not necessarily bleeding edge for the time.
I think that the AT added the external clock card, but putting the chip in the mainboard instead (the chip had a small cmos memory maintained with a small battery... used for keeping bios config as well). But not all machines at the time had that, I think, I remember that the commands date and time where typically at the start of the autoexec.bat, and most people ignored it, which meant that all files were dated Jan 1 1980. ...
As for clock speed alteration I do recall people telling me they used sequences of Null OP instructions to ensure accurate timing in some circumstances (!??)... Depends really what you mean by changing the clock speed, changing the hardware clock speed itself could have unfortunate consequences, changing the translation between cpu or clock ticks and calender increments is another matter. I assume the latter is meant, but I cannot find any direct reference to this in the ntp (the program) docs...
Yes, I remember using that timing method. It worked very well... in the eighties. Turbo Pascal used it in the "delay" routine. Right at program start it measured one tick of the timer in noops loops, and recorded a constant for time adjustment. At some version (6? 7? around 1998?) the variable that held the initial count overflowed, or wrapped back, causing many problems in problems already written. There were patches for the run time library, and also a method to patch released binaries. I guess that borland c had a similar problem, but I dunno. But it surprises me a lot to think that people may be using a method of that kind to time anything nowdays. Perhaps for microseconds :-? - -- Cheers, Carlos E. R. -----BEGIN PGP SIGNATURE----- Version: GnuPG v2.0.9 (GNU/Linux) iEYEARECAAYFAkrvU1sACgkQtTMYHG2NR9Vb4QCfVMaz+yzBrSjqDJV/jPepBO7h pbwAn3nuAcquwlnkhR8iwVIvtqGEe58o =Rt6Q -----END PGP SIGNATURE----- -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org