News for hackers. Stuff that matters.

A comment on Apple AMX instruction set (M1/M2 matrix coprocessor). Parent comment

Re:Apple AMX instruction set (M1/M2 matrix coprocessor) by samwillis on Monday September 05, 2022 @08:40AM (#32723623)
Apple are religious about interoperability within their platform, making it easy and reliable. If they were to do as you suggested with some of their proprietary tech, there will be products that implement it badly. To the user they would have no idea who’s at fault, and would probably blame the tech in general, damaging Apple.

Standardisation, in combination with certification to use the “label”, ensures that people developing on top of their innovation do so well enough that it doesn’t damage the brand.

(Somewhat less relevant to an instruction set, and not something I particularly agree with)

[ Parent | on HN ]
Re:Apple AMX instruction set (M1/M2 matrix coprocessor) by stephc_int13 on Monday September 05, 2022 @08:57AM (#32723789)
>interoperability within their platform

Isn't this called an oxymoron?

[ Parent | on HN ]
Re:Apple AMX instruction set (M1/M2 matrix coprocessor) by samwillis on Monday September 05, 2022 @09:35AM (#32724134)
Quite right, I should have said “ecosystem” really. And obviously this comment, about Apple in general, was a little off topic as a reply to the parent. I know what I was trying to get at, but didn’t convey it well in this context. Oh well.
[ Parent | on HN ]
Re:Apple AMX instruction set (M1/M2 matrix coprocessor) by Jcowell on Monday September 05, 2022 @12:05PM (#32725853)
No I’ve seen platforms where interoperability is non existent but it’s still a platform.
[ Parent | on HN ]
Re:Apple AMX instruction set (M1/M2 matrix coprocessor) by viraptor on Monday September 05, 2022 @08:52AM (#32723739)
This doesn't really make sense. There are already systems implementing connections to Apple stuff badly due to lack of documentation. The situation would only improve with the publication. For example we already have most of m1 hardware reverse engineered in Asahi - that's not going away. We've had things like air drop and earbuds charge state RE'd too. We'll get amx libraries as well soon.
[ Parent | on HN ]
Re:Apple AMX instruction set (M1/M2 matrix coprocessor) by scarface74 on Monday September 05, 2022 @09:46AM (#32724239)
“Apple’s stuff” is there hardware + operating system.

The benefit of buying “Apple stuff” is the integration between their software and hardware ecosystem.

[ Parent | on HN ]
Re:Apple AMX instruction set (M1/M2 matrix coprocessor) by viraptor on Monday September 05, 2022 @09:54AM (#32724309)
Yes, and... That seems irrelevant to the point of my post?
[ Parent | on HN ]
Re:Apple AMX instruction set (M1/M2 matrix coprocessor) by scarface74 on Monday September 05, 2022 @09:58AM (#32724340)
It’s relevant because

> For example we already have most of m1 hardware reverse engineered in Asahi - that's not going away. We've had things like air drop and earbuds charge state RE'd too. We'll get amx libraries as well soon.

You’re trying to use Apple hardware with non Apple software.

If you put Windows on an x86 Mac, do you expect the same experience (or battery life) that you get if you’re running MacOS on an x86 Max?

[ Parent | on HN ]