r2000-analyze-routine

name	r2000-analyze-routine
description	Analyzes a disassembly subroutine to determine its function by examining code, cross-references, and memory usage, leveraging target system expertise.

Analyze Routine Workflow

Use this skill when the user asks to "analyze this routine" or "what does this function do?"

1. Determine System & Context

• Use r2000_get_binary_info to get the system, filename, description, and may_contain_undocumented_opcodes hint.
• CRITICAL: The system field tells you the target computer (e.g., Commodore 64, VIC-20). You MUST become an expert in that specific target computer's memory map, hardware registers, and KERNAL routines.
• CONTEXT: The filename and description tell you the specific software being analyzed. Use this to identify standard libraries (e.g., "Hubbard music driver", "Exomizer decompressor") and to understand the likely purpose of routines based on the game's genre (e.g., "check_collision" in a shooter).
• UNDOCUMENTED OPCODES: If may_contain_undocumented_opcodes is true, expect illegal/undocumented MOS 6502 opcodes (e.g., LAX, SAX, SLO, DCP, ISC) within routines. These are valid instructions — do not treat them as disassembly errors.

2. Identify the Bounds

• If the user provides an address, start there.
• If no address is given, call r2000_get_disassembly_cursor first — the routine starts at the cursor address.
• Look for the start (entry point or label) and end (RTS, JMP, or RTI) of the routine.
• Note: Some routines end with a JMP to a shared epilogue — that still marks the end of this routine's body. Others fall through into the next routine with no explicit return; use cross-references and logic flow to determine the boundary.

3. Read the Code

• Use r2000_read_region (with "view": "disasm") to get the instructions.
• Analyze the flow:
  • Does it loop?
  • Does it call other known routines or OS entry points?
  • Does it access hardware registers?
• Look for common routine patterns (generic 6502):
  • SEI / CLI bracketing → IRQ setup/teardown.
  • Loop with LDA/STA and DEX/DEY/BNE → Memory copy or fill.
  • Bit-shifting, ADC/SBC chains → Math or decompressor.
  • Reads a memory-mapped I/O address then branches → Hardware polling.
• Use your knowledge of the target system (from r2000_get_binary_info) to recognize system-specific patterns:
  • Calls to OS/KERNAL entry points → System service calls.
  • Reads/writes to hardware register addresses → I/O, video, sound, or input handling.
  • Writes to interrupt vector locations → IRQ/NMI setup.
  • Writes to sound chip registers → Music/SFX driver tick.
  • Reads input port registers → Joystick, keyboard, or controller polling.

4. Check Context

• Use r2000_get_cross_references on the entry point to see who calls it.
• This often provides a decisive hint:
  • Called from an initialization block → likely a setup routine.
  • Called from a main loop → likely a per-frame update.
  • Called from an IRQ → must be fast; likely a music tick or raster update.
  • No callers found → may be a dispatch target reached via a pointer/jump table.

5. Analyze Data Usage

• For each memory address accessed by the routine (e.g., LDA $C000, STA $02):
  • Check whether it falls in the target system's hardware register range. Use your knowledge of the target system's memory map (hardware registers, OS variables, ROM entry points) based on the system value from r2000_get_binary_info.
  • Otherwise, call r2000_get_cross_references on that address to understand:
    • Is it written only once (init)? → likely a constant or config variable.
    • Is it written and read by multiple routines? → shared state / global variable.
    • Is it in Zero Page ($00–$FF) and used with ($addr),Y? → indirect pointer.

6. Synthesize

Combine findings into a summary:

• Purpose: What does it do? (e.g., "Clears screen memory").
• Inputs: Registers or memory locations used as arguments.
• Outputs: Registers or memory locations modified.
• Side Effects: Hardware changes, screen updates, etc.

See Step 7 for the exact comment block format to use when documenting your findings.

7. Document

Add a multi-line comment block on top of the routine, rename the label to a descriptive one, and add side-comments to key instructions to explain the logic flow.

The multi-line comment block must be placed above the first instruction of the routine using r2000_set_comment with "type": "line". It should follow this exact format — the separator line must be used as both the first and last line of the comment:

=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
<Summary of what the routine does>

Inputs:  <registers or memory locations used as arguments, or "None">
Outputs: <registers or memory locations modified, or "None">
Side Effects: <hardware changes, screen updates, etc., or "None">
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-

To document the routine, use:

• r2000_set_label_name — to give the routine a descriptive name.
• r2000_set_comment with "type": "line" — to add the multi-line comment block above the entry point.
• r2000_set_comment with "type": "side" — to annotate key instructions within the routine body with short inline notes (e.g., explaining what a register holds, why a branch is taken, or what a memory address represents). Crucial for making the code readable for others.

---

Common Pitfalls

1. Fall-through into next routine: If there is no RTS/JMP/RTI at the apparent end, the routine may intentionally fall through. Check if the next label is also a valid entry point called independently.
2. Tail calls: A JMP sub_routine at the end is a tail call — the routine ends there. Do not include the target routine's body in this routine's analysis.
3. Shared epilogue: Multiple routines may converge to a single RTS. The epilogue belongs to none of them specifically; note this in the comment.
4. Indirect dispatch targets: A routine with no callers may still be active — it could be referenced by a pointer in a jump table. Check nearby data blocks for address tables.
5. Undocumented opcodes: Some programs use illegal/undocumented opcodes (LAX, SAX, SLO, etc.). Check the may_contain_undocumented_opcodes hint from r2000_get_binary_info. If true, these are expected — don't stop disassembly at them. Even if false, remain aware of their existence.
6. IRQ re-entrancy confusion: If an IRQ handler calls JSR routines, those routines may share Zero Page with the main program — context is IRQ-relative.

---

Reporting Results

After completing the analysis, report to the user:

• Purpose: One-sentence description of what the routine does.
• Inputs / Outputs / Side Effects: As determined in Step 6.
• Evidence: Key instructions or cross-references that led to the conclusion.
• Actions taken: What was renamed, what line-comments were added, and which key instructions received side-comments for clarity.
• Uncertain areas: Any instructions or addresses whose purpose is still unclear.