qemu PowerPC BSP?

Till Straumann strauman at slac.stanford.edu
Fri Sep 18 15:58:45 UTC 2009


Joel Sherrill wrote:
> Till Straumann wrote:
>> Joel Sherrill wrote:
>>  
>>> Hi,
>>>
>>> I am curious if we already have a BSP that will run on
>>> the PowerPC simulator provided by qemu.
>>>
>>> $ qemu-system-ppc -M -?
>>> Supported machines are:
>>> g3beige    Heathrow based PowerMAC (default)
>>> mac99      Mac99 based PowerMAC
>>> prep       PowerPC PREP platform
>>>     
>> I recently tried the mvme2307 (which is PreP).
>> IIRC I had to fix and hack quite a few things (in qemu)
>> to get it to boot but gave up after it seemed that
>> the OpenPIC didn't work out of the box.
>> Could probably be hacked, too w/o too much effort
>> but I ran out of time.
>>
>>   
> Do you think the mtx603e variant might work?
Honest answer: I don't know. However, I suspect that it
won't out of the box anyways. You'll probably have to hack
qemu and maybe even the openbios (I had to do both to
get the 2307 to boot).

T.
> Wasn't that a no bus version of the motorola_powerpc?
>> IMO it would be useful to be able to use qemu
>> but we should not abandon PSIM. The latter does
>> have (quite buggy but I have filed a patch fixing
>> many things) ALTIVEC emulation which qemu
>> unfortunately lacks.
>>
>>   
> There is no intention to drop PSIM.  :)
>> OTOH the networking support of qemu is far
>> superior.
>>
>>   
> And qemu is being enhance to support coverage analysis
> at the instruction and branch taken/not taken level.  It
> already supports this on the PowerPC, x86, and leon2.  I
> am trying to get coverage results on x86 and was hoping
> that the PowerPC had a prayer of working.  That would
> enable us to have object coverage testing on more CPUs.
> There will be a more formal announcement in the near
> future but here is the status of coverage testing on
> SPARC and ARM.
> http://www.rtems.com/ftp/pub/rtems/people/joel/coverage/
>
> The subset of code approaching 100% instruction coverage
> is the cpukit directories posix, sapi, rtems and score when
> at -Os.  I want to get to 100% at -O2 and remove the few
> conditionals that are needed.  But comparing coverage
> percentage from 4.7 to now, it is gone from ~75% to 99+%
> on the same code at -Os.
>
> So for me, PowerPC on qemu is primarily for coverage and
> just more to test is better. :)
>
> --joel
>> T.
>>  
>>> ref405ep   ref405ep
>>> taihu      taihu
>>> bamboo     bamboo
>>> mpc8544ds  mpc8544ds
>>>
>>> The taihu is an AMCC 405EP reference platform.
>>> The bamboo is an AMCC 440EP reference platform.
>>>
>>> Shouldn't we have some BSPs that work on some of
>>> those already?
>>>
>>>     
>>
>>   
>
>




More information about the users mailing list