gb/src/instruction.rs

2163 lines
81 KiB
Rust

use super::cpu::{Cpu, Flags, HaltState, ImeState, Register, RegisterPair};
use std::{convert::TryFrom, fmt::Debug};
#[derive(Debug, Copy, Clone)]
#[allow(clippy::upper_case_acronyms)]
pub enum Instruction {
NOP,
LD(LDTarget, LDTarget),
STOP,
JR(JumpCondition, i8),
ADD(MATHTarget, MATHTarget),
INC(Registers),
DEC(Registers),
RLCA,
RRCA,
RLA,
RRA,
DAA,
CPL,
SCF,
CCF,
HALT,
ADC(MATHTarget), // ADC A, MATHTarget
SUB(MATHTarget), // SUB A, MATHTarget
SBC(MATHTarget),
AND(MATHTarget), // AND A, MATHTarget
XOR(MATHTarget), // XOR A, MATHTarget
OR(MATHTarget), // OR A, MATHTarget
CP(MATHTarget), // CP A, MATHTarget
RET(JumpCondition),
LDHL(i8), // LD HL, SP + d
POP(RegisterPair),
RETI,
JP(JumpCondition, JPTarget),
DI,
EI,
CALL(JumpCondition, u16),
PUSH(RegisterPair),
RST(u8),
RLC(InstrRegister),
RRC(InstrRegister),
RL(InstrRegister),
RR(InstrRegister),
SLA(InstrRegister),
SRA(InstrRegister),
SWAP(InstrRegister),
SRL(InstrRegister),
BIT(u8, InstrRegister),
RES(u8, InstrRegister),
SET(u8, InstrRegister),
}
#[derive(Copy, Clone)]
pub enum JPTarget {
RegisterPair(RegisterPair),
ImmediateWord(u16),
}
#[derive(Copy, Clone)]
pub enum Registers {
Byte(InstrRegister),
Word(RegisterPair),
}
#[derive(Copy, Clone)]
pub enum MATHTarget {
Register(InstrRegister),
RegisterPair(RegisterPair),
ImmediateByte(u8),
}
#[derive(Copy, Clone)]
pub enum LDTarget {
IndirectC,
Register(InstrRegister),
IndirectRegister(InstrRegisterPair),
ByteAtAddress(u16),
ImmediateWord(u16),
ImmediateByte(u8),
RegisterPair(RegisterPair),
ByteAtAddressWithOffset(u8),
}
#[derive(Copy, Clone)]
pub enum InstrRegisterPair {
AF,
BC,
DE,
HL,
SP,
PC,
IncrementHL,
DecrementHL,
}
#[derive(Debug, Copy, Clone)]
pub enum InstrRegister {
A,
B,
C,
D,
E,
H,
L,
IndirectHL, // (HL)
}
#[derive(Debug, Copy, Clone)]
pub enum JumpCondition {
NotZero,
Zero,
NotCarry,
Carry,
Always,
}
#[derive(Debug, Copy, Clone)]
struct Table;
#[derive(Debug, Copy, Clone, PartialEq, Eq, PartialOrd, Ord, Default)]
pub struct Cycle(u32);
impl Instruction {
pub fn execute(cpu: &mut Cpu, instruction: Self) -> Cycle {
match instruction {
Instruction::NOP => Cycle::new(4),
Instruction::LD(lhs, rhs) => match (lhs, rhs) {
(LDTarget::ByteAtAddress(nn), LDTarget::RegisterPair(RegisterPair::SP)) => {
// LD (nn), SP | Put Stack Pointer at address nn
cpu.write_word(nn, cpu.register_pair(RegisterPair::SP));
Cycle::new(20)
}
(LDTarget::RegisterPair(pair), LDTarget::ImmediateWord(nn)) => {
// LD rp[p], nn | Put value nn into register pair
use RegisterPair::*;
match pair {
BC | DE | HL | SP => cpu.set_register_pair(pair, nn),
_ => unreachable!("There is no \"LD {:?}, nn\" instruction", pair),
}
Cycle::new(12)
}
(LDTarget::IndirectRegister(pair), LDTarget::Register(InstrRegister::A)) => {
let a = cpu.register(Register::A);
match pair {
InstrRegisterPair::BC | InstrRegisterPair::DE => {
// LD (BC), A | Put A into memory address BC
// LD (DE), A | Put A into memory address DE
let addr = cpu.register_pair(pair.to_register_pair());
cpu.write_byte(addr, a);
}
InstrRegisterPair::IncrementHL => {
// LD (HL+), A | Put A into byte at address HL, then increment HL
let addr = cpu.register_pair(RegisterPair::HL);
cpu.write_byte(addr, a);
cpu.set_register_pair(RegisterPair::HL, addr + 1);
}
InstrRegisterPair::DecrementHL => {
// LD (HL-), A | Put A into byte at address HL, then decrement HL
let addr = cpu.register_pair(RegisterPair::HL);
cpu.write_byte(addr, a);
cpu.set_register_pair(RegisterPair::HL, addr - 1);
}
_ => unreachable!("There is no \"LD ({:?}), A\" instruction", pair),
}
Cycle::new(8)
}
(LDTarget::Register(InstrRegister::A), LDTarget::IndirectRegister(pair)) => {
match pair {
InstrRegisterPair::BC | InstrRegisterPair::DE => {
// LD A, (BC) | Put value at address BC into A
// LD A, (DE) | Put value at address DE into A
let addr = cpu.register_pair(pair.to_register_pair());
let byte = cpu.read_byte(addr);
cpu.set_register(Register::A, byte);
}
InstrRegisterPair::IncrementHL => {
// LD A, (HL+) | Put value at address HL into A, then increment HL
