!!!It's not an entry for challenge!!!
lags a lot and it's the only problem i see at the moment, but 650ms per frame is too much for me.

lib_in_p0...2_x/y/z three points of triangle
lib_in_clr color, one digit red, one digit green, one digit blue
lib_v_t_0...2_0...3 first 4x3 transformation matrix (lowest row is omitted because it's always [0,0,0,1])
lib_fb_t_0...2_0...3 second transformation matrix; same as above.
after both transformations x,y coordinates are in pixels (starting from left-lower corner), and z is depth (towards screen; 0 is far plane and 0.999 is nearest possible value)
lib_draw and lib_clear first one, if true, draws triangle; second one, if true, clears screen. iirc clear command is prioritized.
lib_fb_0...(lib_fb_s_x-1)_0...(lib_fb_s_y-1) is frame buffer, it contains color and depth of pixels (three digits for color and fractional part for depth, RGB.ZZZZ)