You can not select more than 25 topics
Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
280 lines
11 KiB
280 lines
11 KiB
// This is all new to me so it's heavily commented so we can understand what is happening. |
|
|
|
#[derive(Debug)] |
|
#[allow(non_camel_case_types)] |
|
// AddressingMode is a mehtod used by a CPU to determine where the operand (data or memory |
|
// location) for an instruction comes from. It basically defines how the CPU finds the data it |
|
// needs to execute a given instruction. |
|
pub enum AddressingMode { |
|
// Operand is provided directly as part of the instruction |
|
Immediate, |
|
//Operand is located in the zero page (memory address 0x00 to 0xFF) |
|
ZeroPage, |
|
// Operand is in the zero page, with an offet from the X register |
|
ZeroPage_X, |
|
// Operand is in the zero page, with an offet from the Y register |
|
ZeroPage_Y, |
|
// Operand is at the full 16bit address (0x0000 to 0xFFFF) |
|
Absolute, |
|
// Operand is at an absolute address, with an offset from the X register |
|
Absolute_X, |
|
// Operand is at an absolute address, with an offset from the Y register |
|
Absolute_Y, |
|
// Operand is at an address stored in a zero-page address plus an X offset |
|
Indirect_X, |
|
// Operand is at an address stored in a zero-page address plus an Y offset |
|
Indirect_Y, |
|
// No addressing mode (used for instructions that don't require an operand like NOP) |
|
NoneAddressing, |
|
} |
|
|
|
pub struct CPU { |
|
// the accumulator register is a specific register used for arithmetic and logic operations |
|
// the cpu instruction loads a value into the accumulator register and then updates certain |
|
// flags in the processor status register to relect the operation of the result |
|
pub register_a: u8, |
|
// THe Process Status Register is a collection of individual bits (flags) that represent the |
|
// current state of the CPU. Each bit has purpose such as if a calculation resulted in zero or |
|
// if the result is negative |
|
// |
|
// Zero flag: |
|
// - Bit 1 of the status register |
|
// - Set if register_a == 0, cleared otherwise |
|
// |
|
// Negative flag (N) - indicates whether the result of the most recent operation is negative |
|
// - Bit 7 of the status register (most significant bit is Bit 7 because it's zero based) |
|
// - Set if the most significant bit of register_a is 1, cleared otherwise |
|
pub status: u8, |
|
// track our current position in the program |
|
pub program_counter: u16, |
|
// most commonly used to hold counters or offsets for accessing memory. The value can be loaded |
|
// and saved in memory, compared with values held in memory or incremented and decremented. |
|
// |
|
// X register has one special function. It can be used to copy a stack pointer or change it's |
|
// value. |
|
pub register_x: u8, |
|
pub register_y: u8, |
|
// this creates an array with size of 0xFFFF and initializes all elements to 0 (see below in t |
|
// he new function). 0xFFFF is hexadecimal for 65535 in decimal. This defines the size of the |
|
// array with 65535 elements, the typical size for a 6502 CPU system. |
|
memory: [u8; 0xFFFF], |
|
} |
|
|
|
impl CPU { |
|
pub fn new() -> Self { |
|
CPU { |
|
register_a: 0, |
|
status: 0, |
|
program_counter: 0, |
|
register_x: 0, |
|
register_y: 0, |
|
memory: [0; 0xFFFF], |
|
} |
|
} |
|
|
|
fn get_operand_address(&self, mode: &AddressingMode) -> u16 { |
|
match mode { |
|
AddressingMode::Immediate => self.program_counter, |
