🚀 go-pugleaf

Benjamin R. Liblit writes:
 > On Wed, 21 Oct 1992 18:06:46 GMT, Mark.Bush@prg.ox.ac.UK (Mark Bush) said:
 > 
 > > [...] I have just finished implementing classes
 > > such as Form, BitBlt, DisplayObject, DisplayScreen, Controller, etc...
 > >
 > > [...] `notify:' is defined to open a Notifier window displaying the
 > > appropriate message in the titlebar.
 > 
 > Is it just me, or does it sound like this gentleman is actually
 > implementing a genuine GUI for GNU Smalltalk?  :-)
 > 
 > Tell us more, Mark:  enquiring minds want to know.

It sure looks like it to me :-)

This will make things *much* more interesting.  Now we just need some speed
improvements and we're all set.  I'm toying with the idea of having the
Smalltalk byte code compiler generate C code instead, which can be compiled and
dynamically loaded into the running Smalltalk image.  Now, this won't be nearly
as efficient as optimized assembler, but it should be an improvement
over the current interpretation scheme.  I would like to have compiled C code
and byte coded Smalltalk routines work side by side, and have a different form
of compile: method to create and link in the C code.  As I say, this is a dream
at this point, but who knows...a little extra time after the experimental OOP
table removal and it may appear.  I'll use the GNU dynamic linking facility to
load the code in since it's portable, and have an extra instance variable in
instances of CompiledMethods to hold onto the pointer to the C code.  The
interpreter (really sendMessage) will check for a non-sentinal value in this
variable, and invoke the corresponding routine when appropriate.  This of
course is just a place holder for the real 2.0 compiler, but it may make a
significant difference in the overall performance of the system.

Steve