Ungo (a.delete@this.b.d.c.e) on May 23, 2022 9:01 pm wrote:
> --- (---.delete@this.redheron.com) on May 23, 2022 6:49 am wrote:
> > Ungo (a.delete@this.b.c.d.e) on May 23, 2022 2:05 am wrote:
> > > Doug S (foo.delete@this.bar.bar) on May 22, 2022 8:46 pm wrote:
> > > > I knew there were ranges of encodings that could be used as licensees wanted for new stuff like
> > > > AMX, I just hadn't realized Apple was also extending existing facilities like NEON.
> > >
> > > Maynard only has a patent to cite here. There's no known opcodes, no reverse engineering,
> > > and Apple files patents on tons of things they don't actually ship.
> >
> > It's true that patents are not dispositive.
> > However the dismissive attitude of so many towards them strikes me as just one more
> > step in the denialism that we've been seeing about Apple since pretty much the A7.
>
> You're so strange, Maynard. I own two different M1 Macs. They're amazing. Not perfect
> by any means, but performance per watt is off the charts and absolute performance
> is extremely good. I'm not sure what I'm supposed to be in denial about?
>
> Thinking that Apple has done some impressive work doesn't obligate me to go along with your spin on things.
> Opinions contrary to yours aren't attacks on Apple, that's the self-absorbed superfan in you talking.
>
> > On the one hand plenty of what I've found in patents has been
> > validated, either by my explorations or by those of others.
> > On the other hand, when you look at enough of these things, you start to see the patterns,
> > the patents that seem like something out of left field unrelated to the rest of the corpus
> > vs the patents that seems like a natural continuation of so much else.
>
> On the third hand, you're the guy who once got really angry at me just because I dared suggest that in production
> Mac hardware, unlike the A12Z DTKs, Apple might put Rosetta support into the efficiency cores, not just the performance.
> I don't remember exactly what your pretzel logic was, but what I do remember was the broader message: 'I, Maynard,
> have declared that Apple couldn't possibly want this, so shut up you ignorant fool'.
>
> You were, of course, completely wrong. Am I in denial about something if I remember
> that, and am inclined to think that maybe your ideas shouldn't be accepted uncritically,
> without reservation? Or if I remember all the times you've bitterly attacked subject
> matter experts here for telling you the truth about the thing they're SMEs in?

In both that episode and this one,
(a) you seem to like to remember/interpret cautious "it is possible" statements as extreme "respect my authoritah" statements.
I can't do anything about that.

(b) you seem to be very concerned about exactly what counts as *evidence*.
Understanding progresses by making leaps, sometimes unfounded. The problem is not the leaps when no evidence exists, it is an insistence when contrary evidence exists.
Some people naturally are afraid to go beyond absolute certified evidence, some are more willing to try to look for patterns in whatever evidence is available. This is an unending issue in Philosophy of Science and is ultimately an issue of politics more than anything else, Democracy vs Aristocracy.

You seem to be under the impression that I do these explorations to CONVINCE people (and especially you) of something.
I could not care less about *convincing* you or anyone else.
I do this exploration because I find it fascinating, and the only person I care about convincing is myself. As I see new data my understanding evolves and changes, sometimes dramatically (as happened with my understanding of the L1D between the first publication of my M1 PDF and the second publication, and as is happening right now with my understanding of the Fabric). Hell I flipped back and forth maybe five times as to whether the M1 is really performing address value prediction to perform early loads, as I'd write some code that seemed to confirm the hypothesis, then I'd come up with an alternative explanation that did not require such value prediction.
I'm frequently convinced (or at least have my understanding improved) by people whose opinion I respect and who make good points. And sometimes there's standoff -- Dougallj and I continue to have rather different ideas about how the scheduling queues in the M1 work -- we've both seen the same evidence, but we each see a different pattern behind the evidence. Yet we continue to respect each other (well, I don't know what's in his mind!) and we don't need to fight because we're both on the same page that the goal is understanding, not evangelism.

Those who have never bothered to engage in *serious* exploration of a new field don't understand how this plays out, the balance of tentative hypotheses that all have to be retained as one looks for dispositive evidence in one direction or another.
Well I can't do anything about that.


I write up occasional notes here about things I find especially interesting, or amusing, or fact points that I think add to the discussion. The ANS in NEON patent (plus the Asahi list of as-yet-uninterpreted ISA opcodes) are two such facts. If you choose not to find them interesting or convincing, that's your choice.
But to respond as you do, with the implication that you want me to stop posting such bits and pieces seems, umm, anti-social? My guess is that more people find them interesting (whether they're convinced of my [always tentative] current explanation or not) than are offended as you are.

But hell, I'm happy either way. I have patents to explore, Mathematica to write up my analyses and to run simulations, an M1 macbook air to run test code. What more do I need?
Before my first publication I hoped that by publishing (small pieces here, and the large PDFs) others would join in and the project could run faster. But I'm discovering the same thing discovered by so many pioneers -- many many more people prefer to criticize than want to help.
I do not have time to do everything else I am doing right now and also run experiments to try to find opcodes that might match ANS in NEON. You'd hope at least one person would think "that sounds like a fun project, I'll try doing that". But there are a million people who'd prefer to complain "you don't have evidence, only a suggestion" for every one that willing to actually do something about that, like help look for evidence, one way or another.
Unfortunately it's very difficult to hold any sort of discussions about tentative matters in a public space given that the mass of the public simply do not comprehend the very notion of "tentative", the concept of "strong opinions weakly held", or the idea that the search for truth is more interesting than the search for allies.
Well, like I said, I also can't do anything about that.