Messages in this thread | | | Date | Thu, 25 Jun 1998 01:42:29 +0200 (MET DST) | From | MOLNAR Ingo <> | Subject | Re: Scheduler fixes |
| |
On Thu, 25 Jun 1998, Chris Wedgwood wrote:
> I guess CONFIG_EXACT_PROCTIMES would make everyone happy (having this > dependent upon the processor selection and perhaps APM).
no, we want to keep the number of kernel config variations as low as possible. There are different conflicting issues here, but mass-ifdefs in the core kernel code is not nice, especially as the ifdef separates two completely different concepts. So the debugging work will be hard. This is not a driver ifdef, it's not a separate kernel feature. This ifdef separates two _competing_ pieces of code. We really dont want to do that, unless justified by other factors.
note that we can do the 'timestamp' trick with non-cycle-counter CPUs too, we just use 'jiffies' instead of cycles. If you think about it, it will give the current system usage metrics ... this way the difference between 'old' and 'new' CPUs is much visible, and only in include files. Think of them as CPUs with a 100 Hz cycle counter :)
of course it's more expensive than doing the sampling in the interrupt handler, and the overhead is not '100 times a second', but 'once per reschedule'.
-- mingo
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu
| |