Ben Pfaff wrote: > How much does this commit trim from the binary size? Alexandru Ardelean wrote: > Roughly a 34k decrease in binary size, decrease being in libofproto and > libopenvswitch libs > Quite a small decrease, so more --disable-<feature> flags would be needed. > My next thought would have been to go after making RSTP and STP > disable-able.
I am not sure that this is the best approach to follow first. It doesn't seem to me that we've actually written 500 kB of new code from OVS 2.0 to OVS 2.5. I'd be curious to find out where the size is coming from. I can think of a couple of approaches that one could take to find that out. One is to look at intermediate points (e.g. v2.1, v2.2, ..., or bisected points in between them) and see if it is a steady growth or whether there are big jumps, and if the latter, whether those jumps are for good reason or not. Another is to start from what we've got not, look at the sizes of the various object files, and do something akin to profiling but for code size: is the largest object file large for a reason, or is there something unfortunate going on? I assume that OpenWRT builds optimize for size with -Os, is that true? _______________________________________________ dev mailing list dev@openvswitch.org http://openvswitch.org/mailman/listinfo/dev
