2026-01-16 00:00:06 But... we had the have the software ALREADY OUT THERE run faster. That's the tail wagging the dog if you ask me, and it has dictated the entire arc of microprocessor development for decades. 2026-01-16 00:01:05 had "to" have 2026-01-16 00:29:07 KipIngram, as a 95% electronics only guy I have to admit, my interest drops off below about 90nm node sizes 2026-01-16 00:30:59 KipIngram, I remember being awed at the Motorola 6800 when it came out and I grew up with valves and uses them for probably 4 years to learn electronics before I even read about Germanium transistors 2026-01-16 00:32:40 I still remember the famous OC71 transistor fondly 2026-01-16 01:31:45 KipIngram: In terms of how recently we learned some things, I've got a few that I've been boggling over recently, too. 2026-01-16 01:33:24 dzho, talking about mind boggling, China is currently installing 100 solar pannels every SECOND on their national power grid : https://e360.yale.edu/digest/china-renewable-photo-essay 2026-01-16 01:33:25 Even though Rutherford predicted the existence of the neutron circa 1919, James Chadwick is credited with its discovery only in 1932. 2026-01-16 01:33:37 tpbsd: also an awesome fact, yeah! 2026-01-16 01:34:15 so, we are still 6 years out from that centenary 2026-01-16 01:42:41 dzho, i'm in a bit of a 'zen state' atm after a 43 hour fast so all these mind boggling facts are extra enjoyable right now 2026-01-16 01:43:18 dzho, extra clarity for my simple electronics technician brain :) 2026-01-16 01:45:30 dzho, the pics in the china-renewable-photo-essay are just so awesome, I'm blown away by their ethereal nature 2026-01-16 01:48:04 tpbsd: they are striking, yes 2026-01-16 01:49:39 also, I'd been taught about atomic theories stretching back all the way to Democritus but in the early modern era also attributed to Dalton. 2026-01-16 01:50:09 What I didn't realize that there was fierce debate as to whether atoms were real, or a convenient model, all the way up into the early years of the 20th century. 2026-01-16 01:50:26 https://www.bbc.com/audio/play/m00057s5 2026-01-16 01:51:08 atomic theory won the day, but it was hard fought at the time by a number of scientists with familiar names 2026-01-16 01:51:16 familiar, at least, in the field 2026-01-16 01:51:27 Ernst Mach and Ostwald were opposed to the idea 2026-01-16 01:52:05 Maxwell and Boltzmann in favor, but it took a toll particularly on the latter. 2026-01-16 09:43:42 KipIngram: which processor would this intelligent compiler optimize for? 2026-01-16 09:44:06 since different intel processors have different microcode schemes and number of steps for the sameinstructions 2026-01-16 11:29:32 KipIngram: While executing the processor has more up-to-date info to optimise reordering than you can possibly have statically, bear in mind 2026-01-16 11:30:49 Beyond even the microarch etc; you don't know until executing what state cache, pipeline, etc will be in; you can guess but it's a fragile estimation compared to just being in the processor right now and actually knowing, which the processor does 2026-01-16 15:51:42 veltas: I don't think you understand how far I'm talking about having adjusted our philosophy. I think we should have kept cores brutally simple. So simple that it was easy for one person to understand the thing in its entirity, which would have made situations like Spectre and Meltdown a lot less likely. And then use all that logic that we're using for these other things now for "more cores." 2026-01-16 15:51:45 We tried for a long time to maintain a single-core mentality, but in the end we had to give it up anyway - I just think we should have swallowed that pill right off the bat and dealt with it. So, none of those "fragility inducing" things would even BE THERE. The state WOULD be uttlerly predictable. And as a fringe benefit we'd still be able to cycle count in an accurate way and so on, making 2026-01-16 15:51:47 real-time applications easier. So I'm not talking about any kind of "minor nudge" to our extant situation - I'm talking a a wholesale different approach. 2026-01-16 15:52:20 MrMobius: with brutally simpler processor cores it would be a lot easier to optimize for any given one, so the compiler could just "have a list of supported targets." 2026-01-16 15:53:03 So, I'm talking about a totally different ecosystem that I think would result in more user work being delivered and, importantly, a much higher level of security reliability. 2026-01-16 15:53:23 Fewer esoteric bugs, weird performance gotchas, etc. etc. 2026-01-16 16:44:18 i've made a matrix animation with my toy rpng lang xd 2026-01-16 16:44:23 https://gitlab.com/vms14/oh/-/blob/master/oh.oh 2026-01-16 16:45:20 that's a video of the animation itself 2026-01-16 16:45:23 https://youtu.be/6g_Bi1e2IKM