Based on the results I am getting, for PARSEC benchmarks, the OoO preformance is really bad, there are to many hw_mfpr and hw_mtpr, instructions. I am trying to figure out why I am in the PALcode so often (any ideas on how to figure this out?). I am running Blackscholes, which is a relatively simple PARSEC benchmark.
I need to do more research, but I dont think this is caused at all for itb and dtb misses (i made them really large just in case). Right now for blackscholes, which isnt close to finish executing (about 30%) I have it so it will execute 1e9 instructions (running two cores), From my perl scripts it seems to have it fetched 13 million hw_mfpr and mtpr instructions (fetched, not committed). There is something really wrong with that ratio. Thanks, EF On Mon, Oct 12, 2009 at 9:28 PM, nathan binkert <[email protected]> wrote: >> I noticed that the M5 pal code is based on the ev5, is it possible to >> make it ev6 or ev67? Or are there no benefits? >> >> My main overall issue I am thinking about is I am trying to see if >> their is any problems in my compiled benchmarks, as I have made them >> ev67 code, ev5 has given me lower ipc as well as a few problems in >> unimplemented SIMD like instructions. I am not happy about my compiled >> PARSEC benchmarks IPC, the main problem is I my simulations keep going >> into palmode, stalling my OoO processor, and killing IPC. > > While we model EV5 PAL, we more or less model the EV67 ISA (though, > there may very well be missing instructions as you've mentioned.) The > issues that you're seeing have more to do with the fact that most pal > instructions are serializing. It's just never been anything that > we've decided to optimize. Fixing it would likely require both > rewriting some of the palcode and changing the way that we deal with > the pal instructions in o3. > > My first question is, why are you entering PAL so often? > > Nate > _______________________________________________ > 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
