Daily Linux & Open Source News - S01E200 - Valve talks Steam Deck 2
Added 2024-10-16 09:00:54 +0000 UTCHey everyone!
In this episode, we have:
00:05 Nvidia keeps patching multi GPU code on Linux
https://www.phoronix.com/news/Linux-GPU-P2P-DMA-Device-Priv
02:44 Intel & AMD partner up to form an x86 board
06:28 Valve talks Steam Deck 2
https://www.reviews.org/au/games/valve-steam-deck-australia-interview/
Have a great day!
Comments
I won't pretend myself to be an expert on the matter but the technical design of x86 as an instruction set is just god-awful. Writing x86 assembly was without a doubt one of the most painful programming experiences I've ever had (and I've played around with JS and Java!). Needless to say, RISC-based ISAs (like ARM/aarch64, RISC-V, etc.) are MUCH nicer to work with. Fundamentally, the design of UNIX and by extension Linux (as a UNIX-like) and further macOS runs perfectly contrary to that of x86. Interrupts should not live at the hardware level. Power "state" should not live at the hardware level. Predictive execution should not live at the hardware level. There's a reason Apple has done so well for themselves since the release of M1 silicon: the DX on RISC-based architectures is _objectively_ better suited to UNIX-like systems. Their profits weren't just from sales--it was also from how much development time (and thereby money) they saved by not having to constantly fight x86. The UNIX philosophy isn't just DOTADIW--it's also PNWOAYUS: play nice with other, as yet unwritten software. x86 and everything Int\*l and/or MICROS~1.EXE have touched totally fails at both. They implement half of one thing and make it inaccessible to anyone who doesn't want to bend the knee. x86 is the "dominant" ISA just like W\*ndows NT is the "dominant" OS architecture, and DOS was "dominant" before that. That doesn't make it good. It makes it the popular choice of the herd--the herd who are not necessarily well-versed in what is "good". The fact is that nine times out of ten, discounting either money or time, popularity simply doesn't matter. UNIX proved that. And then Linux proved that when you don't need to discount those things, even more people will use it. GitHub is popular. GitLab is better. Same thing there: someone needs to invest the money and time for GitLab to be an objectively better experience than GitHub. GitHub had to do the same to beat SourceForge.
darkwater4213
2024-10-19 22:06:15 +0000 UTCI mean, that used to be true, yeah. But AMD is now a sizeable competitor, and they do own a lot of the patents on the AMD64 (x86_64) part of the architecture, they even have cross licensing agreements. So, sure, x86 is proprietary, and owned by Intel, but also owned by AMD. it' s not like AMD couldn't have refused this alliance or couldn't exist without Intel today. As per being a crappy ISA, I don't know enough about that, but it's certainly not DOA, since it's been the reference architecture for what, 20-30 years now? It has evolved to the point where it was able to resume power consumption and efficiency parity with recent ARM CPUs, in a matter of a few years. I don't think it's that dead... Then again, my knowledge of architectures and instruction sets isn't the best.
The Linux Experiment
2024-10-19 07:24:55 +0000 UTCI think representing talks between Int\*l and AMD, as if AMD does not owe their very existence as a major x86 player _to_ Int\*l is deceptive at best. Int\*l financed a competitor (AMD) precisely so they couldn't be broken up as a monopoly. That's not AMD being independent of Int\*l. AMD is very much dependent on Int\*l for licensure of the x86 ISA and for not squashing them like a bug. So first, x86 is proprietary. Second, Int\*l is a monolithic basilisk with only one goal (murder everyone on the way to conquering the world). Third, x86 is genuinely a $#!£ (pardon my French) ISA and was dead on arrival.
darkwater4213
2024-10-18 16:49:58 +0000 UTC
