Perhaps we should change m5.simulate() to take an absolute tick number
rather than a delta... it was really an arbitrary choice on my part, and
now it looks like maybe it wasn't the best one...
Steve
Stewart, Chris (Solutions Alliances) wrote:
Never mind -- I found my own problem (which I had previously fixed in a
different implementation). In fs.py, the event loop can sometimes
generate a num_cycles value less than zero. I solved this by replacing
the event loop with the following code:
while exit_event.getCause() == "checkpoint":
m5.checkpoint(root, "cpt.%d")
if maxtick == -1:
exit_event = m5.simulate(maxtick)
else:
exit_event = m5.simulate(maxtick - m5.curTick())
This accounts for a test being run without a maxtick specified (i.e. run
until complete), which sets maxtick to be -1. If you don't address this
issue, then the test will exit with a fatal error after the first
checkpoint (because maxtick - m5.curTick() will invariably be a negative
value).
Chris Stewart
------------------------------------------------------------------------
*From:* [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]
*On Behalf Of *Stewart, Chris (Solutions Alliances)
*Sent:* Tuesday, October 17, 2006 11:14 AM
*To:* [email protected]
*Subject:* [m5-users] M5stats.txt
I'm running into a problem running the v2 Specweb benchmark. The
benchmark works fine and completes fine, but when it completes on its
own, I get a zero-length m5stats.txt file. However, when I ctl-C out of
the test before it completes, I get data written to m5stats.txt. I've
included an explicit command in the client-side rcS file, m5 dumpstats,
to get these stats out, but I'm confused as to this behavior.
Chris Stewart
------------------------------------------------------------------------
_______________________________________________
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
