On Thu, 2004-09-02 at 05:05, Mike Barnes wrote: > On Wed, 01 Sep 2004 09:17:41 -0700, Ulrich Drepper <drepper@xxxxxxxxxx> wrote: > > > 3. It would almost double the size of i386.rpms (These sse2 libs would > > > have to be part of i386.rpms) - Is it worth it? > > > > The size of the actual DSOs is not the only factor in the RPM size. > > This means that two RPMs are bigger then one RPM with two DSO versions. > > Just playing Devil's Advocate here, but if the extra optimised > libraries are in a separate directory, wouldn't it be trivial to > define a subpackage for them? > Say we have libinfinite, which is a special library for executing > infinite loops. There's an option to have an SSE2 optimised version of > the library, which executes them even faster. > > libinfinite-0.1-1.i386.rpm contains > /usr/lib/libinfinite.so.0 > (and other common docs, utils, etc) > > A subpackage, libinfinite-sse2-0.1-1.i386.rpm, contains > /usr/lib/sse2/libinfinite.so.0 > (just the optimised version, depends on libinfinite) There is one major drawback of this approach: The libinfinite-sse2*.rpm would not get automatically installed by apt/yum etc. If /usr/lib/sse2/*.so.* were part of i386-rpms, they would get automatically installed on all ix86 systems and the dynamical linker would have to decide on which library to use at run-time. Ralf
