Single instruction, multiple data

... uh, um, numeric values.

And, uh, you know, and you can try to pause and convert that, uh, in a single call, so that's called a SIMD, Z, um...

Sorry, a SWAR SIMD within a register, so, you know, this kind of thing.

Oh, uh, yeah, SIMD was also involved.

SIMD was then also involved, so people used it, um, incubating a vector API.

Um, but it's really a combination of things, and people were, you know, improving for an entire month, and even they would have gone further had we not, uh, finished the challenge at the end of the month.

Uh, I, I think people would still have, uh, found more ways to actually make, make it faster.

I mean, you absolutely can, right? And I think people actually realize this, and maybe many folks still have this idea from 15 years ago, whatever, you know, like Java is like legacy and old school and it's, it's slow and w- and super verbose, and I hope this challenge helped a little bit that, you know, to overcome those misconceptions.

That's the difference, right? But name a thing, whatever you want.

SIMD, caches, loop caching, virtual memory, whatever it is, it was there.

It was there since like at least 1970, but usually a little earlier than that.

And so if you went and learned about how computers work, meaning you were like, here is a survey of how they work.

That's the only way you do that in a language, right? In some higher level language, that's great.

In C, yeah, the vanilla language C doesn't have these things in it, but that's not the point.

The point is that if you already kind of know roughly how to learn something like C or Rust, you can then just right there, without teaching the student anything new go, now we are going to go use this crazy instruction on your CPU by using an intrinsic to go do this thing that looks at a SIMD vector and takes the mask register and packs the ones that are zeros into like, and you can understand it with knowing nothing more than just the CU already learned and the little bit that they're trying to teach you now.

And so at least what I got out of her statement, and I thought she said it pretty clearly, she even used the word like extending down to memory management or things like that.

They can like hang on for, you know.

Slightly more relevant package, I feel, or technical, I guess, package is the new portable SIMD package.

Not a ton to say about this other than what does SIMD do? So SIMD stands for single instruction, multiple data.

And what it does is it lets you in new processors.

They come with these new, I'll take a step back.

It's like a table of...

MOV and MOV, you know, takes these two parameters and it does this, uh, and ADD, which is like ad and, uh, JMP, which lets you jump to a specific instruction in a memory and whatever.

It's like the very low level instructions that the processor runs, right? There are new SIMD instructions, right? Again, SIMD stands for single instruction, multiple data.

I mean, it's always been somewhat divorced, but is even dramatically more divorced from the building things field.

And there will be people that are interested in computer science and there probably are, um, there's a market demand for people who are finicky about how SIMD works and like these like low level details of things.

And but the...

Like so many people, the vast majority of the software work is just like the building things side, right? And I s- I sort of...

All right.

Uh, my next one, uh, sort of in congress, I think, with a lot of the stuff we've been talking about so far, but that's okay, is everyone should know SIMD.

So we talked about SIMD, uh, number of episodes ago.

Yeah.

Someone's gonna be like, "Oh, it was 100 episodes ago." Um-

[laughs]

But I don't think it was.

Um, SIMD is a single instruction multiple data.