Re:Driving Amazon’s electric delivery vehicle: Rivian EDV [v...
by thechao on Wednesday December 07, 2022 @08:50AM
(#33894290)
I seem to remember that a good pencil has lag on the order of 1ms (how long it takes the tip to settle out after high speed motion), but I don't remember where this number came from, so it's untrustworthy.
I always view the IBM selectric as "beyond the horizon" in terms of responsiveness: that's 25–30ms.
Re:Driving Amazon’s electric delivery vehicle: Rivian EDV [v...
by TeMPOraL on Wednesday December 07, 2022 @09:18AM
(#33894567)
> I seem to remember that a good pencil has lag on the order of 1ms (how long it takes the tip to settle out after high speed motion), but I don't remember where this number came from, so it's untrustworthy.
Not sure what lag there is on actual pencil, but there's this somewhat famous video[0] from Microsoft Research, where they test various values of input lag on a touch screen, and demonstrate that even 10ms is noticeable - but 1ms is about enough for it to feel instant. With 1ms lag, drawing on a tablet seems to feel just like drawing on paper (I personally haven't had an opportunity to test a tablet with 1ms lag, so I can't tell).
--
Re:Driving Amazon’s electric delivery vehicle: Rivian EDV [v...
by goodpoint on Wednesday December 07, 2022 @08:41AM
(#33894198)
...until you start comparing lag and CPU performance. Then it becomes obvious what kind of bloated software the world is producing.
Re:Driving Amazon’s electric delivery vehicle: Rivian EDV [v...
by TeMPOraL on Wednesday December 07, 2022 @09:22AM
(#33894605)
And then you notice that recently there was an article pointing out that "software is IO-bound anyway" is no longer true, and this claim - for the first time since I remember - wasn't laughed out of the room. I await a more detailed study, but it seems quite likely the article was correct: IO is quite fast these days, and software is likely to be CPU-bound, particularly on parsing stuff as it's being read and operated on.
Re:Driving Amazon’s electric delivery vehicle: Rivian EDV [v...
by tormeh on Wednesday December 07, 2022 @11:07AM
(#33896000)
Wasn't the limiting factor RAM latency? Anyway, yes, computers parse way too much JSON, SQL, and whatever else. I've recently started working with gRPC, and boy is it nice. The current paradigm where every action requires a text blob to be written, compressed, then sent to a server that decompresses it, parses it, acts on it, then creates a new text blob that is compressed and sent out... It's a bit much. And that's before ypu add load balancers and microservices. gRPC cuts down on the parsing and compressing bits, but it's still quite wasteful.
It seems like there might be a slight negative correlation (newer is worse). The data is fairly all over the place though.