> On May 30, 2012, 3:26 p.m., Nilay Vaish wrote: > > Ali, can report the size of the BaseDynInst class before and after > > the application of your patch? > > Ali Saidi wrote: > 696 bytes before; 512 after for ARM. > > Nilay Vaish wrote: > That's a lot of improvement. Did you notice improvement in the performance > of any regression / benchmark tests? How did you go about rearranging > the variables of the class? Is the improvement also coming from > eliminating > some of the variables?
There seemed to be some small benefit. I re-arranged them based on my best guess about which variables would be most commonly accessed and which would be done together. I'm sure there is a better arrangement, but this is the best I could do without some tools to help. It also cleans up the code to put al the variables in one places which makes this a good be easier in the future. - Ali ----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: http://reviews.gem5.org/r/1241/#review2862 ----------------------------------------------------------- On May 30, 2012, 3:13 p.m., Ali Saidi wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > http://reviews.gem5.org/r/1241/ > ----------------------------------------------------------- > > (Updated May 30, 2012, 3:13 p.m.) > > > Review request for Default. > > > Description > ------- > > Changeset 9037:350f47d32f56 > --------------------------- > O3: Clean up the O3 structures and try to pack them a bit better. > > DynInst is extremely large the hope is that this re-organization will put the > most used members close to each other. > > > Diffs > ----- > > src/cpu/base_dyn_inst.hh 8b9f227b64d8 > src/cpu/base_dyn_inst_impl.hh 8b9f227b64d8 > src/cpu/o3/bpred_unit.hh 8b9f227b64d8 > src/cpu/o3/comm.hh 8b9f227b64d8 > src/cpu/o3/commit_impl.hh 8b9f227b64d8 > src/cpu/o3/decode_impl.hh 8b9f227b64d8 > src/cpu/o3/dyn_inst.hh 8b9f227b64d8 > src/cpu/o3/dyn_inst_impl.hh 8b9f227b64d8 > src/cpu/o3/iew_impl.hh 8b9f227b64d8 > src/cpu/o3/inst_queue_impl.hh 8b9f227b64d8 > src/cpu/o3/lsq_unit.hh 8b9f227b64d8 > src/cpu/o3/lsq_unit_impl.hh 8b9f227b64d8 > src/cpu/o3/rename_impl.hh 8b9f227b64d8 > > Diff: http://reviews.gem5.org/r/1241/diff/ > > > Testing > ------- > > > Thanks, > > Ali Saidi > > _______________________________________________ gem5-dev mailing list [email protected] http://m5sim.org/mailman/listinfo/gem5-dev
