A slight re-organization of Priest's "Efficient Scaling for Complex Division" to make it compatible with "try to divide the dumb fast way inline, then branch to rescale only if necessary" while preserving scale invariance of rounding.
Also fixes it up to work for Float16, which the original approach does not.
Further optimization possible and pretty straightforward.
kind of weird to think that the last time most people learned something truly new (e.g. they have no intuition for the thing) is during childhood/adolescence
I didn’t learn to do anything athletic to any level of skill until adulthood (olympic weightlifting). It was immensely hard to learn, I think in part because everyone else doing it, and everyone teaching it, had some sort of sports intuition to draw from (football, gymnastics, volleyball)
Eventually I figured out the movements, and understood what I was supposed to be doing, and was able to link that back to the simple cues they were giving me. And I found that I was really good at explaining stuff to other unathletic people because I, just like them, completely was missing any standard athletic foundation, so I had to explain everything without shorthand/cues
It was pretty uncomfortable to suck at something so much. I saw a lot of other folks similar to me show up for a couple classes and then give up, probably because… as an adult it’s really uncomfortable to actually truly be bad at something and feel completely lost.
It’s completely reasonable to not want to spend time on things that you’re bad at. But also, it can be rewarding, and the new experiences can help you develop new perspectives on the things you’re already comfortable with.
Also being able to just… accept that you might not be “good” at something is a useful skill I think.
Learning to look at other people who are better than you at something and say “wow so and so is really amazing” rather than “wow I suck” is really nice, and allows you to experience more in your life.
fwiw if it doesn’t work out i’ve generally found fleo’s stuff comfortable and good for moving around in, and they actually have humans reply to emails about fit etc
Wrote up my notes from Community.o on newcomer paths into LLVM. Main theme is that we need to get documentation out of people’s heads into the actual project, and empower beginners to take ownership over pieces of the documentation that confuse them. We should try to decentralize knowledge in the community.
I managed to clean up a lot of cmake garbage today but every time I use this language it feels like I’m like, trying to hammer a nail with a screwdriver
CAT SAGA: something that kind of sucks about being scammed by a california business (the pet relocation company) is that basically all standard approaches for like, getting your shit back assume that you live in california (or at least in the usa)
so like, basically this company seems to feel fairly confident that they can just walk away with tens of thousands of people’s dollars which is like, awesome
The cat saga continues: the pet relocation guy won’t respond to me and keeps coming up with weird excuses to not buy a plane ticket so I’m just…. going to fly the cat myself with a friend and figure out the pet import process on my own?
Unfortunately I paid in full and there are reviews popping up about the company saying that they’ve been dropping the ball and not refunding people so uh that’s great
also before I started moving I didn’t know rabies is mostly just a North America thing and that the rest of the world really really does not want the zombie virus