I think that kevin may be on to something, it may be because in 1.1 stats
were not reset between the 1e9 period and the 200e6 period. I believe
that they are now reset.
What changes did you make to the ethernet dev file? If you only get 2k
packets, maybe it's only during the 200e6 period that it's actually running
(the 1e9 period is much longer). If you can figure out when packets stop
transmitting, you can try using the various Ethernet trace flags (see
base/traceflags.py) to try to track what's going wrong with the Ethernet
device by outputting a trace around that time. To start tracing at a
specific cycle, add "--Trace.start=<cycle num>" to the command line, and use
"--Root.max_tick=<max cycle num>" to limit your simulation length.
I hope this helps,
Kevin
Zhang, Yu wrote:
Hi all,
I'm using M5_1.1 version. Under original condition, when I'm running
netperf-maerts benchmark, and do the sampling with 'Sampler(periods = [1e9,
200e6])', it always gives me about 67k packets transmitted during this
period. However, after I modified the ethernet dev file a little bit, it
only gave me about 2k packets transmitted, and there is almost no
statistics in the output file. It seems like nothing is excuting during the
200e6 cycles. How to figure out this problem?
Thanks,
Yu
------------------------------------------------------------------------
_______________________________________________
m5-users mailing list
[email protected]
http://m5sim.org/cgi-bin/mailman/listinfo/m5-users
_______________________________________________
m5-users mailing list
[email protected]
http://m5sim.org/cgi-bin/mailman/listinfo/m5-users
