🚀 go-pugleaf

garst@sunchem.chem.uga.edu (John Garst) wrote:

>Can anyone provide (even a very gross) estimate of the execution speed
>advantage that might be obtained by using PowerOberon, compared with a
>compiled BASIC on the Amiga 3000 (and the PowerMac 7100/66, should one
>become available soon)?

My experience is that the speed of Oberon in general, and Oberon on
the 7100 (I have one) is pretty respectable. I don't notice any
conspicuous problems in doing numerical applications that include
things like splines, Fourier transforms, FIR filters, and the like. I
also have done image processing and neural network work in this
language and it hasn't given me a bad time.

>Is Oberon a good number crunching language?

It depends. I like using it because I can write small
data-manupulation modules that can operate on data managed by other
modules. The data-holding modules can stay in memory while I write,
compile, and run the analysis routines. This way, the time I used to
spend loading and preparing data to test the analysis routines is
dramatically reduced. I don't know of another convenient compiled
language that lets me do this.

OTOH, this sort of assumes you want to work in the Oberon environment.
I enjoy doing this and find it pretty workable, but your mileage (and
tastes in user interfaces) may vary. Certainly without a 3-button
mouse it will drive you _mad_. (I have invested heavily in Logitech!)

I have used the external linker to take MacOberon modules and make
them into Mac programs, and it works OK, but you can't debug things
that way, and you'd better be ready to wallow in the depths of the Mac
Toolbox to get anything done. I was a pretty experienced Mac
programmer, and I made a nice framework to make this (sort of) easy,
but I didn't have a lot of fun there. (We did this to make a
commercial Mac product out of Oberon code. The resulting program is
actually pretty nice, but if you're just trying to get your own work
done, it's not worth the trouble.)

>Do I face a very steep learning curve if I try to use it?

Not really. It has lots of on-line docs and examples. I'd ignore most
of the how-to-use-the-user-interface stuff for a while and just do
reads and writes from the In and Out modules until you're sure the
language can cut what you want to do numerically. _Then_ program the
fancy display windows...

FWIW, I've seen the complex_numbers_vs_no_complex numbers bit come up
quite a lot here. It's pretty interesting to note that in the
Numerical Recipes book (the original one, in Fortran) virtually no use
was made of the Fortran complex type. I've been able to translate just
about any numeric code I've seen in any procedural language into
Oberon without much trouble, and haven't missed native complex numbers
at all. YMMV, I guess.

(Note: I did a real-world project at a former employer using Oberon &
Gadgets on a SparcStation, and completed _way_ early and under budget,
for a neural-net and image-processing application. So there's a data
point for you.)

(Note #2: I'm teaching a software engineering class where the students
have learned some Oberon & Gadgets and are implementing a small
spreadsheet. They don't seem to be having any particular problems,
which is remarkable, since they are coding collaboratively on about
four different platforms.)

(Aside comment: IMHO, you'll be a long time waiting for those
native-code Basic compilers (unless you're impressed with FutureBasic,
which I'm not.)

>*********************************************************************
>John Garst  garst@sunchem.chem.uga.edu
>*********************************************************************
>Laws of Tradition:  (1) Nothing is ever lost.
>                    (2) Nothing ever stays the same.

Hope this helps! Sorry about the verbosity!

Best wishes,
-greg
---