Cube Processing for Dimension Schema changes

Hi,

I had a general question about how cube processing works.

If my dimension schema changes (an addition of a property), will I have to do a full reprocess of all cube partitions for which that dimension was member of? Or do I only need to reprocess the dimension?

I am working on a system in which our data will contain millions of records, but our schema is extensible and the user can add new attributes to a dimension. The perf impact if we need to do a full reprocess on all partitions of the dimension would be huge. I'm still new to cubes, so I'm not sure if a change to the dimension schema will require a full reprocess of all affected partitions.

Thanks,
Danny