Re: 25 microseconds?
- From: "Gary G. Little" <gglittle.nospam@xxxxxxxxxxxxx>
- Date: Mon, 11 Apr 2005 14:35:33 GMT
There is a "rule of thumb" I have used for decades, and I have heard touted
in every seminar I have sat in --- you take as much time in your ISR as you
need; no more and no less. Draining the serial FIFO in the ISR is by better
than sipping one character at a time since each character in the FIFO is
going to cause another interrupt to be thrown which will incur additional
overhead. Moving all of the data from the FIFO to a software ring buffer
takes much less time than incurring interrupt overhead for every character.
--
The personal opinion of
Gary G. Little
"Richard Lee" <richard.lee@xxxxxxxx> wrote in message
news:1112893167.700666.244360@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
> Just my opinion, the DPC is not a magic bullet.
> Both ISR and DPC execution can kill performance.
> I guess that's why Microsoft limits both.
>
> I'm the last person to defend LPC.
> OEM's like it because it's cheap. So I support it.
>
> Anyway, I better quit while I'm ahead. I saw an old
> posting from Eliyas Yakub where he said:
> ISR - 10 microseconds
> DPC - 25 microseconds
> Yikes!
>
>
.
- References:
- 25 microseconds?
- From: Richard Lee
- Re: 25 microseconds?
- From: Pavel A.
- Re: 25 microseconds?
- From: JimE
- Re: 25 microseconds?
- From: Alexander Grigoriev
- Re: 25 microseconds?
- From: Pavel A.
- Re: 25 microseconds?
- From: Richard Lee
- Re: 25 microseconds?
- From: Ray Trent
- Re: 25 microseconds?
- From: Alexander Grigoriev
- Re: 25 microseconds?
- From: Richard Lee
- 25 microseconds?
- Prev by Date: Re: HCT-machine frozen
- Next by Date: Re: printui.dll,PrintUIEntry internals...
- Previous by thread: Re: 25 microseconds?
- Next by thread: how to Sync Irp->AssociatedIrp.SystemBuffer when Overlapped I/O.
- Index(es):
Relevant Pages
|