[pc386 BSP] tm[01|02].exe doesn't run on K6-2 (raw handler connexion failed issue)

[Joel Sherrill <joel@OARcorp.com>]

Karel Gardas wrote:
> On Tue, 26 Jul 2005, Joel Sherrill <joel at OARcorp.com> wrote:
> 
>> Karel Gardas wrote:
>>
>>>
>>> Hello,
>>>
>>> testing today's sources of RTEMS 4.7, I've found that I'm not able to 
>>> run time meassuring tests on my K6-2/128MB box. All tests so far 
>>> starts well, but then print "raw handler connexion failed" message.
>>>
>>> Any idea how to get this working?
>>
>>
>> The timer driver comes in two style on the pc386 BSP.  If the
>> CPU has the RDTSC instruction, then the CPU cycle counter is
>> used.  Otherwise, a counter/timer is used.
>>
>> If the K6 has the rdtsc instruction, then the conditional
>> in pc386/timer/timer.c just needs to be changed to reflect
>> that it is a supported instruction.  A quick google shows
>> that RDTSC is supposed to work on the K6 so the macro is
>> too tight.
>>
>> Patch welcomed.
> 
> 
> OK, I've also tested original binary on PIII and it shows the same 
> issue. Do you prefer to support dynamic detection of target CPU 
> features? Or just should I simply define `pentium' and be done there?

Ahh .. I kind of see what has happened.  In older gcc versions,
the CPU dependent CFLAGS used specified the architecture level
and tuning.  The ones we are using with gcc 4.x only specify
the tuning level.  So no higher model level CPU instructions
are allowed and the cpp predefines reflect this.

I think that the dynamic detection would be great.

> Thanks,
> Karel
> -- 
> Karel Gardas                  kgardas at objectsecurity.com
> ObjectSecurity Ltd.           http://www.objectsecurity.com


-- 
Joel Sherrill, Ph.D.             Director of Research & Development
joel at OARcorp.com                 On-Line Applications Research
Ask me about RTEMS: a free RTOS  Huntsville AL 35805
    Support Available             (256) 722-9985