|
|
||||
|
Re: More thoughts on compile-time performance
by
Michael Kay
I've trimmed another 200ms off the compile time (whether or not optimization is enabled) by a more intelligent implementation of setContainer(). Every node in the expression tree has a pointer to the local root of the tree, and when the tree is grafted onto a new root, becoming part of a larger container, it is "repotted" by updating this pointer in every node. It turns out this is happening rather frequently; in particular, every time an XSLT instruction is compiled, the subordinate instructions and their subtrees are repotted in this way. Again this is quadratic in the size of the function or template.
|
Search
Recent Comments
Recent Articles
Month Archive
|
|||