i am a little surprised that programmable hardware in not more prevalent for large scale computations. i don't mean program once like gpus, mining stuff, but more on the fly, sort of like jit with hardware.
The pipeline for x86 is already a sort of JIT in how it operates on the hardware level.
The frontend looks for instructions that can be executed immediately and has a lookahead as well. This is how the out of order execution is implemented which, really, is tantamount to parallel execution, and it is effectively parallel execution, if not actually, in some circumstances.
If something can be executed out of order, it means it can also be executed in parallel.
@copper.hat Noice.I've mostly been a software kind of guy, but I'm getting more and more into hardware, and I want to eventually be able to create (or at least design) hardware as easily as I make software.
Hardware isn't that much different of course. They're analogs of each other.
For square root, any integer under the radical that is not a perfect square is irrational, and if the integer is negative, it is imaginary, and the imaginary part of that number is irrational.
@copper.hat I'm only saying that building circuit diagrams and writing assembly are analogous operations. I don't have to be an electrical engineer to design the circuit itself, but to get something that is optimal, having knowledge of the hardware might affect my design choices.
@copper.hat Well I don't know what you call it. They're extremely related. One is restricted to logic gates, and the other is restricted to the instructions of an architecture. The latter can be entirely interpreted as a circuit diagram. That's all.
Back when I was in high school, the closest I got to designing hardware was making an ALU using Minecraft redstone. They used the same circuit diagrams that you can find on wikipedia, but optimized with the understanding of how redstone physically propagates and how many ticks it takes for redstone to propagate, etc., so I'm sure the experience would be similar, just with electricity, or whatever I would decide to use if not electricity.
@copper.hat most of the downvotes on my answers are to good answers. In some cases, they were on answers to PSQs (usually before we were monitoring for such things), sometimes they are simply spiteful.
i kind of like spite downvotes. on my own account. i'm not saying anybody do any of them but it's funny when it's like "what could this possibly be" and it's nothing.
@leslietownes once in a while, my attention is drawn to an old answer by an unreasonable downvote, and I see that I can improve it. However, the downvote usually does not go away.
I think I should just study how libdivide works and then make my own implementation using their algorithm so I don't have to use their code (and include their license).
Actually, most of the pre-cleaning person cleaning is to hide the things you don't want the cleaning person to put some place where you'll never find it.
we had a person who would come every two weeks to dust and mop floors. we paid her through most of the pandemic because she had a whole family to feed even though she did not come here. then she moved.
our apartment in oakland had the world's dodgiest handyman. he was a lot of fun but never did anything on schedule and eventually he was fired because he was stealing stuff from people's apartments.
you gotta draw the line somewhere.
he was a nice guy, though. our landlord was a complete hippie and did not get the police involved.
i used to play nethack and one mechanic you can use to succeed in that game is to do the same stuff over and over. i did that.
it satisfied my lizard brain
my daughter is going through a stack of my business cards and scribbling on every one of them. i never use them so i'm fine with it, but she's done 20 cards.
If a game doesn't challenge my intellect in some way, or I can't find a way to make a challenge out of it myself, then I don't find it very worthwhile. Unless of course I don't want to be challenged, in which case I want it to have virtually no challenge at all.
@leslietownes Hahahaha.. I won't get mad or disappointed for playing such games with children. However, I find it hard to respect people who play cookieclicker for more than an hour.
mas if you have previously identified forms of cube roots of 1 this would be an opportunity to use that information. alternatively there is the course of just cubing that thing and seeing what you get.
Suppose $c=\sum_im_i \sigma_i\in S_q(\Delta^n)$ is affine. Then $meshSd^q(c) \leq (\frac{n}{n+1})^qmesh c$.
, where $S_q(\Delta^n)$ is the free abelian group generated by singular $q$ simplices in $n$ and $Sd$ is the subdivision operator and $Sd^q$ is the its composition $q$ times.
I'm having tro...
Oh, by the way, decided on complex exponential with 16 taylor terms and 2^16 for argument reduction. Turns out I get more precision for free that way since it only requires 4 multiplies. Originally used only 12.
I'm glad addition-chain exponentiation is a thing :)
How's everybody's proficiency with multiple linear regression models? Actually @copper.hat, you may be able to help because it has more to do with the linear algebra treatment of the idea and not the statistical stuff per say.
The idea revolves around "deleting an observation" from a vector of observations. So in order to calculate the Cook's distance I do the following (numerator is all vectors, in fact I'm going to ignore the denominator here):
Where $(i)$ is the deleted case from a vector of $n$ observations. So if I "delete" an observation that means $\hat{Y}$ and $\hat{Y_{(i)}$ are of different dimensions....so how would the algebra actually work?
damn it...... give me a sec to write properlu
The idea revolves around "deleting an observation" from a vector of observations. So in order to calculate the Cook's distance I do the following (numerator is all vectors, in fact I'm going to ignore the denominator here):
Where $(i)$ is the deleted case from a vector of $n$ observations. So if I "delete" an observation that means $\hat{Y}$ and $\hat{Y_{(i)}}$ are of different dimensions....so how would the algebra actually work?
how do you guys designate the ith component of $f(\vec x)$. $f_i (\vec x)$ looks like partial derivative notation and $f^i (\vec x)$ looks like the ith derivative
Ok....I'm going to post it in Stat Exchange.....I know it is a minor thing in the grand scheme of it all, but it does need to be clarified. Continue on with your saturday enjoying a fine wine :)
I can just use a lookup table for fixed point log_2, either for four bits or eight bits depending on whether or not I want to have 4c or 2c 64-bit divide respectively.
No idea why I didn't think of this earlier, but it's just what I need in the absence of an efficient hardware implementation for division.
Still, tsk tsk. I would prefer a hardware instruction... I hope my software project doesn't take too long to get sufficient returns to make my own CPU someday...
Anyways, quick question: is there a way that I can use $\frac{i}{i+x}-\frac{x^{17}}{x+i}$ to compute my approximation of $\exp(ix)$ even faster?
That expression happens to be all the terms of the Taylor polynomial for complex exponential but without the unique coefficients.
