Famicom Party

8. Refactoring

Table of Contents:

Before we move on to learning more about how the NES draws graphics, let's take some time to reflect on what we've already built. There are a number of improvements we can make now that will be useful in the future. By refactoring now, we will create a useful template for our upcoming projects.


There are several places in our code where we use a particular number that doesn't change, such as MMIO addresses for talking with the PPU. It's hard to tell what these numbers are referring to when looking at the code.

Thankfully, we can replace these abstract numbers with descriptive text, by declaring constants. A constant is essentially a name for one number, which can't be changed. Let's create constants for the PPU addresses we've used so far: These names are the generally agreed-upon standard naming for the various MMIO addresses available in the NES. When looking through references like the NESDev wiki, you will see these names used.

PPUCTRL   = $2000
PPUMASK   = $2001
PPUADDR   = $2006
PPUDATA   = $2007

With these constants, our main code becomes much more readable:

25.proc main
27 LDX #$3f
29 LDX #$00
31 LDA #$29
33 LDA #%00011110
36 JMP forever

Where do we put these constants? The common approach is to make a separate constants file, which can be included into our main assembly file. We'll call the constants file constants.inc. Why does this file end in .inc instead of .asm? The constants file is not exactly assembly code; it doesn't have any opcodes. We will use the .asm extension for assembly code files, and .inc for files which are included into an assembly code file. Then, we include the constants file at the top of our .asm file like this:

.include "constants.inc"

Header File

We can do the same thing with the .header segment, since it will generally be the same from project to project. Let's make a header.inc file to hold our header content. Now would also be a good time to add some comments:

.segment "HEADER"
.byte $4e, $45, $53, $1a ; Magic string that always begins an iNES header
.byte $02        ; Number of 16KB PRG-ROM banks
.byte $01        ; Number of 8KB CHR-ROM banks
.byte %00000001  ; Vertical mirroring, no save RAM, no mapper
.byte %00000000  ; No special-case flags set, no mapper
.byte $00        ; No PRG-RAM present
.byte $00        ; NTSC format

Now we can delete the .segment "HEADER" section of our main .asm file, and include our new header file. The top of our .asm file should now look like this:

.include "constants.inc"
.include "header.inc"

.segment "CODE"

When the assembler and linker run, they will take the contents of header.inc and put them in the correct place in the output ROM, exactly the same as if we had put it directly into the assembly file.

ca65 Imports and Exports

A full reset handler can become quite large, so it can be useful to put it into a separate file. But we can't just .include it, because we need a way to reference the reset handler in the VECTORS segment.

The solution is to use ca65's ability to import and export .proc code. We use the .export directive to inform the assembler that a certain proc should be available in other files, and the .import directive to use the proc somewhere else.

First, let's create reset.asm, including the .export directive:

1.include "constants.inc"
3.segment "CODE"
4.import main
5.export reset_handler
6.proc reset_handler
9 LDX #$00
14 BPL vblankwait
15 JMP main

There are a few things I'd like to point out in this file. First, the file ends in .asm, because it contains opcodes. Second, we include the constants file so that it can be used here. Third, we need to specify which code segment this .proc belongs in, so the linker knows how to put everything together. Finally, note that we are importing main. This way, the assembler knows what memory address the main proc is located at, so the reset handler can jump to the correct address.

Now that we have a separate reset file, we'll use reset_handler inside our code:

4.segment "CODE"
5.proc irq_handler
9.proc nmi_handler
10 RTI
13.import reset_handler
15.export main
16.proc main
17 ; contents of main here
20.segment "VECTORS"
21.addr nmi_handler, reset_handler, irq_handler

On line 13, where our .proc reset_handler used to be located, we now import the proc from an external file. Note that you do not need to specify which file the proc comes from - the assembler scans all .asm files for exports before it starts assembling, so it already knows what external procs are available and where they are located. (Note that this also means you can't export two procs with the same name - the assembler will have no way to tell which one you are referring to in an .import.) You may have noticed that reset.asm uses .segment "CODE", and our main assembly file also uses .segment "CODE". What happens when we assemble and link these files? The linker finds everything that belongs to the same segment and puts it together. The order does not particularly matter, since labels are converted into addresses at link time. We also have to export our main proc, so that the reset handler can import it and know where to jump to when it is finished.

Custom Linker Configuration

When we linked our sample project back in Chapter 3, we used this command:

ld65 helloworld.o -t nes -o helloworld.nes

The -t nes tells ld65 to use the default linker configuration for the NES. This is why we have the "STARTUP" segment, despite never using it. While the default configuration works for the sample project, it can lead to problems as our code becomes larger and more complicated. So, instead of using the default configuration, we will write our own linker configuration with only the segments and features that we need.

Our custom linker config will be in a file called nes.cfg, which will look like this:

  HEADER: start=$00, size=$10, fill=yes, fillval=$00;
  ZEROPAGE: start=$10, size=$ff;
  STACK: start=$0100, size=$0100;
  OAMBUFFER: start=$0200, size=$0100;
  RAM: start=$0300, size=$0500;
  ROM: start=$8000, size=$8000, fill=yes, fillval=$ff;
  CHRROM: start=$0000, size=$2000;

  HEADER: load=HEADER, type=ro, align=$10;
  ZEROPAGE: load=ZEROPAGE, type=zp;
  STACK: load=STACK, type=bss, optional=yes;
  OAM: load=OAMBUFFER, type=bss, optional=yes;
  BSS: load=RAM, type=bss, optional=yes;
  DMC: load=ROM, type=ro, align=64, optional=yes;
  CODE: load=ROM, type=ro, align=$0100;
  RODATA: load=ROM, type=ro, align=$0100;
  VECTORS: load=ROM, type=ro, start=$FFFA;
  CHR: load=CHRROM, type=ro, align=16, optional=yes;

The MEMORY section lays out the regions of memory that segments can be placed into, while the SEGMENTS section describes the segment names we can use in our code and which memory areas they should be linked into. I won't be going into detail on what each setting means, but you can find in-depth documentation in the ld65 docs.

To use this custom linker configuration, we first need to update the segment names in our code to match the config file's segment names. In our case, the only needed changes are moving "CHARS" to "CHR" and removing "STARTUP".

Putting It All Together

Finally, we need to update the structure of our files a bit. We will move all of the .asm and .inc files into a sub-directory, src, with our new linker config at the top level. The code we have after all of our refactoring should now look like this:

   |-- nes.cfg
   |-- src
      |-- constants.inc
      |-- header.inc
      |-- helloworld.asm
      |-- reset.asm

To assemble and link our code, we will use the following commands (run from the top-level 08-refactoring directory):

ca65 src/helloworld.asm
ca65 src/reset.asm
ld65 src/reset.o src/helloworld.o -C nes.cfg -o helloworld.nes

To be clear, what we are doing here is first assembling each .asm file to create .o files. Once that is done, we pass all of the .o files to the linker. Instead of using the default NES linker config (-t nes), we use our new custom config (-C nes.cfg). The output from the linker is placed into the same helloworld.nes ROM file.

If you would like to download a copy of the files listed above, here is a ZIP file of everything so far. We'll be using this setup as a base for our future projects, so be sure that you are able to assemble, link, and run the code before moving on.