Re: [Gems-users] A few doubts regarding garnet, gems and errors faced during simulations


Date: Sun, 11 Oct 2009 10:22:58 -0400
From: Niket Agarwal <niketa@xxxxxxxxxxxxx>
Subject: Re: [Gems-users] A few doubts regarding garnet, gems and errors faced during simulations
All the shortest path calculation is done at the level of nodes which are L1s, L2s and directories. Even in the routing tables in the routers, the information is stored at the level of nodes. In Garnet, there is a NIC corresponding to each node and hence the numbers in the matrices/routing tables are for nodes as well as NICs.

Niket

On Oct 9, 2009, at 8:37 PM, Joshi,Madhura R wrote:

Hi,
Thanks for the reply Niket. I am not trying to create my own
messagebuffer class but the ordering property was set to false,
true ordering was not needed at that point on simulation.

I am trying to play with netdest, as I want to change the
destination of message.I am trying to decide between the
destination points when a packet arrives at routing unit and route
the packet accordingly.Destination points are decided based on
current destination in the flit message.Whenever using this scheme
, my packets never reach destination/ a few packets reach
destination but system faults after that.Destination gets changed
appropriately as I can see the change in flit but still I face the
error. (I verified my approach using code in NetworkInterface
class -flitisizing the message part)

I have a question regarding the 'dist' matrix used in topology to
hold shortest route values from a source to a destination.If there
are 16 nodes in system , the matrix size is not 16*16 but much
larger than that.
If each number in source place corresponds to a unique source;
which places/numbers correspond to which routers/ network
interfaces? I think the table gives total view of the system
topology but I am not able to make out the candidates represented
by the numbers.
Can anyone please elaborate/give pointers on this part?

Thanks in advance.
Sincerely,
Madhura



_______________________________________________
Gems-users mailing list
Gems-users@xxxxxxxxxxx
https://lists.cs.wisc.edu/mailman/listinfo/gems-users
Use Google to search the GEMS Users mailing list by adding "site:https://lists.cs.wisc.edu/archive/gems-users/ " to your search.


[← Prev in Thread] Current Thread [Next in Thread→]