Hi Jiachen, (1) There are three sections labeled simulator and base_machine; the first is kernel mode stats, the second is user mode stats, and the third section is the total (sum) of both the kernel and user stats. You can always determine which section is which by referring to the tags in the header of each section:
2 simulator: 3 tags: [shared_l2, wrigley.(none), 2012-11-09, kernel] 4 version: 1341 simulator: 1342 tags: [shared_l2, wrigley.(none), 2012-11-09, user] 1343 version: (2) Depending on what you are focusing on, you'll generally want to look at different fields. A good starting point might be IPC. However (just as an example), if you're studying the effects of cache size, you might be more interested in cache hits and misses. As for determining time, use either the seconds field for time spent simulating, or calculate the simulated time using the native_hz/cycles/cycles_per_second/etc. fields accordingly. Best, Tyler On Sun, Dec 30, 2012 at 11:15 AM, jcxue <[email protected]> wrote: > Hi, > > I have two questions regarding how to interpret the stats file: > > (1) There are multiple sections in the stats file, each starting with > > --- > base_machine: > > what does each section correspond to and which section should be looked at > for a particular simulation? > > (2) In multicore configuration, to get a rough idea of performance improvement > of one design over the other, which field should be looked at? Since each core > has a field "cycles", I am just wondering if there is a place giving the > "total > time" for the simulation. > > Thank you, > > Jiachen > > > _______________________________________________ > http://www.marss86.org > Marss86-Devel mailing list > [email protected] > https://www.cs.binghamton.edu/mailman/listinfo/marss86-devel _______________________________________________ http://www.marss86.org Marss86-Devel mailing list [email protected] https://www.cs.binghamton.edu/mailman/listinfo/marss86-devel
