Hello, I am currently working on introducing a change to the way stats output within gem5. Currently, certain stats output under different names depending on the CPU, such as insts in the O3 CPU versus numInsts in the Simple CPU. Additionally, other stats only output in certain CPUs when they may be useful to all CPUs, such as cpi and ipc only outputting in Minor CPU.
This change aims to unify stats with similar names, as well as move stats from certain CPUs that are more generally relevant into the Base CPU output. However, doing this interrupts the way that one can parse the stats.txt output from running tests, so I wanted feedback on the way that this change is being implemented in order to make sure certain stats weren't accidentally dropped, naming conventions weren't unclear, values weren't incorrectly outputted and so on. Below, I've attached links to the Jira Issue addressing this change, as well as a link to the beginning of this change set. Within the Jira Issue, I have links to a mapping of all the old stats names, how they would be represented after these changes go through, and their origin CPU. In addition, I've uploaded the stats outputs from running a variety of CPU and Ubuntu boot tests within gem5 before and after these changes for comparison. In addition, this change is split into two halves: introducing the new stats groups to use going forward (FetchCPUStats, ExecuteCPUStats, CommitCPUStats) in the first half and removing the old stats that have been moved to new locations in the second half. I did this so that anyone interested in using diff to observe the changes could have a version of stats.txt before the changes, one with both the old and new stats all in one file, and finally a version with just the changes made. Similarly, the Jira post contains versions of the CPU and Ubuntu boot tests from each of these changes in folders named before-stats, middle-stats, and after-stats. Hopefully, this should make it easier for anyone who wanted to take a closer look at how these changes were incorporated. Jira Issue: https://gem5.atlassian.net/browse/GEM5-1304 Gerrit Changes: https://gem5-review.googlesource.com/c/public/gem5/+/69097 I would appreciate any feedback on this, so we can make sure this change isn't extremely disruptive to the way stats are used within gem5. Thank you for all the help, and if you had any questions regarding these changes please let me know! Best, Melissa Jost
_______________________________________________ gem5-dev mailing list -- gem5-dev@gem5.org To unsubscribe send an email to gem5-dev-le...@gem5.org
