> On July 24, 2012, 9:24 a.m., Ali Saidi wrote: > > Ship It! > > Andreas Hansson wrote: > There doesn't seem to really be any measurable performance > difference, so I am inclined to discard this patch and avoid > the extra noise in terms of configuration options. > > Ideas? Suggestions?
fine with me. - Ali ----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: http://reviews.gem5.org/r/1282/#review3139 ----------------------------------------------------------- On July 2, 2012, 10:32 a.m., Andreas Hansson wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > http://reviews.gem5.org/r/1282/ > ----------------------------------------------------------- > > (Updated July 2, 2012, 10:32 a.m.) > > > Review request for Default. > > > Description > ------- > > Changeset 9088:d0e979f2644c > --------------------------- > gcc: Enable tuning for native CPU features for gcc and clang > > This patch adds a scons flag to indicate that compilation should be > done using gcc and clang's automatic identification of CPU > features. This flag is expanded based on the CPUID, for example on my > Core i3 CPU, "-march=native" turns into "-march=core2 -mcx16 -msahf > -mpopcnt -msse4.2 --param l1-cache-size=32 --param > l1-cache-line-size=64 --param l2-cache-size=256 -mtune=core2", thus > tuning both for the cache organisation, and potentially using > instructions which are specific to the CPU. > > If you compile and execute on different machines this may obviously > cause problems, so beware. The flag is also known to cause problems > with distcc (as you can imagine). When compiling for local execution, > this flag should be the default option, and we might want to invert it > at some point to say --no-native and be default on. > > Performance benchmarks are in progress and I will bump the patch once > some numbers are available. > > > Diffs > ----- > > SConstruct 5f0321c03a26 > > Diff: http://reviews.gem5.org/r/1282/diff/ > > > Testing > ------- > > Compiled correctly using a variety of gcc versions. Performance testing in > progress. > > > Thanks, > > Andreas Hansson > > _______________________________________________ gem5-dev mailing list [email protected] http://m5sim.org/mailman/listinfo/gem5-dev
