4 ms·

Hm, github.com's diff doesn't let you ignore changes in whitespace? Bummer. Anyway that's not basis enough to invalidate their claims. They're published in E

Hm, github.com's diff doesn't let you ignore changes in whitespace? Bummer. Anyway that's not basis enough to invalidate their claims. They're published in EuroSys [1]. Is that not a reputable journal/conference? Plus this dude has a sweet CV layout [2]. I'm inclined to believe him on that basis alone. [1] http://www.i3s.unice.fr/~jplozi/wastedcores/ http://www.i3s.unice.fr/~jplozi/wastedcores/ [2] http://www.i3s.unice.fr/~jplozi/ http://www.i3s.unice.fr/~jplozi/

kazinator 11y ago
Fact is, the "dude" doesn't know how to get his text editor to use the same tabbing as the kernel code he's editing, and doesn't know how to run a patch through the kernel's "checkpatch.pl" script, without which it will be rejected from upstreaming. The idea that the kernel runs idle tasks on cores while actual tasks are runnable (and that this situation persists for seconds) is ridiculous. It's such a gaping problem, that everyone doing anything with Linux anywhere would be running into it on a regular basis. Think about all the people out there who carefully profile the performance of their multi-threaded apps. Like all of them wouldn't notice missing seconds of execution?
- studentrob 11y ago
  I've not studied this branch of CS since college a decade+ ago. As far as I know, certain elements of the kernel aren't modified very often. Also the researcher claims to introduce new tools that didn't exist before. Again I think if the journal, EuroSys, is reputable, then that journal would vet the findings properly. Replicating people's work is standard procedure. Maybe people need time to attempt that. Right on if you are proven right!
  - kazinator 11y ago
    Certain elements of the kernel aren't modified often, but the scheduler isn't one of them. It had been subject to a lot of churn. There is no generalization you can state about the Linux kernel scheduler that is valid for an entire decade of kernel versions.
    - studentrob 11y ago
      > There is no generalization you can state about the Linux kernel scheduler that is valid for an entire decade of kernel versions That sounds like a generalization about the Linux kernel scheduler for an entire decade of kernel versions =) (just kidding haha!)
otterley 11y ago
It does (add ?w=1 to the query parameters), but this is a file containing a diff, not a GitHub diff.

kazinator 11y ago

Fact is, the "dude" doesn't know how to get his text editor to use the same tabbing as the kernel code he's editing, and doesn't know how to run a patch through the kernel's "checkpatch.pl" script, without which it will be rejected from upstreaming. The idea that the kernel runs idle tasks on cores while actual tasks are runnable (and that this situation persists for seconds) is ridiculous. It's such a gaping problem, that everyone doing anything with Linux anywhere would be running into it on a regular basis. Think about all the people out there who carefully profile the performance of their multi-threaded apps. Like all of them wouldn't notice missing seconds of execution?

studentrob 11y ago
I've not studied this branch of CS since college a decade+ ago. As far as I know, certain elements of the kernel aren't modified very often. Also the researcher claims to introduce new tools that didn't exist before. Again I think if the journal, EuroSys, is reputable, then that journal would vet the findings properly. Replicating people's work is standard procedure. Maybe people need time to attempt that. Right on if you are proven right!
- kazinator 11y ago
  Certain elements of the kernel aren't modified often, but the scheduler isn't one of them. It had been subject to a lot of churn. There is no generalization you can state about the Linux kernel scheduler that is valid for an entire decade of kernel versions.
  - studentrob 11y ago
    > There is no generalization you can state about the Linux kernel scheduler that is valid for an entire decade of kernel versions That sounds like a generalization about the Linux kernel scheduler for an entire decade of kernel versions =) (just kidding haha!)

otterley 11y ago

It does (add ?w=1 to the query parameters), but this is a file containing a diff, not a GitHub diff.