m-j-fox

2035 is a long time from now. Hope someone hatches a practical energy plan in the meantime.

Sure, but in silicon valley one thing to do with your smalltime IPO winnings is buy a house and rent out your old one to someone who has yet to hit paydirt. Maybe part of what pushes prices up, but I like to think my renters are getting a better deal than they would from old-man Irvine. (The incentives to play this game are razor-thin, mostly down to tax advantages.)

Seems like a good opportunity for someone to register disparage.org and start recording complaints against the uncomplainables.

The one that confuses me is AB 71. Aren't 2nd and 3rd homes basically rental stock? Owned by individuals and not the Irving Company if you care about that.

Not really, publication quality. Just a Sunday-morning effort to understand the ceiling on the phased-execution model. All resources like decode, #pipes, cache and branch prediction are unmodeled, assumed infinite. I'm sure anyone can improve on it, but I lost interest with disappointing results.

https://pastebin.com/14E7X41j

I dump the instruction stream for coremark out of LLI interpreter, hacked a tiny bit. To save trouble, here's one iteration of coremark:

https://drive.google.com/open?id=0B0ygb7T_Ab3kM01DQnlvQi1teG...

gunzip the data, run it through the script and count the lines out vs lines in. (I get "IPC" ~2.12.) For a bit more insight, the number on the left of each output line is how many instructions fused. You can make a histogram like:

    cat tmp.out | cut -f 1 -d " " | sort -n|uniq -c

    76056 1
    95805 2
    32931 3
    4401 4
    1172 5
    970 6
    ...

It's good he recognizes that most code won't benefit from switch optimization. But certain tasks like pattern matching benefit greatly. For that, I suggest:

http://www.micronautomata.com/

Micron will eat 1000 nested switches 100 cases wide every cycle. This sort of TCAM tech was rare 5 years ago, but everywhere now.

If true, that'd be a very high number. Intel can get 2 on some benchmarks, but system-average IPC tends towards 1. (In terms of instructions standardized for cross-platform comparison. I take data on ARM servers and I'm jellous of 1.)

Are any Intel engineers on this thread saying, whelp if they can do 6 we may as well give up? Or does the lack of any data help you sleep soundly and skeptically?

I think security was not fixable, at-least not to the user's and browsers liking. You had to disable flash for safety and stopping the most annoying ads.

The other issues could be left to consenting adults. But browsers and IT departments turned against plugins in general. Now even Java is almost gone (from browsers) for the same reason.

Yes, but the embedded platforms that are interesting for high-level scripting are mostly phones. Worse than desktops, they won't bother with 32-bit compatibility layers.

Using the other meaning of embedded (I embedded a Lua script in my C application), compatibility libraries don't help. It simply won't link.

I looked over the code to guage how big this 32-bit problem is. It's going to be a lot of work. 32-bit is hardcoded everywhere. Compiler, linker, libc calls.

So, on one hand, there's massive cross-platform support. On the other hand, those platforms are all obsolete, or about to be. Even the mobile and embedded ones have all gone 64-bit.

Maybe backwards compatibility will be maintained for a few years. But it ruins the embed-red-in-C story already.

> I've spotted tens of times more ticks on me than I've actually been bitten

How many total is that?

Never mind. You convinced me to never go outside, even if you were advocating the opposite.

Is Linyos Torovoltos a joke name?

What's the application for DBM? Sounds like a way to trace and JIT already-compiled programs for performance reasons, maybe to take advantage of newer CPU features than the original compiler could target, or maybe to port between platforms.

Is any of that close?

Emacs offers to let you edit ASTs in Paredit mode for lisp if you want to experience it.

Side question: Do you use your Yubikeys for GPG? I tried to make two identical keys with the same GPG key, but still I get "Please remove the current card and insert the one with serial number: ..." if I try to decrypt a file with the other key. I asked the internet a couple times but no one seems to know.

Does it matter a lot for this sort of application (which I assume is interactive data analysis)? I think if you're worried about performance then you want a back-end to map your computation out to a big cluster, no? Like, write OCaml, target hadoop would be the logical thing if you need HPC. Of course someone will have to write that compiler (not it). :P

Please tell me about the parallelism problem? I love that Haskell can auto-parallelize and is very safe. But, I'm ready to kick it to the curb over stupid lazyiness-induced memory explosions. But I'd be sad to go back to having to worry about threads and safety.

I'd love to give OCaml a go. Tell me why I'll end up disgruntled.

I find this design clean and appealing. Compared to the current fashion of all these Python libraries, it's a breath of fresh air. But I wonder if it can work. Python has going for it: kitchen sink of tools, easy to learn, tooling that took 20 years to develop, you already know it.

Question: What's wrong with Matlab? Is it that it's closed and costs money? Other than that it seems clean, purpose-built, complete and not really expensive. Why all the open-source attempts to replace it?

I really thought Julia was going to do that, but it's been years and they never got the tooling. They never solved the ridiculous, self-inflicted problem of having to recompile the entire system every time you run it. You're code will execute in 300ms, but only after 10 minutes of compiling. (Or did they make progress? I haven't checked in in a while.)

Anyway, serious question: Why not Matlab?

And like, why? To save power? Money? Performance? Maybe run some benchmarks and compute the bottom line.

One idea that I think gives the best of both worlds is to use lisp to generate your C++ (or whatever) boilerplate instead of dealing with the maddening vaugeries of template metaprogramming, boost, STL, etc to achieve genericity.

Wow. So damp. Someone's going to get electrocuted. Where's Chinese OSHA?

Ha! Now I know.

You, sir, are a badass. You should sell wifi power cycling sockets. I will be your first customer.

Agreed. But more specifically, how can VCs be more discriminating and also not make getting money harder?

There was another thread today about an interview with some rockstar VC and his wisdom was: invest in ideas that sound stupid, but turn out not to be. I mean, here I've been trying to not sound stupid in front of investors and it's been holding me back.

I don't know about SGX but worked on HDCP and maybe the license charge is for the same reason: lazy or malicious partners could leak a key and compromise the whole system. All the charges pay into the fund to pay damages to the injured if the system suddenly collapses like DVD encryption did.

Classy. I've seen too many situations where the old man wouldn't die or relinquish power but also wouldn't contribute anything but obstruction. There should really be more of a mentorship model rather than the assumption that a director will be around indefinitely.

Yeah, I've been working on LLVM lately and it seems like they've solved all the problems. Somehow, they solved them in standard C++. I've been through a bunch of compilers and I could barely hang on. In LLVM, everywhere I go, I'm like: oh, so that's how you do that.

One thing that helped get started was slides from a CMU class. Hang on:

https://www.cs.cmu.edu/~15745/syllabus.html

Excellent work and so nice of them to post it for free.

We're still doing Kickstarter? Thought we all got burned on that AR/wifi deal and never went back.

Hadn't heard about Hello. Their product sucked but I assumed PT was in for more than one hand to prove his point.

But they can sell 1 FPGA used to prototype the machine-learning ASIC.