* The top 4 bits of $D018 (VM10-VM13) control the base of graphics memory
* The bottom 3 bits of $D011 (YSCROLL) and $D016 (XSCROLL) let you delay screen display by 0-7 pixels in either axis
* The 4th bit of $D011 (RSEL) and $D016 (CSEL) let you shrink the screen viewport so it's visibly 304x192 pixels (in NTSC) even though the screen memory is still 320x200.
VM10-13 lets you establish a double-buffer, you can switch instantly between a front and back buffer by changing $D018. While you are scrolling, you copy the front buffer to the back buffer, offset by 8 horizontal pixels (1 byte) or 4 vertical pixels, then add in new map data on the edge you're scrolling towards. On each frame, you update XSCROLL/YSCROLL by one pixel so the visible screen is moving towards your new screen, then once it's moved all 8/4 pixels, switch to the new buffer.
The cost of copying all that graphics memory is expensive, so split it across the 8/4 frames.
There's a lot that can be improved in the code, such as just using FOR...TO...STEP would be much faster.
In general, this is a routine that is super easy to write in 6502 assembly. Most magazine BASIC games had something like that, with a few such routines that were POKEd in memory at startup. Several were written in a relocatable fashion so that they could be copied easily from one game to the next.
What you really need:
http://1amstudios.com/2014/12/07/c64-smooth-scrolling/
https://www.c64-wiki.com/wiki/Scrolling
* The top 4 bits of $D018 (VM10-VM13) control the base of graphics memory
* The bottom 3 bits of $D011 (YSCROLL) and $D016 (XSCROLL) let you delay screen display by 0-7 pixels in either axis
* The 4th bit of $D011 (RSEL) and $D016 (CSEL) let you shrink the screen viewport so it's visibly 304x192 pixels (in NTSC) even though the screen memory is still 320x200.
VM10-13 lets you establish a double-buffer, you can switch instantly between a front and back buffer by changing $D018. While you are scrolling, you copy the front buffer to the back buffer, offset by 8 horizontal pixels (1 byte) or 4 vertical pixels, then add in new map data on the edge you're scrolling towards. On each frame, you update XSCROLL/YSCROLL by one pixel so the visible screen is moving towards your new screen, then once it's moved all 8/4 pixels, switch to the new buffer.
The cost of copying all that graphics memory is expensive, so split it across the 8/4 frames.
There's a lot that can be improved in the code, such as just using FOR...TO...STEP would be much faster.
In general, this is a routine that is super easy to write in 6502 assembly. Most magazine BASIC games had something like that, with a few such routines that were POKEd in memory at startup. Several were written in a relocatable fashion so that they could be copied easily from one game to the next.