The default L1 block size is 64 bytes.  Do you get errors if you
change the cache size without changing the block size?

I realize that some real machines have different block sizes at
different cache levels, but that's not officially supported in M5, so
I wouldn't be surprised if having different block sizes is the source
of your problem.

Steve

On Wed, Jun 9, 2010 at 5:10 PM, Wang, Weixun <[email protected]> wrote:
> 1) I am using the latest version and I did not change anything other than
> the cache size.
>
> 2) For the default case, I use 32B block size for L1 and 64B for L2. It
> works fine. The errors are produced due to change in cache size. I did not
> change block size. Furthermore, it is common for L2 to have a larger block
> size than L1
>
> Any help will be appreciated.
>
> Thanks.
>
> -Weixun
>
> On 2010/6/9 20:06, Steve Reinhardt wrote:
>>
>> Just changing cache sizes should not affect the results.  Are you
>> using the latest m5 from the development repository?  Have you changed
>> anything other than the cache sizes?
>>
>> I noticed you said the default L1 cache is 32B, but the default block
>> size is 64, and all caches must use the same block size.
>>
>> Steve
>>
>> On Wed, Jun 9, 2010 at 4:23 PM, Weixun Wang<[email protected]>  wrote:
>>
>>>
>>> Hi all,
>>>
>>> I'm trying to run selected benchmarks from MiBench in M5. The system is
>>> configured into multi-core (CMP) mode with each core has its own L1
>>> caches
>>> and all cores share a unified L2 cache. Each core is assigned a different
>>> benchmark. However, I found the following error:
>>>
>>> When I use the L1 cache configuration of (32kB, 2 assoc, 32B) and L2 of
>>> (4MB, 8 assoc, 64B), which are the default values, everything is OK.
>>> However, when I try to change the cache configuration, for example, L1
>>> with
>>> a size of 8kB and L2 with 128kB or L1 with 16kB and L2 with 1MB, the
>>> simulation terminates with the error:
>>>
>>> 0: system.remote_gdb.listener: listening for remote gdb on port 7000
>>> 0: system.remote_gdb.listener: listening for remote gdb on port 7001
>>> panic: Tried to execute unmapped address 0.
>>>  @ cycle 1494306000
>>> [invoke:build/ALPHA_SE/arch/alpha/faults.cc, line 186]
>>> Memory Usage: 119756 KBytes
>>> For more information see: http://www.m5sim.org/panic/95542d88
>>> Program aborted at cycle 1494306000
>>> Abort
>>>
>>> Am I missing something?
>>>
>>> Thanks.
>>>
>>> --
>>> Best Regards,
>>>
>>> Wang, Weixun
>>>
>>> _______________________________________________
>>> m5-users mailing list
>>> [email protected]
>>> http://m5sim.org/cgi-bin/mailman/listinfo/m5-users
>>>
>>>
>>
>> _______________________________________________
>> m5-users mailing list
>> [email protected]
>> http://m5sim.org/cgi-bin/mailman/listinfo/m5-users
>>
>>
>
>
> --
> Best Regards,
>
> Wang, Weixun
>
> Department of Computer&  Information Science&  Engineering
> Gator College of Engineering
> University of Florida
> Gainesville, FL 32611
> http://www.cise.ufl.edu/~wewang
>
> _______________________________________________
> m5-users mailing list
> [email protected]
> http://m5sim.org/cgi-bin/mailman/listinfo/m5-users
>
_______________________________________________
m5-users mailing list
[email protected]
http://m5sim.org/cgi-bin/mailman/listinfo/m5-users

Reply via email to