Re: PC386 boot problems (solved)

[Quality Quorum <qqi@world.std.com>]

On Mon, 6 Dec 1999, Joel Sherrill wrote:

> VALETTE Eric wrote:
> > 
> > >>>>> "Quality" == Quality Quorum <qqi@world.std.com> writes:
> > 
> > Quality> On Mon, 6 Dec 1999, VALETTE Eric wrote:
> > Quality> This loop is called once during start up, so the the point should be
> > Quality> a reliable minimum delay time (this one will give you 1us on 1Gz
> > Quality> processor), versus precision achived through a strange trick.
> > 
> > >> Could the guy that suggested the change test with 0x80?
> > 
> > Quality> The problem I found was that start16.s does not have enough delay
> > Quality> time while doing Gate 20 operation. I just changed ldsegs.s along
> > Quality> same lines, hence '%'.
> > 
> > Proabbly untested since it does not compile...
> 
> The does not compile part is my fault from merging the patch by hand.
>
> 
> > Furthermore :
> > 
> >    1) Linux does otherwyse and has been much more tested than RTEMS on pc386
> 
> This part is hard to argue. :)

There are some spots even on the sun.

 
> >    2) Your loop does not work on our PC...

And your loop does not work on mine ...


If delay from ldsegs.s is used outside start.s and ldsegs.s then it 
is a bug in a caller. Rationale: delay function in ldsegs.s is used 
before there are any chances to do processor speed calibration.
I suppose that once processor speed is known, then there is a full
possibility to use calibrated loop and it sould be used (and written
if there is none).
 
   
> > 
> > -- eric
> 
> -- 
> Joel Sherrill, Ph.D.             Director of Research & Development
> joel@OARcorp.com                 On-Line Applications Research
> Ask me about RTEMS: a free RTOS  Huntsville AL 35805
>    Support Available             (256) 722-9985
> 

Thanks,

Aleksey