On Thu, 2005-08-18 at 08:39 -0700, Praveen Ray wrote: > I suspect its because we've written a large number of > moderately complex macros
Could be. You can probably test this fairly simply by globally removing the macro calls and benchmarking. > Or, is it the fact that TT uses string concatenation for > every MACRO and BLOCK's output? Not likely, although that may increase the memory used. > I wish TT could write onto output stream as the output is > generated instead of first building a giant string. There have been patches on the list to flush output periodically, but I don't think they remove all buffering. The buffering is what makes things like filters, wrappers, and exception handling possible. However, TT is pretty hackable if you want to change that. - Perrin
