RT if you, too, are taking the breathless, sky-is-falling, performance-apocalypse "journalism" being done based on limited information about this Intel bug with several truckloads of salt, and taking a more wait-and-see attitude instead.
-
no, speculation is fun! Seriously though, why the rush job? If it's not "critical" why are they rewriting the entire process memory model?
As I understand it, it involves Intel's speculative execution model, and using side-channel attacks to get at the speculatively executed results. Beyond that, moving the kernel into its entirely own address space is non-trivial and a lot of work.
-
-
Essentially my gripe comes down to people freaking out about "30% performance loss!!!" without any additional info as to what *kind* of code would be affected. Because I really doubt user-land code would be affected at all. 30% slower calls into the kernel? I could believe that.
-
Well, the measured impact is nearer 50% (complete TLB rebuild) on phoronix, for syscall mostly workloads. On a desktop, the 5% number is probably more realistic. AWS and Azure and all those cloud providers? 30% is probably right on the money. That's a lot of CPU to lose.
- Näytä vastaukset
Uusi keskustelu -
Lataaminen näyttää kestävän hetken.
Twitter saattaa olla ruuhkautunut tai ongelma on muuten hetkellinen. Yritä uudelleen tai käy Twitterin tilasivulla saadaksesi lisätietoja.