The speed problem was twofold - some cruft in callbacks coupled with the fact that I wasn't doing make cleans, so something about the way that gcc was putting together the old .o files was slowing things down. I learned my lesson: always do make clean before testing a patch.
A full make clean brings compilation time on Keith's test case down to: PATCH real 0m7.324s user 0m6.944s sys 0m0.192s CURRENT MASTER real 0m7.407s user 0m6.828s sys 0m0.304s Usually about a 3% increase for scores with lots of cross staff slurs, almost nothing otherwise. Note that this is with an optimized build that is not debugging or profiled enabled. All of those flags add about 4ish%. http://codereview.appspot.com/6498077/ _______________________________________________ lilypond-devel mailing list lilypond-devel@gnu.org https://lists.gnu.org/mailman/listinfo/lilypond-devel
