> On Feb. 25, 2013, 2:39 p.m., Ali Saidi wrote: > > this seems fine with me. If you set the buffer size to 64, do you see any > > statistics changes? If not, feel free to commit. > > > > > > Anthony Gutierrez wrote: > There are a few more misses, not a significant number, but there are > more. I'm still trying to figure out why. > > Ali Saidi wrote: > Any updates? Should we get this committed?
Took a different approach and just turned cacheData into an actual fetch buffer. When set to 64B this change makes no difference over the previous code. Setting to 16B in ARM makes it closer to results observed in HW. Regressions pass. - Anthony ----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: http://reviews.gem5.org/r/1726/#review4080 ----------------------------------------------------------- On Aug. 13, 2013, 9:15 p.m., Anthony Gutierrez wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > http://reviews.gem5.org/r/1726/ > ----------------------------------------------------------- > > (Updated Aug. 13, 2013, 9:15 p.m.) > > > Review request for Default. > > > Repository: gem5 > > > Description > ------- > > Changeset 9821:9e2898b0e9bd > --------------------------- > cpu: allow the fetch buffer to be smaller than a cache line > > the current implementation of the fetch buffer in the o3 cpu > is only allowed to be the size of a cache line. some > architectures, e.g., ARM, have fetch buffers smaller than a cache > line, see slide 22 at: > http://www.arm.com/files/pdf/at-exploring_the_design_of_the_cortex-a15.pdf > > this patch allows the fetch buffer to be set to values smaller > than a cache line. > > > Diffs > ----- > > configs/common/O3_ARM_v7a.py 2f9aecba23623a188e3c775530cdc47dc78dec6e > src/SConscript 2f9aecba23623a188e3c775530cdc47dc78dec6e > src/cpu/o3/O3CPU.py 2f9aecba23623a188e3c775530cdc47dc78dec6e > src/cpu/o3/fetch.hh 2f9aecba23623a188e3c775530cdc47dc78dec6e > src/cpu/o3/fetch_impl.hh 2f9aecba23623a188e3c775530cdc47dc78dec6e > > Diff: http://reviews.gem5.org/r/1726/diff/ > > > Testing > ------- > > ran several of the shorter SPEC CPU2006 benchmarks with test input. > icache.overall_accesses::total stat was validated against real > cortex A15 hardware, the value is much closer now. > > regressions pass > > > Thanks, > > Anthony Gutierrez > > _______________________________________________ gem5-dev mailing list [email protected] http://m5sim.org/mailman/listinfo/gem5-dev
