FMSLogo Code

Include turtle.bmp in the release.

Add a bitmap that resembles a turtle that is to be used in place of the triangle in a Sri Lankan school system.

Add some tests for trying grow a bitmap with BITFIT to something that is larger than it could reasonably allocate.

Used integral math, instead of floating point math, when interpolating RGB values, since the domain is so limited. This results in a small (1%) speed improvement.

Bypass the bilinear interpolation for sprite rendering if all adjacent pixels are identical. This yeilds a 10% speed gain when drawing squirals and circles.

Add bilinear interpolation when drawing rotated sprites. This is very slow, but otherwise the sprites look terrible when rotated at oblique angles.

Add special cases when drawing sprints of rotations of 90, 180, and 270 degrees to make sure that the sine and cosine are exactly as they should be, and not slightly off due to the imprecision of the trig functions in Borland's libc. This was done to get crisp-looking bitmaps at those angles.

Add support for rotating bitmapped turtles, called "sprites". The implementation is a bit slow, clumsy, and doesn't quite solve everything it was supposed to solve. This will remain an undocumented and unsupported feature until these problems are taken care of.

Refactoring: Centralize the logic for computing the turtle's bounding box in ibmturt(), now that I better understand what it is doing and why.

Simplify ibmturt() a bit by keeping the bounding box in screen coordinates, rather than mapping it to world coordinates, then mapping it back. This has a small gain in performance (measured by benchmark.lgo), but the main motivation was to make it easier to later centralize the way bounding boxes are computed in turtlepaste() and ibmturt().