You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Sorry if this is a completely n00b question but how can I go about figuring out the e2e latency of a design generated by nngen? In the generated RTL (verilog) I can see the main_fsm and in the timing reports generated by Vivado I can see critical path/delay and etc. but I am not sure how long each step of your FSM takes.
I think I can synthesize one layer at a time and then combine that with the schedule_table in some way but again my question is how to figure out the e2e latency of a layer?
Btw thank you very much for releasing this, I have been struggling for several months to come up with feasible RTL designs for NNs and this package is very informative!
EDIT: actually I guess the n00b answer to my n00b question is to simulate using verilator and multiply the number of cycles times the delay along the critical path.
The text was updated successfully, but these errors were encountered:
Sorry if this is a completely n00b question but how can I go about figuring out the e2e latency of a design generated by
nngen
? In the generated RTL (verilog) I can see themain_fsm
and in the timing reports generated by Vivado I can see critical path/delay and etc. but I am not sure how long each step of your FSM takes.I think I can synthesize one layer at a time and then combine that with the
schedule_table
in some way but again my question is how to figure out the e2e latency of a layer?Btw thank you very much for releasing this, I have been struggling for several months to come up with feasible RTL designs for NNs and this package is very informative!
EDIT: actually I guess the n00b answer to my n00b question is to simulate using verilator and multiply the number of cycles times the delay along the critical path.
The text was updated successfully, but these errors were encountered: