Comments by "Anony Mousse" (@anon_y_mousse) on "Linus Torvalds Is Testing Linux On M1/M2 Macs" video.

  1. I personally dislike RISC architectures, but the competition is great and they are innovating more on product lines. So I'll use them, but I would prefer if a new platform would rise up to compete with both and be a more unified architecture. One of the reasons I like CISC better is because you can really pack instructions in the cache a lot better and it does have a noticeable speed increase on your system. Speed testing also proves that as they add more extensions the speed suffers when mixing modes a lot. That's why a lot of ARM devices still haven't fully gone 64-bit, and it's annoying as heck. If a new CISC architecture cropped up with 8 to 64 bit instructions baked in from the start, it would kill everything else. For that matter, I'd love to see the FPU be more integrated and work more similarly to the ALU. More registers, wider variety of sizes, and all usable whether we're talking about multimedia instructions or floating point. No more duality, and just be 64-bit on startup. It'd require a lot to get it going, but any of the larger chip makers could do it, if they had the testicular fortitude. No one will read this far, if they read this at all, but I'm sure people will disagree with this vehemently for one reason or another, if they respond at all.
    3
  2. I pronounce it Lie-nucks, and I don't care. ;p
    1
  3.  @rohithmekala2608  And the rest of what I said, no opinion?
    1
  4.  @hishnash  Who writes the compiler? If they're auto-generating it using a toolkit, who writes that? If they're auto-generating things it won't be very optimized, so who optimizes it? How are instructions generated in these circumstances? If the CPU handles the cache the right way, any branch prediction mismatch is easily handled and these days it's far less of a problem than back 20 years ago.
    1