|
From: Ulrich D. <dr...@re...> - 2006-11-22 08:56:18
|
Nicholas Nethercote wrote: > This is not such a hot path, the set1 !=3D set2 case is much less commo= n > than the set1 =3D=3D set2 case. True, but still far too frequent to be ignored. Even the wrong case=20 (i.e., when the tag needs to be corrected) appeared in measurable=20 amounts. I ran the broken and fixed cachegrind binaries and the=20 introduced errors were measurable. And that's only a fraction of cases=20 handle in this if branch. > Cache optimisations are sufficiently subtle that I would consider very=20 > few of them obvious. Well, it _is_ obvious that if in one version two cache lines are used=20 and in the other just one, the second is faster. --=20 =E2=9E=A7 Ulrich Drepper =E2=9E=A7 Red Hat, Inc. =E2=9E=A7 444 Castro St = =E2=9E=A7 Mountain View, CA =E2=9D=96 |