Re: CFS and O(1) scheduler
- From: Michael Schnell <mschnell_at_bschnell_dot_de@xxxxxxx>
- Date: Sat, 08 Dec 2007 12:52:39 +0100
What I find disturbing on that issue is that the software developers and the hardware developers seem to live on different planets.
E.g. the extremely talented (free open source) software developers don't know about the revolutionary "explicit multithreading" CPU design. Instead they concentrate mainly on mainstream hardware like IA32, IA32-64, IA64, PPC and ARM. They even ignore the greatly evolving virtual CPUs in FPGAs (such as NIOS, MicroBlaze and MICO), that are extremely suitable for many embedded projects. There are Linux ports for all of them, but supposedly there is no chance to see them fully supported in the main Linux tree some day soon. I would wish for these CPUs and the IP5K (no Linux port on the horizon yet) to get more acknowledgment in the open source community.
And the extremely talented (commercial) hardware developers (e.g. at Ubicom, maybe not the "inventors" of this CPU architecture, but those who did the first <the IP3K), and up to now single million-selling multithreaded chip) don't build the OS that comes with their SDK on the work of the Linux community, even though the 5K Chip would be able to run µCLinux with great performance and is very suitable for vitalization in order to support "virtual peripherals" (i.e. extremely hard realtime). Instead the 5K SDK comes with their propriety (though provided with all sources) OS. As of next year, same is said to be upgraded to offer a POSIX API. IMHO, a Linux port would be the much better investment.
- Prev by Date: Re: CFS and O(1) scheduler
- Next by Date: Re: CFS and O(1) scheduler
- Previous by thread: Re: CFS and O(1) scheduler
- Next by thread: Re: CFS and O(1) scheduler