2012/2/24 Ouabache Designworks <[email protected]>:
>
>
> On Fri, Feb 24, 2012 at 6:58 AM, R. Diez <[email protected]>
> wrote:
>>
>>
>> I also noticed that, although the top-level test driver is a makefile, it
>> breaks when running in parallel mode (when invoking GNU Make with option -j
>> N). The tests must be run sequentially, so that only one CPU ever gets busy.
>>
>> Regards,
>>   R. Diez
>> _______________________________________________
>> OpenRISC mailing list
>> [email protected]
>> http://lists.openrisc.net/listinfo/openrisc
>
>
>
> I don't think that you can run the suite in parallel because all the sims
> run in the sim/run directory and they all use the same sram.vmem file to
> link to their rom image file.
>
> To fix this you would need to create a separate subdirectory for each sim
> AND pass the filename for the rom image into the sim as a parameter.  That
> way you can also support multicore designs that have more than one rom image
> file. It also eliminates the need for the /sim/out directory since all log
> and vcd dump files are stored inside a separate directory.
>
> Thats what I do in my design environment. I create a sim/icarus directory
> and underneath that I have a separate subdirectory for each simulation. You
> want to give each tool its own separate space in case you ever want to use
> more than one at a time. I don't link the rom image file but rather pass the
> relative path name into the sim. The bits only live in the /sw directory
> where they were compiled.
>
> With all the multicore processors out there this would be a useful change to
> make for orpsocv3.
>
> You also want to break the dependency where each sim is locked in with a
> software test and you call the sim by calling the software test. I do sims
> where one piece of software is used by many different sims. I can change the
> parameters for each sim as well and stimulus and responses on a sim by sim
> basis. You need that flexibility if you ever want to build a complete test
> suite.
>
>
>
> John Eaton
>
>
>
> _______________________________________________
> OpenRISC mailing list
> [email protected]
> http://lists.openrisc.net/listinfo/openrisc
>

Yes, this should be fixed in ORPSoCv3. The structure allows for it,
but I haven't implemented proper test support yet, so I will put a
note on the wiki for now.

-- 
Olof Kindgren
______________________________________________
ORSoC
Website: www.orsoc.se
Email: [email protected]
______________________________________________
FPGA, ASIC, DSP - embedded SoC design
_______________________________________________
OpenRISC mailing list
[email protected]
http://lists.openrisc.net/listinfo/openrisc

Reply via email to