Timings

[Jon L White <edsel!jonl@labrea.stanford.edu>]

Thanks for you explanatory note.  Indeed, what may be the problem with
the gross discrepancies is that your automated benchmark driver is
combining together bits and pieces of the Gabriel suite, and reporting
those times.  Yesterday's message used the same names found in Gabriel's 
book for these combinations of tests, but without actually mentioning that
it was a particular summary.

The one very positive thing to come out of this interchange, from my
viewpoint, is that you've found reasonable cases where the obvious
declarations for one Common Lisp implementation seem to impede the
performance in another, and vice versa.  The importance of this for 
programmers trying to port from one vendor to another should not be 
overlooked.

Additionally, I would agree with you that it is a substantial improvement 
to KCL to remove the requirement for block-compilation in each file.  
Although this may not be a step forward from the standpoint of the 
"Gabriels", it certainly is a step forward from the standpoint of the 
programmer who has to develop programs in such a lisp.  You also imply 
that you speeded up the general function-to-function interface of KCL?

Even though the "realistic application" clause of benchmark criticism
hasn't been invoked here, one can appreciate your testimonial about the 
usability of KCL and AKCL for the Computational Logic group.


-- JonL --