Should be quite similar to the requirements of an old MS-DOS COM file. 64kB of data is quite a lot, and as long as the code doesn't try flow-filling any regions, there will be very little need for recursive calls or extra state information besides a reasonably large bitmap buffer.

If printing 1k dots in one dimension, then the memory will be able to fit 512 dots in the other direction. But a big advantage with a matrix printer is that it doesn't need to get all data fluidly. It is ok to stop the transfer to recompute.

So it is possible to draw a very large graph in memory, but to clip the drawing to a narrow vertical band and then emit this band to the printer. Then redraw the graph again in memory, but clipping to the next narrow vertical band.

For printing text, you either switch to text mode, or store one or more nice character sets in the code space and render these characters (easiest with bitmapped) into the same clipped memory region before you emit the text.

The LX-300 is a 9-pin printer, so even if it may handle 2k pixels horisontally, it will not be meaningful to try too high horisontal resolution unless you spend time doing multiple passes with fractional line feeds. For really high graphic resolutions, a printer with a 24-pin high-resolution printer head should be used instead.

As long as the processor can fit at least the bitmap for a single horisontal scan, everything should be fine. But you should maybe pick up an entry-level book on computer graphics. Most of the text can be skipped, but it will contain the basics about line drawing, clipping etc. Peeking in a computer-graphics book, you will see that it is quite easy to quickly draw circles, ellipses, or lines in any angle.