On Sun, 2006-10-22 at 19:32 -0400, Matthew Miller wrote: > On Sun, Oct 22, 2006 at 09:18:47AM +0200, Marcus Camen wrote: > > http://lists.opensuse.org/opensuse-factory/2006-10/msg00173.html > > We run SPECcpu2000 benchmarks on x86-64 to compare -Os and -O2. The > > results are 1011 (-Os) and 1140 (-O2) for SPECint and for a subset of > > SPECfp 1202 (-Os) vs. 1270 (-O2), so the speed difference is too high > > to make this change. > > Does SPEC CPU2000 really measure the workload appropriately enough to be a > good basis for this decision? Especially when they're just giving a subset > of the integer and fp benchmarks? also they're microbenchmarks, they don't show the system level gains you get from having more compact code (memory savings, more efficient cache, less disk seeks etc etc) -- fedora-devel-list mailing list fedora-devel-list@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/fedora-devel-list