|
AddressingMode::ZeroPage => self.mem_read(self.program_counter) as u16, |
|
AddressingMode::Absolute => self.mem_read_u16(self.program_counter), |
|
AddressingMode::ZeroPage_X => { |
|
let pos = self.mem_read(self.program_counter); |
|
let addr = pos.wrapping_add(self.register_x) as u16; |
|
addr; |
|
} |
|
AddressingMode::ZeroPage_Y => { |
|
let pos = self.mem_read(self.program_counter); |
|
let addr = pos.wrapping_add(self.register_y) as u16; |
|
addr |
|
} |
|
} |
|
} |
|
|
|
fn mem_read(&self, addr: u16) -> u8 { |
|
self.memory[addr as usize] |
|
} |
|
|
|
fn mem_write(&mut self, addr: u16, data: u8) { |
|
self.memory[addr as usize] = data; |
|
} |
|
|
|
// combines two consecutives bytes in memory into a single 16 bit value |
|
fn mem_read_u16(&mut self, pos: u16) -> u16 { |
|
// read the low byte (pos) |
|
let lo = self.mem_read(pos) as u16; |
|
// read the high byte (pos + 1) |
|
let hi = self.mem_read(pos + 1) as u16; |
|
// combine the bytes. Shelf the high byte 8 bits to the left, effectively moving it into |
|
// the upper 8 bits of a 16-bit value. This will combine to a single 16 bit value. |
|
(hi << 8) | (lo as u16) |
|
} |
|
|
|
// split a 16-bit number into 2 8-bit pieces |
|
fn mem_write_u16(&mut self, pos: u16, data: u16) { |
|
// take the top half of the 16 bit number |
|
let hi = (data >> 8) as u8; |
|
// take the bottom half of the 16 bit number |
|
let lo = (data & 0xff) as u8; |
|
// store low byte in the first compartment of memory (pos) |
|
self.mem_write(pos, lo); |
|
// store the high byte in the next compartment (pos + 1) |
|
self.mem_write(pos + 1, hi) |
|
} |
|
|
|
pub fn load_and_run(&mut self, program: Vec<u8>) { |
|
self.load(program); |
|
self.reset(); |
|
self.run(); |
|
} |
|
|
|
pub fn reset(&mut self) { |
|
self.register_a = 0; |
|
self.register_x = 0; |
|
self.status = 0; |
|
|
|
self.program_counter = self.mem_read_u16(0xFFFC); |
|
} |
|
|
|
// load the program code into memroy starting at address 0x8000. 0x8000 .. 0xFFFF is reserved |
|
// from program ROM and we can assume that the instructions should start somewhere here. |
|
pub fn load(&mut self, program: Vec<u8>) { |
|
self.memory[0x8000..(0x8000 + program.len())].copy_from_slice(&program[..]); |
|
self.mem_write(0xFFFC, 0x8000); |
|
} |
|
|
|
// LDA stands for Load Accumulator. It loads a value into the accumulator register. It affects |
|
// the Zero Flag and Negative Flag. |
|
fn lda(&mut self, value: u8) { |
|
self.register_a = value; |
|
self.update_zero_and_negative_flags(self.register_a); |
|
} |
|
|
|
fn tax(&mut self) { |
|
self.register_x = self.register_a; |
|
self.update_zero_and_negative_flags(self.register_x); |
|
} |
|
|
|
fn update_zero_and_negative_flags(&mut self, result: u8) { |
|
// update the status register |
|
if result == 0 { |
|
// 0b000_0010 represents a number where only the second bit (bit 1) is set |
|
// to 1 and all other bits are 0 |
|
self.status = self.status | 0b000_0010; |
|
} else { |
|
// 0b1111_1101 represents a number where only bit 1 is 0 and the rest are 1 |
|
self.status = self.status & 0b1111_1101; |
|
} |
|
|
|
if result & 0b1000_000 != 0 { |
|
self.status = self.status | 0b100_0000; |
|
} else { |
|
self.status = self.status & 0b0111_1111; |
|
} |
|
} |
|
|
|
// INX stands for Increment Index Register X - increase the value of the X register |
|
// by one and update specific processor flags based on the result. |
|
fn inx(&mut self) { |
|
