[Date Prev][Date Next] [Thread Prev][Thread Next]
[Thread Index]
[Date Index]
[Author Index]
Re: Laptop loosing time
- From: Daniel Senderowicz <daniel bicho SynchroDS COM>
- To: redhat-install-list redhat com
- Cc: daniel bicho SynchroDS COM
- Subject: Re: Laptop loosing time
- Date: Tue, 6 Aug 2002 11:35:05 -0700 (PDT)
Rick,
Thanks for the explanation.
>So, either resynchronize the software clock to the hardware clock
>on a regular basis,
What would be the best setup to do this?
Dan
>Daniel Senderowicz wrote:
>> Robert,
>>
>>
>>>'Twould appear that something is eating enough processor time to cause
>>>it to miss clock interrupts, and cause the clock to "run slow". Next
>>
>>
>> I don't understand this. I thought that the hardware clock is a
>> piece of logic (counters) totally independent of the CPU, therefore
>> whether the CPU runs busy or not wouldn't make any difference. What
>> should be running on interrupts is the transfer of the data from
>> the hardware clock registers to the system clock and if the system
>> is busy, then it may miss a few transfers but whenever it gets it,
>> it will be OK, or am I missing something here?
>
>Speaking as a kernel geek, the hardware clock is, essentially, a big
>counter that's clocked by either the CPU clock source or an internal
>crystal). As long as the clock source is stable AND the scaling
>between clock cycles and the counter are correct, it should be fairly
>accurate (the clock runs when power is off via the lithium battery on
>the mobo--it doesn't simply keep your BIOS settings alive).
>
>The software clock is a software counter that's clocked by an
>interrupt, typically the one on IRQ0. If the system spends an
>inordinate amount of time in interrupt handlers of higher priority
>(IRQ0 is the lowest priority), you may miss clock increments.
>
>Many graphic cards make use of IRQ9, normally tripped during vertical
>retrace. If your X driver is inefficient, this can also lead to
>lost clock increments. There are certain DMA operations that can
>cause interrupts to occur, and again, ineffecient coding can lead to
>lost clock increments. And so on and so on.
>
>So, either resynchronize the software clock to the hardware clock
>on a regular basis, or use something like ntp to keep it in line.
>Of the two, I prefer ntp but that depends on your ability to get to
>an ntp server.
[Date Prev][Date Next] [Thread Prev][Thread Next]
[Thread Index]
[Date Index]
[Author Index]