On 10/06/2017 01:56 PM, Paul Koning wrote: >> On Oct 5, 2017, at 6:19 PM, allison via cctalk <[email protected]> wrote: >> >> Moore's law only worked for hardware, software lagged typically two >> years behind. > There's a more cynical view, sometimes called "the virtual disease", which is > that software performance is constant because all the gains from Moore's law > have been consumed by increased software inefficiency and complexity. The key is mature technologies got speed from efficient software and hardware when programmers were skilled and they were many. Now it though hardware at it. Best example is Rpi 3b using Raspbian linux to control and monitor a garage door over the Internet. Seems like Atomic weapons to kill ants to me. But it was cheap.
The rule is still to go twice as fast the cpu has to be 3x faster and run the same crappy code. IF the code updates then the CPU has to faster still. None of the 64 bit PC OSs seem faster than the previous 32bit despite being on faster platforms. > A related aspect is an observation from my boss at DEC: interrupts (context > switches) always take 10 microseconds no matter how fast the CPU is. Likely I worked with him back when. VAXen did the context switch fairly fast and the PDP11 was really good but, PCs yuck. And the more complex the OS the worse it is no matter how many cores! > Both of these are slightly unfair but much too close to the truth for comfort. Actually right on the nose if you ask me. Then again I still use a Epson PX-8 Z80-CP/M system because it works! Allison > paul >