self.register_x = self.register_x.wrapping_add(1); |
|
self.update_zero_and_negative_flags(self.register_x); |
|
} |
|
|
|
// The interpret method takes in mutalbe reference to self as we know we will need to modify |
|
// register_a during execution |
|
// |
|
// - Fetch next instruction from instruction memory |
|
// - Decode instruction |
|
// - Execute the instruction |
|
// - Repeat |
|
pub fn run(&mut self) { |
|
// We need an infinite loop to continuously fetch instructions from the program array. We |
|
// use the program_counter to keep track fo the current instruction. |
|
loop { |
|
// set the opscode to the current byte in the program at the address indicated by |
|
// program counter |
|
let opscode = self.mem_read(self.program_counter); |
|
// we increment program counterto point to the next byte |
|
self.program_counter += 1; |
|
|
|
match opscode { |
|
// this will implement LDA (0xA9) opcode. 0xA9 is the LDA Immediate instruction in |
|
// the 6502 CPU. |
|
// |
|
// 0x42 tells the CPU to execute a specific operation: LDA Immediate. An opscode is |
|
// a command for the CPU, instructing it what to do next. Essentially, this means |
|
// "Load the immediate value from the next memory location into the accumulator" |
|
// |
|
// Immediate value refers to the constant value that is directly embedded in the |
|
// instruction itself, rather than being fetched from memory or calculated |
|
// directly. |
|
0xA9 => { |
|
// fetch the next byte in program. This byte is the immediate value to load |
|
// into the accumulator (register_a) |
|
let param = program[self.program_counter as usize]; |
|
// Increment program counter to point to the next instruction after the |
|
// parameter. The program counter must always point to the next instruction to |
|
// be executed. |
|
self.program_counter += 1; |
|
// store the fetch parameter in register_a - handling the actual loading of the |
|
// value into the accumulator and updates the CPU flags. |
|
self.lda(param); |
|
} |
|
// implement the 0xAA opcode which corresponds to the TAX (transfer acculumater to |
|
// X register) |
|
0xAA => self.tax(), |
|
// INX stands for Increment Index Register X - increase the value of the X register |
|
// by one and update specific processor flags based on the result. |
|
0xe8 => self.inx(), |
|
0x00 => return, |
|
_ => todo!(), |
|
} |
|
} |
|
} |
|
} |
|
|
|
#[cfg(test)] |
|
mod tests { |
|
use super::*; |
|
|
|
#[test] |
|
fn test_lda_sets_register_a() { |
|
let mut cpu = CPU::new(); |
|
cpu.load_and_run(vec![0xa9, 0x05, 0x00]); |
|
assert_eq!(cpu.register_a, 0x05); |
|
assert!(cpu.status & 0b0000_0010 == 0b00); |
|
assert!(cpu.status & 0b1000_0000 == 0); |
|
} |
|
|
|
#[test] |
|
fn test_0xa9_lda_zero_flag() { |
|
let mut cpu = CPU::new(); |
|
cpu.load_and_run(vec![0xa9, 0x00, 0x00]); |
|
assert!(cpu.status & 0b0000_0010 == 0b10); |
|
} |
|
|
|
#[test] |
|
fn test_0xaa_tax_move_to_a_to_x() { |
|
let mut cpu = CPU::new(); |
|
cpu.register_a = 10; |
|
cpu.load_and_run(vec![0xaa, 0x00]); |
|
|
|
assert_eq!(cpu.register_x, 10) |
|
} |
|
|
|
#[test] |
|
fn test_5_ops_working_together() { |
|
let mut cpu = CPU::new(); |
|
cpu.load_and_run(vec![0xa9, 0xc0, 0xaa, 0xe8, 0x00]); |
|
|
|
assert_eq!(cpu.register_x, 0xc1) |
|
} |
|
|
|
#[test] |
|
fn test_inx_overflow() { |
|
let mut cpu = CPU::new(); |
|
cpu.register_x = 0xff; |
|
cpu.load_and_run(vec![0xe8, 0xe8, 0x00]); |
|
|
|
assert_eq!(cpu.register_x, 1) |
|
} |
|
}
|
|
|