Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Tales of the M1 GPU

asahilinux.org

Post reply on HN
Print view View original

341–350 of 390 posts

Re: Tales of the M1 GPU

#341

Post by saagarjha »

p_l wrote: post #237
monocasa wrote: post #49

Earlier quoted context omitted.

> Hector Martin claims so which (frankly) suggests he has some insiders. Eh, I've heard the same thing form people I trust, and I'm not the Apple news version of deep throat or something. It's simply not a very well kept secret. Although part of me wonders if the code will flow the other way. Now that marcan has put in the elbow grease to upstream concepts like 16KB pages, the non standard ordering for regular MMIO a…

16kB pages weren't too hard to get running before, iirc, it's just that most devices people care about went to 64kB. Generally linux is happy unless you try to make pages smaller than 4kB,then all sorts of hell break lose in VFS

Linux is generally fine with 16K pages; userspace programs often are not.

Re: Tales of the M1 GPU

#342

Post by saagarjha »

gjsman-1000 wrote: post #41
monocasa wrote: post #39

Earlier quoted context omitted.

I think there's larger barriers to getting windows running on Apple Silicon that would need to be addressed first. For one example, Windows ARM kernels are pretty tied to the GIC (ARM's reference interrupt controller), but Apple has its own interrupt controller. Normally on ntoskrnl this distinction would simply need hal.dll swapped out, but I've heard from those who've looked into it that the clean separation has br…

Apple Silicon doesn't use GIC, but uses AIC (Apple Interrupt Controller). "Apple designed their own interrupt controller, the Apple Interrupt Controller (AIC), not compatible with either of the major ARM GIC standards. And not only that: the timer interrupts - normally connected to a regular per-CPU interrupt on ARM - are instead routed to the FIQ, an abstruse architectural feature, seen more frequently in the old 32…

Apple silicon doesn't use GIC, but it has some sort of vGIC that I believe the Asahi Linux team is using.

Re: Tales of the M1 GPU

#343

Post by saagarjha »

punnerud wrote: post #308

Not one comment here about the “GPU drivers in Python”. I like the idea of iteration speed, over pure speed. And the coprocessor called “ASC” also have similarities with Python, where the GPU is doing the heavy lifting, but the ASC (like Python) interact using shared memory. The same Python is doing with a lot of its libraries (written in C/C++)

> And the coprocessor called “ASC” also have similarities with Python

It's a processor, not a programming language :) The team has essentially strapped the API into something that you can poke with Python instead of with a native driver.

Re: Tales of the M1 GPU

#344

Post by saagarjha »

comprambler wrote: post #138
helf wrote: post #57

The fact so much hardware these days is running a full real-time OS all the time annoys me. I know it is normal and understandable but everything is such a black box and it has already caused headaches (looking at you, Intel).

Technically it is not a real time OS. There are very few OSs that have this moniker (vxworks, qnx, etc)

"Real-time" isn't a trademark, you can assign it to other things if they meet the typical guarantees of "real-time".

Re: Tales of the M1 GPU

#345

Post by saagarjha »

jbverschoor wrote: post #183
helf wrote: post #57

The fact so much hardware these days is running a full real-time OS all the time annoys me. I know it is normal and understandable but everything is such a black box and it has already caused headaches (looking at you, Intel).

I sometimes wonder about how fast tings could go if we ditch the firmware, and also just bake a kernel / os right into the silicon. Not like all the subsystems which run their own os/kernels, but really just cut every layer, and have nothing in between.

Then you can't improve them.

Re: Tales of the M1 GPU

#346

Post by saagarjha »

the__alchemist wrote: post #166
helf wrote: post #57

The fact so much hardware these days is running a full real-time OS all the time annoys me. I know it is normal and understandable but everything is such a black box and it has already caused headaches (looking at you, Intel).

Same. It's not about the principle, but that generally these OSes increase latency etc. There's so much you can do with interrupts, DMA, and targetted code when performance is a priority.

What makes you think performance was not a priority here?

Re: Tales of the M1 GPU

#347

Post by saagarjha »

TimTheTinker wrote: post #77
liuliu wrote: post #64

Earlier quoted context omitted.

I don't know. This sounds very computer-sciency-ish. We build smaller tools to help build big things. Now the big things are so good and versatile we can replace our smaller tools with the big things too. With the more powerful tools, we can build even bigger things. It is just compiler bootstrapping happening in hardware world.

The problem is that there's so much unexplored territory in operating system design. "Everything is a file" and the other *nix assumptions are too often just assumed to be normal. So much more is possible.

You do realize that RTKit is not a Unix-like OS, right?

Re: Tales of the M1 GPU

#348

Post by psychphysic »

yazaddaruvala wrote: post #264
psychphysic wrote: post #132

Earlier quoted context omitted.

There's no point thinking of this as a cognition the pilomotor response is a reflex. One that evolved before the dinosaurs were wiped out. I can only tell you my feelings, thoughts and emotions. Feel free to come up with a better explanation!

I think any first reaction is always an entirely fair first reaction. Biases exist. Snap judgements exist. No point in pretending otherwise. I think like with anything "new", life is always about disconfirming our biases. As a result, now that you're aware of this bias, have the additional time and can access System 2: Going forward you find a way to have a less aggressive reaction to learning about a the next VTuber…

I almost certainly won't.

Unless this stuff goes entirely mainstream I will just avoid it.

This is the first time I've watched this kind of livestream and as well has as making my teeth itch with creepiness, not being an efficient means of conveying the topic it is boring.

I guess I'm older than the average HN user but I prefer reading to watching a video for education.

Re: Tales of the M1 GPU

#349

Post by saagarjha »

kajaktum wrote: post #261

Half the comments here are talking about the vtuber herself. Who cares. It's been talked before. Just imagine if half the thread is discussing what gender she is. What I am interested in is the claims here https://asahilinux.org/2022/11/tales-of-the-m1-gpu/#rust-is-... . (what is it called if it comes with a proof?). The resident C/C++ experts here would have you believe that the same is possible in C/C++. Is that tr…

Possible? Definitely. Easier? Probably not. At least for the most part, there are a couple things which C(++) can sometimes be more ergonomic for and those can be isolated out and used independently.

Re: Tales of the M1 GPU

#350

Post by drooopy »

fareesh wrote: post #304

I've never played games on my M1 Macbook - what are some popular reasonably graphics intensive games that it would support? Could it run Dota2 for example?

Resident Evil Village recently came out and it performs surprisingly well even on the low end MacBook Air M1 with only 7 GPU cores. What's even more impressive is that the game is playable (low gfx settings, 30fps) when running that machine on low power mode.
Post reply on HN