Lars Gullik Bjønnes wrote: > Peter Kümmel <[EMAIL PROTECTED]> writes: > > | Lars Gullik Bjønnes wrote: > | > Peter Kümmel <[EMAIL PROTECTED]> writes: > | > > | > | Andre Poenitz wrote: > | > | > Have you proof from the profiler that this is necessary? > | > | > | > | The patch has already been dropped because of my profiling, > | > | see the end of the thread. > | > | > | > | But I will test the same trick for the text function, > | > | then the most used function of LyX is not virtual any more. > | > > | > I am not sure that template tricks is the way to go. > | > > | > And the gain will always be minimal, but go after the huge offenders > | > instead of micro-optimizations that makes the code more complex and > | > the binaries larger. > | > | I'm not proposing such a patch ATM, I'm only interested in the > | costs of such an additional virtual call. > > Compared to the rest of the call-chain, the virtual call has minimal > overhead. > > And to figure out the cost of an additional virtual call is easy, > create a test-case that test exactly this. >
You are right, I would have wasted my time. Peter
