working with non-powers of 2 might throw such issues.
Yes you can comment this out … But not sure if it’ll throw up more. Good luck.

> On Aug 23, 2017, at 9:37 PM, [email protected] wrote:
> 
> Dear Professor,
> Thanks for your reply.
> But I wonder if mesh_row=3 can work or not? Can I set mesh_row with 9 cpus 
> and 16 dirs? Because I encountered the errors as assert(node.type == 
> 'DMA_Controller')
>  assert(num_columns * num_rows == num_routers)
> 
> Or can I comment out these assertion?
> Thanks in advance.
> Best,
> Evania
> 
> 2017-08-23 5:31 GMT+08:00 Tushar Krishna <[email protected] 
> <mailto:[email protected]>>:
> gem5 needs directories to be powers of 2 unfortunately.
> One hack is as follows - 
> With mesh_row = 3, I would suggest still giving num_dirs as 16, multiple 
> directories will just get connected to the same node.
> And in the traffic generator (http://www.gem5.org/Garnet_Synthetic_Traffic 
> <http://www.gem5.org/Garnet_Synthetic_Traffic>) you can change it to only 
> send up to dir 9 instead of till 16 … 
> 
> Cheers,
> Tushar
> 
> 
> 
>> On Aug 22, 2017, at 5:25 PM, [email protected] 
>> <mailto:[email protected]> wrote:
>> 
>> Dear Professor,
>> May I inquire about why can't set dir_num = 9? even in mesh_row =3?
>> I got this error as follows,
>> fatal condition intlvBits && intlvMatch >= ULL(1) << intlvBits occurred: 
>> Match value 8 does not fit in 3 interleaving bits
>> 
>> Thanks for your help.
>> Best,
>> Evania
>> 
>> 2017-08-18 4:19 GMT+08:00 Tushar Krishna <[email protected] 
>> <mailto:[email protected]>>:
>>> May I ask  is it possible to set every data_msg_size differently when 
>>> generating packet each time?
>> 
>> 
>> You can do it in the NI then as thats where packets are being generated.
>> Right now the packet size is determined by the message type from the 
>> coherence protocol.
>> But nothing stops you from generating packets of different sizes (i.e., 
>> different number of flits) inside the NI.
>> Take a look at the flitisize function in the NI.
>> 
>>> And is ni_flit_size equal to rubysystem::getBlockSizeBytes()?
>> 
>> 
>> No. The block size is the cache line size (64B) say.
>> The ni_flit_size is just the link width.
>> 
>> Data msg size typically is control_msg_size + 
>> rubysystem::getBlockSizeBytes()?
>> 
>> 
>> 
>> In that case, you can do it in 
>> 
>>> On Aug 17, 2017, at 4:07 PM, [email protected] 
>>> <mailto:[email protected]> wrote:
>>> 
>>> Dear Professor,
>>> Thanks for your reply.
>>> May I ask  is it possible to set every data_msg_size differently when 
>>> generating packet each time?
>>> And is ni_flit_size equal to rubysystem::getBlockSizeBytes()?
>>> Thanks for your kind help.
>>> Best,
>>> Evania 
>>> 
>>> 2017-08-18 3:36 GMT+08:00 Tushar Krishna <[email protected] 
>>> <mailto:[email protected]>>:
>>> Take a look at
>>> http://www.gem5.org/Garnet2.0#Configuration 
>>> <http://www.gem5.org/Garnet2.0#Configuration>
>>> 
>>> control_msg_size, data_msg_size, and ni_flit_size are the parameters you an 
>>> play with for what you want to do.
>>> 
>>> Cheers
>>> Tushar
>>> 
>>> 
>>>> On Aug 17, 2017, at 1:22 PM, [email protected] 
>>>> <mailto:[email protected]> wrote:
>>>> 
>>>> Dear Professor,
>>>> I would like to inquire about how to set the packet size from garnet2.0 
>>>> network tester.
>>>> From the network tester, I know there are two types of message data and 
>>>> control and respectively for 1 flit and 5 flit.
>>>> Is there any possibility that I could change these length?
>>>> Or there is only one change to implement it in networkInterface.cc 
>>>> <http://networkinterface.cc/>?
>>>> Then may I inquire about how can I carry the packet size information there?
>>>> 
>>>> Thanks in advance.
>>>> Best,
>>>> Evania
>>> 
>>> 
>> 
>> 
> 
> 

_______________________________________________
gem5-users mailing list
[email protected]
http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users

Reply via email to