Timer units are the same for the S/360 and the S/370. From GA22-7000-4 : "In each case, the frequency is adjusted to give counting at 300 cycles per second in bit position 23. The cycle of the timer is approximately 15.5 hours. "
-- Shmuel (Seymour J.) Metz http://mason.gmu.edu/~smetz3 ________________________________________ From: IBM Mainframe Discussion List <[email protected]> on behalf of Bill Godfrey <[email protected]> Sent: Sunday, February 25, 2018 3:41 PM To: [email protected] Subject: Re: Timer Unis (was: ... time change ...) On Fri, 23 Feb 2018 23:32:07 -0500, Tony Harminc wrote: >On 23 February 2018 at 19:16, Paul Gilmartin < >[email protected]> wrote: > >> But now I'm confused. The description of TIMER says: >> For TUINTVL, the address is a fullword containing the time interval. >> The time interval is presented as an unsigned 32-bit binary number; >> the low-order bit has a value of one timer unit (approximately >26.04166 >> microseconds). >> >> That has to be right, or else programmers would have noticed. >> >> And a less official source (but agreeing with Tony) says: >> >http://secure-web.cisco.com/1Z8ALEZ494MZ0AYNZseGKLTpxMpIbkNznfeV9qOsGHKl-Uaw9Pn03mrEFQSjmMZbAm580pDXaNv6yz24MwahOT-0zyMY9MdAK5eXIfp-4hN4hvfEanA6Ctt0KyJpT6RyanLaq156xXkFcLlsqlcTWUlejaqoTwg2TKFmF5jCUSMOFMYNUHdqAWpYOuvdQMg5qAh5mRT_DVnZtNGbHmutRT5YIS1uU9JG08Fw8NW77SFHbG7OsKHdJCzz7ZfRNyMtEMaNgudAkfVJtauNsNE5QG7prB9Xyff4keDg_DrwqyMuGnioWwqrZHObWt5UKxYxHdJiz7wPiHkoKr7Qj6ixvkRkW7YZl0h9DYf2Jjn8usaWRNJZZ0BbprDmuzL2ACmPCPezTLWi2vKPP-J_NSr_jOIQNBJxVPiJtr8KZpKVu89mJFXOcxO7wim5lLsNwXWP5/http%3A%2F%2Fwww10.dict.cc%2Fwp_examples.php%3Flp_id%3D1%26lang%3Den%26s%3Dinterval%2520timer >> >> IBM System/360 architecture >> If the interval timer feature is installed, the processor decrements >the word at >> location 80 ('50'X) at regular intervals; the architecture does not >specify the interval >> but does require that value subtracted make it appear as though 1 >were subtracted >> from bit 23 300 times per second. > >I remembered it as bit 23 when I posted, and I first assumed I had just got >it wrong. But the last S/370 POO (and the earliest for S/360 - both on >Bitsavers) do say, with slightly different wording, that it is bit 23 that >is effectively counted down at 300 Hz. > >That first S/360 POO also has this table: > > BIT >POSITION FREQUENCY RESOLUTION > 23 300 cps 3.33 ms > 24 600 cps 1.67 ms > 25 1.2 kc 833 µS > 26 2.4 kc 417 µS > 27 4.8 kc 208 µS > 28 9.6 kc 104 µS > 29 19.2 kc 52 µS > 30 38.4 kc 26 µS > 31 76.8 kc 13 µS > >> But bit 23 must have 256 times the value of the low-order bit, and >> 26.04167 * 256 = 6666.66752 >> ... which is 1/150 second, not 1/300 sec. > >I'm wondering if the problem relates to the timer itself being a signed >integer (an external interrupt becomes pending when the timer goes from >positive to negative, but it keeps on counting), but the TU arguments to >things like STIMER are said to be unsigned. But that's perhaps more of a >clue than an explanation. So I dunno. > >> Was the interval timer the only source of time-of-day on those early >models? > >Yes. The TOD clock was new with S/370. (Well, one never knows about the >360/85, 91, and 195 unless one has had hands-on experience.) > >> If so, the External interrupt handler must reload its register before >another >> tick is lost -- easy enough at power frequencies, challenging for a higner >> resolution interval timer. > >Heh... They thought of that early on. I can't find it in the POO at the >moment, but it is no accident that the fullwords at locations 4C and 54 are >"reserved" by both S/360 hardware and software. Then an MVC for length 8 >from address 50 to address 4C both saves the current value and sets a new >one. Since the timer is not (visibly) updated during instruction execution, >there is no loss of information. > If the timer unit was defined as the same unit as the System 360 timer word, 13.02083 microseconds, where bit 23 is 1/300 second, then 24 hours (86,400 billion microseconds) would be about 6.6 billion timer units which is too large for a 32-bit unsigned number. If a timer unit is defined as double the System 360 timer unit, 26.04166 microseconds, then 24 hours of timer units would be about 3.3 billion timer units, which fits in a 32-bit unsigned number. I think that's the reason why the timer unit is not the same as the System 360 timer word unit. The 360 timer word unit could not represent 24 hours in 32 bits. What got me thinking about that was an October 2014 post by you, Tony, where you said "Possibly they wanted to be able to represent more than the 7.7ish hours that the positive number range of the architected format provides." Bill ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [email protected] with the message: INFO IBM-MAIN ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [email protected] with the message: INFO IBM-MAIN
