Hi Kypros, Yes they should be injected at some point, they will not get thrown away. But you defined your simulation's simCycles to be 100,000 cycles at which point the simulation ended. If you use the --fixed-pkts option to inject a fixed number of packets or flits say 640,000 with injectionRate = 1.0, and if you give a large enough simCycles, then all 640,000 flits will be injected and received eventually.
The "queueing latency" stat in garnet refers to the queueing latency at the source NI (i.e. the time between when the message came from the cache/dir controller to the time when it is actually injected into the source router). The "network latency" stats refers to the complete latency within the network (time spent in buffers at various routers + time spent to traverse crossbars + time spent to traverse links). Both these latencies are updated in the stats after the corresponding flit is received at the destination NI. - Tushar On Aug 2, 2012, at 12:33 PM, Kypros Chrysanthou wrote: > Hello Tushar, > > I understand that due to the congestion the flits are not actually injected > into the network, > what I am not sure is, shouldn't they be injected at some point? since they > are in the buffer. > Is there any point where these flits get thrown away? > > Also, the queuing latency is indeed big, but since it's queuing latency > shouldn't be referring > to the latency of flits being in the queue until they finally get in the > network? > > Thanks again, > Kypros > _______________________________________________ > gem5-users mailing list > gem5-users@gem5.org > http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users
_______________________________________________ gem5-users mailing list gem5-users@gem5.org http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users
