Jeff Duntemann's Contrapositive Diary Rotating Header Image

Odd Lots


  1. KD says:

    The links to the two books in your first item don’t work. (Same problem on the other version of the blog.) It seems that your page’s address is prepended to the Amazon address. When I strip that away, the remaining address works.

    1. I don’t have the blog text source with me here. I’m working off of a crappy netbook while tending my sister’s girls for the weekend. Will fix when I get back to the condo. Thanks for reminding me!

  2. Bruce C. Baker says:

    Larry O’Brien: “I do wish that there were a performance-oriented text to take the place of Michael Abrash’s classic but obsolete books. I wonder if Duntemann could be convinced to write it using Hyde’s HLA?”


    1. In a word, no. I respect HLA highly but I don’t think it’s the right tool for optimization at the Abrash “Zen” level. Furthermore, I don’t think that’s what Randy created it for.

      Given 2010-era CPUs, the whole question of what low-level optimization actually is comes into play. We may make it fast primarily by making it small: See the Atlantis Word Processor. Staying inside cache seems to outweigh almost every other consideration in my experience, and that’s not something you do by looking for faster machine instructions.

      1. David says:

        > Staying inside cache seems to outweigh almost every other
        > consideration in my experience, and that’s not something
        > you do by looking for faster machine instructions.

        Right, and it’s far more about state changes concerning data than code so it’s almost always perfectly accessible from high-level languages now.

        Twenty years ago the difference between a good compiler and a good assembler programmer was about a factor of ten. These days instruction-level tuning is not so important. Optimizing to minimize memory-related machine state changes (tlb, cpu cache, ram) is where the big wins are to be found and the difference is often two or three orders of magnitude in performance.

        It’s amazing that the gap has grown so much. It’s also disappointing that so few people are writing about it.

  3. Jim Tubman says:

    Scientific American did something else that raised a lot of eyebrows: they printed an article about climatologist Judith Curry, who has been respectfully engaging those who are skeptical of immanently catastrophic anthropogenic global warming.

    I mention this not to endorse her views (or any others) on this topic, or to start a discussion about climate change, but simply to note that this sort of thing would not have appeared in SA before. Has there been a change in its editorial staff?

    1. Magazine editors who are not owners of the company are often at the mercy of ad sales management, who field complaints from advertisers (or potential advertisers) and apply subtle and not-so-subtle pressure on editors to make the mag more attractive on the ad side.

      If I had to guess, this would be it. Magazines are getting less and less revenue from paying subscribers, so alienating readers is much less of a hazard than many think. SA did itself in years ago by “dumbing down” content generally, and while it’s possible that somebody in editorial decided that all the usual usual on global warming was turning readers off, I’m more inclined to think that one or more influential advertisers bitched.

      I did see that item and, like you, I marvel.I wouldn’t draw too many conclusions from it, though.

  4. Bruce C. Baker says:

    Sci Am poll: Dr. Judith Curry & climate change: I tried to take the poll but, like Jerry’s correspondent, I agree that “I must say that it was simply the worst, the most biased poll I have ever seen. The choices essentially boiled down to ‘Yes, I am a sensible person who is convinced of climate change’ or ‘No, I am a mindless climate change denier.’ There was never a ‘None of the above’ option for this complex subject.”

    I mourn the loss of the Sci Am of my youth. 🙁

Leave a Reply

Your email address will not be published. Required fields are marked *