[Gems-users] Creating a new topology network

[Blas <blacuesa@xxxxxxxxxx>]

Hello.

I want to implement the Perfect Shuffle (or Butterfly) topology in GEMS and I 
am not sure if I have done it in a good way or there is a better method than 
I should have used. Basically, I have only changed the Topology.C file:

  1.- I have created a new topology type, and I have added it to the       
"Topology::init( )" function.

  2.- I have created a new function ("makeSwitchesPerChipButterfly"). This 
function is similar to "makeSwitchesPerChip". I think this function assigns 
one node (and other components) to every switch, but in the butterfly network 
there are some switches which have not assigned any node. Therefore, 
"makeSwitchesPerChipButterfly" function does not assigns one node per switch, 
it assigns two nodes to every switch in the first stage (two nodes because I 
want to use 2x2 switches).

  3.- Then, I have added another function ("makePerfectShuffle"), which 
creates the topology.


Now I want to check that the topology I have generated is a perfect shuffle 
topology. I have printed the "nodePairs" variable and everything seems right. 
However, I do not know if the content and format of the variable are right. 
Could anyone help me?

In the ruby configuration, I put "g_PRINT_TOPOLOGY" to "true" and, when I get 
the results, I can see the topology but I do not know to interpret the 
notation of the topology print. I will be grateful if someone could explain 
it to me.

The last question is: Where can I change the switches sizes?

Thanks,

Blas.