Speed now & then

Boris Gimbarzevsky boris at summitclinic.com
Sat Apr 14 20:40:09 CDT 2018


TAhanks for that link which fits with my 
measurements (nowhere as detailed) of ones actual 
ability to do things with "modern" hardware.  In 
the 1980's I was used to being able to measure 
events with 0.2 microsecond precision using a 
PDP-11 and my expectation was that the accuracy 
was only going to improve as processors got faster.

I ported a program I wrote on the PDP-11 to a 
Commodore 64 in 1988 and was using it to measure 
finger tapping with a switch array to 1 msec 
accuracy.  This was done through the simple 
expedient of speeding up the sample rate for the 
keyboard to 1 KHz and the adding in my 4 external 
switches as "keys".  Used a 512 K Mac to get the 
serial data and display results.  To do the same 
now would require custom hardware to do the 
timing and a USB link to a "modern" CPU or implimentation on a microprocessor

When I attempted to get this same type of timing 
accuracy from a PC, found out that it was no 
longer easy to get access to interrupts as easily 
as before and keyboard latency was longer as now 
keystrokes were detected by an on board 
microprocessor and sent out as a series of 
packets for each keystroke.  In DOS and W95 where 
one could still easily get at interrupts, then a 
serial port could be used to do msec 
timing.  Once XP and beyond arrived, then the 
best temporal precision one can expect from a 3 
GHz machine is 15 msec.  I suspect the same holds 
for Macs and haven't tried running real time 
Linux as I either pull out my trusty C64 from 
time to time and use it for precision timing 
(unfortunately have only one copy of the code on 
casette tape so when that goes can't do this 
anymore) or I use various microprocessors to do 
the job.  Have a nice microsecond precision timer 
that I wrote for a Propeller chip and feel much 
more comfortable programming for it than the 
latest windoze bloatware system.  The Propeller 
has the same amount of RAM as the PDP-11's I 
started on, runs 20x faster/core and is fun to 
program.  The microsecond timer is attached to a 
geiger counter to generate random bytes for OTP encryption.

Boris Gimbarzevsky

>On 29 March 2018 at 19:53, Paul Koning via 
>cctalk <cctalk at classiccmp.org> wrote:
> >
> > It would be fun to do a "generalized Moore's 
> Law" chart, showing not just transistor count 
> growth (Moore's subject) but also the many 
> other scaling changes of computing: disk 
> capacity, recording density, disk IOPS, disk 
> bandwidth, ditto those for tape, CPU MIPS, 
> memory size, memory bandwidth, network bandwidth...
>
>This is the most telling I've seen in a long time...
>
>https://danluu.com/input-lag/
>
>--
>Liam Proven • Profile: https://about.me/liamproven
>Email: lproven at cix.co.uk • Google Mail/Hangouts/Plus: lproven at gmaill.com
>Twitter/Facebook/Flickr: lproven • Skype/LinkedIn: liamproven
>UK: +44 7939-087884 • ÄŒR (+ WhatsApp/Telegram/Signal): +420 7002 829 053




More information about the cctalk mailing list