> On 2010-07-11 06:47:22, Nathan Binkert wrote: > > Can you describe exactly how this is used? I'm assuming that when you're > > in a sampling period, that you take stats normally, then before you fast > > forward, you save to backup(), you then continue to take stats in the > > primary and when you're done, you restore() and throw away the stats during > > fast forward? That about right? This sounds like a subset of the "bin" > > feature that I took out of the stats package several years ago. > > > > Anyway, I have a ton of changes to the stats package outstanding, so I need > > to think about how this impacts that. I also have one other question, how > > often do you switch from primary to backup? > > > >
Ok, your description is basically correct. SMARTS has three phases: functional warming, detailed warming and measurement. We gather statistics that we want during the measurement phase. At the end (before draining the CPU) we take a backup and, as you say, stats continue to be gathered during functional warming and detailed warming. Then, right before we start measurement again, we restore from the backup. Roughly speaking, you need around 8,000 measurement phases to accurately characterise each benchmark, with one backup and one restore for each. - Timothy ----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: http://reviews.m5sim.org/r/56/#review70 ----------------------------------------------------------- On 2010-07-09 18:21:43, Timothy Jones wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > http://reviews.m5sim.org/r/56/ > ----------------------------------------------------------- > > (Updated 2010-07-09 18:21:43) > > > Review request for Default. > > > Summary > ------- > > Stats: Allow backing up and restoring of stats which is needed for SMARTS > simulation. > > > Diffs > ----- > > src/base/statistics.hh 249f174e6f37 > src/base/statistics.cc 249f174e6f37 > src/base/stats/info.hh 249f174e6f37 > src/python/m5/simulate.py 249f174e6f37 > src/python/m5/stats.py 249f174e6f37 > src/python/swig/stats.i 249f174e6f37 > > Diff: http://reviews.m5sim.org/r/56/diff > > > Testing > ------- > > > Thanks, > > Timothy > > _______________________________________________ m5-dev mailing list [email protected] http://m5sim.org/mailman/listinfo/m5-dev
