• (disco) in reply to PleegWat

    I'm talking about the "free lunch", and that so many people/tools seem to think it's still valid.

    In the past, you'd write your code and it'd be fine at initially expected usage. When usage grew beyond original expectations, you'd buy a new, faster machine and bingo, it's fast enough again.

    That ended around 5 years ago. CPUs are not getting faster, you can simply buy more of them for the same money/size/power budget.

    If you can't write concurrent code, then you can't use those extra cores, and your code will not be able to go faster on a new machine.

    The "brute force" approach means a lot of additional overhead and will still fail if the code assumed it is only running once.

  • (disco) in reply to lightsoff
    lightsoff:
    CPUs are not getting faster

    That's not true, FWIW. It's just that the speed differences have become a lot smaller for most code. What's happened is that the number of clock cycles per instruction has dropped on average, mostly because of things like the amount of transistor budget for things like floating point multiplies has become larger, allowing those to become faster. If your code is multiply-heavy, that makes a big difference. There's also been quite a bit of success with integrating SIMD techniques into standard code (NB: this covers everything from MMX to GPUs); programs that can use that have accelerated a lot.

    But a lot of general code just has not gained all that much. The real focus of recent developments has been on improving parallel code. Fortunately, multi-process applications can also take advantage of this sort of thing; the PHP fork technique is not actually all that terrible after all. Of course, often what happens is that people use the extra capabilities of the CPU to run VMs; though the individual processes aren't faster, the hardware utilisation is greatly improved…

  • (disco) in reply to dkf
    dkf:
    Why do you think I'm interested in the points you're trying to make?
    Because you replied to me? Quoted even?
  • (disco) in reply to Gaska
    Gaska:
    Because you replied to me? Quoted even?

    http://i2.kym-cdn.com/photos/images/original/000/276/379/103.jpg

  • (disco) in reply to aliceif
    aliceif:
    PHP not being the WTF is TRWTF!

    Just a matter of time with this one.

  • (disco) in reply to Kian
    Kian:
    repo

    I keep reading that as "repossession" instead of "repository". Need to cut back on Gabriel Iglesias videos, I guess...

  • (disco) in reply to redwizard
    redwizard:
    I keep reading that as "repossession" instead of "repository".

    https://www.youtube.com/watch?v=DLGrXGEMOSo

  • (disco) in reply to dkf
    dkf:
    http://i2.kym-cdn.com/photos/images/original/000/276/379/103.jpg
    That's racist.
  • (disco) in reply to CoyneTheDup
    CoyneTheDup:
    I really can't agree. Based on some of the stories we've seen in here, it doesn't look like it works very well grow a boss from inside the company, either.
    kupfernigk:
    that is not designed from the ground up to fit a company.

    My choice of words was careful: I wrote "designed from the ground up". Growing from inside is not the same thing, otherwise cancer would be just what you wanted. The word "designed" is important. There are a few companies that select promising graduates or postgrads, have a careful management training system including plenty of outside education, and reward success. Rolls-Royce, Honda, Toyota, VW, Boeing for instance. Sadly, software design still is not engineering.

Leave a comment on “The Backend”

Log In or post as a guest

Replying to comment #:

« Return to Article