vtkPiecewiseFunction::GetValue() is 6x slower
This issue was created automatically from an original Mantis Issue. Further discussion may take place here.
I've attached a sample python script which shows that vtkPiecewiseFunction::GetValue() is now 6x slower than it was before the pipeline changes. Here is an extract from my mail to the vtk-devel mailing list:
I have attached a very interesting example that illustrates a 6x slow-down with vtkPiecewiseFunction::GetValue() calls. I think that in this case it's due to certain overheads associated with the new pipeline implementation (Update() is called in GetValue(), and this triggers bunches of other logic). The comment in the attached example shows my exact performance figures and the VTK checkouts I tried it with.
This is a significant problem, as for example the VTK RayCasting code routinely does 65536 lookups more than once (for the opacity and colour lookups of shorts). On my machine, a single set of 65536 lookups now takes 2.4 seconds (instead of 0.4 seconds).