Sim: Add functionality to the simulation scripts to allow running with
Review Request #57 - Created July 9, 2010 and updated
Information | |
---|---|
Timothy Jones | |
gem5 | |
Reviewers | |
Default | |
Sim: Add functionality to the simulation scripts to allow running with SMARTS sampling methodology. At the moment the scripts don't support the full range of options (e.g. you can't start from a checkpoint), and nor do they check for incompatibility with other options. However, they do give an example of how sampling could work with M5. With a benchmark that has 10 billion dynamic instructions, try: <m5bin> configs/example/se.py -d --caches --smarts:n 10000 --smarts:b 10
Posted (July 11, 2010, 9:04 a.m.)
-
configs/common/Options.py (Diff revision 1) -
Why exactly do these parameter names have a colon in them? Is there a particular reason that --smarts-k doesn't work? Also, what is the k value? Is there some more descriptive name that you can give to these parameters in general? i.e. --smarts-warmup, --smarts-measure --smarts-samples, etc.
-
configs/common/Smarts.py (Diff revision 1) -
in python 2, you're going to get integer division and the rounding will happen multiple times. If a float value is not OK, then you should probably do the math as a float and round/floor it afterwards.
Posted (July 16, 2010, 11:23 a.m.)
This revision should fix the comments made in the previous review.
Posted (July 16, 2010, 12:13 p.m.)
-
configs/common/Options.py (Diff revision 2) -
You don't have to change this if you don't feel like it, but the dest="" thing isn't necessary. --smarts-sample-interval becomes options.smarts_sample_interval. I prefer that they're consistent myself. The only real reason (I think) to use dest= is if you have multiple options that affect the same variable.