News for hackers. Stuff that matters.
Re:About speculative execution vulnerabilities in ARM-based ... by Strom on Friday January 05, 2018 @07:52AM (#16078251)
You should try defining this mystical true benchmark that you seek.

Apple decided that heavy userspace execution and web performance are representative of common workloads. You are welcome to argue for a different kind of workload. As it stands now, at best you're implying that it's common for users to execute syscalls in a loop.

Re:About speculative execution vulnerabilities in ARM-based ... by HugoDaniel on Saturday January 06, 2018 @01:06PM (#16086566)
Can you tell with any degree of precision what kind of benchmark the "GeekBench" is and does ? What guarantees do you have that it is testing "real world common workloads" ?

Do you think common users execute guassian blurs in a loop and ray tracers before perfoming rigid body physics in their "common workloads" ? As you said, I know I am welcome to argue for a different kind of workload, specially when I am being taken for a fool.

Re:About speculative execution vulnerabilities in ARM-based ... by Strom on Saturday January 06, 2018 @02:45PM (#16087052)
I'm not too familiar with GeekBench, but I'm assuming it's userspace code. Whether it traces rays or does something else isn't that important, because the slowdown is with syscalls.

As for what common users execute, based on my observations over the years it's mostly web browsing. However my personal vision of common workloads is a distraction. I have made no claims to know common workloads. I merely stated that Apple has chosen a particular set to be representative.