Re: Slowup in 2.1.54

[John W. Eaton]

On 17-Feb-2004, Dmitri A. Sergatskov <address@hidden> wrote:

| David Bateman wrote:
| > I've just rerun the sciview tests, sinced I'd presumed that all of the 
recent
| > changes in Octave would result in significantly faster code. It basically
| > is just that the last test is much slower. I think the [] operator is
| > slower in 2.1.54, though I need to check some more. Any clues?
| 
| I also noticed that recursion is slower. For sciview tests it is in
| grand common divisors tests (gcd2.m).

There were some changes to the parse tree code to store more data to
allow for better information from warning messages in the future.
Saving and restoring the data requires some additional calls to
functions in the unwind_protect class.  It's possible that this is the
cause of what you are seeing.

But, recursion and function calls are known to be slow in Octave and
there is no simple fix.  If you want them to be faster, then I think
you will need to write a new interpreter that takes the current
parse tree information and converts it to something that can be uased
by a stack based interperter or a just-in-time compiler.

jwe