let addr = cpu.register_pair(RegisterPair::HL);
let byte = cpu.read_byte(addr);
cpu.set_register(Register::A, byte);
cpu.set_register_pair(RegisterPair::HL, addr + 1);
}
InstrRegisterPair::DecrementHL => {
// LD A, (HL-) | Put value at address HL into A, then increment HL
let addr = cpu.register_pair(RegisterPair::HL);
let byte = cpu.read_byte(addr);
cpu.set_register(Register::A, byte);
cpu.set_register_pair(RegisterPair::HL, addr - 1);
}
_ => unreachable!("There is no \"LD A, ({:?})\" instruction", pair),
}
Cycle::new(8)
}
(LDTarget::Register(reg), LDTarget::ImmediateByte(n)) => {
// LD r[y], n | Store n in Register
use InstrRegister::*;
match reg {
A | B | C | D | E | H | L => {
cpu.set_register(reg.to_register(), n);
Cycle::new(8)
}
IndirectHL => {
let addr = cpu.register_pair(RegisterPair::HL);
cpu.write_byte(addr, n);
Cycle::new(12)
}
}
}
(LDTarget::IndirectC, LDTarget::Register(InstrRegister::A)) => {
// LD (0xFF00 + C), A | Store value of register A at address 0xFF00 + C
let addr = 0xFF00 + cpu.register(Register::C) as u16;
cpu.write_byte(addr, cpu.register(Register::A));
Cycle::new(8)
}
(LDTarget::Register(InstrRegister::A), LDTarget::IndirectC) => {
let addr = 0xFF00 + cpu.register(Register::C) as u16;
let byte = cpu.read_byte(addr);
cpu.set_register(Register::A, byte);
Cycle::new(8)
}
(LDTarget::Register(lhs), LDTarget::Register(rhs)) => {
// LD r[y], r[z] | Store value of RHS Register in LHS Register
use InstrRegister::*;
match rhs {
B | C | D | E | H | L | A => {
let right = cpu.register(rhs.to_register());
match lhs {
B | C | D | E | H | L | A => {
cpu.set_register(lhs.to_register(), right);
Cycle::new(4)
}
IndirectHL => {
let addr = cpu.register_pair(RegisterPair::HL);
cpu.write_byte(addr, right);
Cycle::new(8)
}
}
}
IndirectHL => {
let addr = cpu.register_pair(RegisterPair::HL);
let right = cpu.read_byte(addr);
match lhs {
B | C | D | E | H | L | A => {
cpu.set_register(lhs.to_register(), right);
Cycle::new(8)
}
IndirectHL => {
unreachable!(
"There is no \"LD ({:?}), ({:?})\" instruction",
lhs, rhs
)
}
}
}
}
}
(LDTarget::ByteAtAddressWithOffset(n), LDTarget::Register(InstrRegister::A)) => {
// LD (0xFF00 + n), A | Store register A at address (0xFF00 + n)
cpu.write_byte(0xFF00 + (n as u16), cpu.register(Register::A));
Cycle::new(12)
}
(LDTarget::Register(InstrRegister::A), LDTarget::ByteAtAddressWithOffset(n)) => {
// LD A, (0xFF00 + n) | Store value at address (0xFF00 + n) in register A
let byte = cpu.read_byte(0xFF00 + (n as u16));
cpu.set_register(Register::A, byte);
Cycle::new(12)
}
(
LDTarget::RegisterPair(RegisterPair::SP),
LDTarget::RegisterPair(RegisterPair::HL),
) => {
// LD SP, HL | Load Register HL into Register SP
cpu.set_register_pair(RegisterPair::SP, cpu.register_pair(RegisterPair::HL));
Cycle::new(8)
}
(LDTarget::ByteAtAddress(nn), LDTarget::Register(InstrRegister::A)) => {
cpu.write_byte(nn, cpu.register(Register::A));
Cycle::new(16)
}
(LDTarget::Register(InstrRegister::A), LDTarget::ByteAtAddress(nn)) => {
let byte = cpu.read_byte(nn);
cpu.set_register(Register::A, byte);
Cycle::new(16)
}
_ => unreachable!("There is no \"LD {:?}, {:?}\" instruction", lhs, rhs),
},
Instruction::STOP => Cycle::new(4),
Instruction::JR(cond, offset) => {
// JR cc[y - 4], d | If condition is true, then add d to current address and jump
// JR d | Add d to current address and jump
let flags: &Flags = cpu.flags();
let prev = cpu.register_pair(RegisterPair::PC);
let addr = Self::add_u16_i8_no_flags(prev, offset);
match cond {
JumpCondition::Always => {
cpu.set_register_pair(RegisterPair::PC, addr);
Cycle::new(12)
}
JumpCondition::NotZero => {
if !flags.z() {
cpu.set_register_pair(RegisterPair::PC, addr);
Cycle::new(12)
} else {
Cycle::new(8)
}
}
JumpCondition::Zero => {
if flags.z() {
cpu.set_register_pair(RegisterPair::PC, addr);
Cycle::new(12)
} else {
Cycle::new(8)
}
}
JumpCondition::NotCarry => {
if !flags.c() {
cpu.set_register_pair(RegisterPair::PC, addr);
Cycle::new(12)
} else {
Cycle::new(8)
}
}
JumpCondition::Carry => {
if flags.c() {
cpu.set_register_pair(RegisterPair::PC, addr);
Cycle::new(12)
} else {
Cycle::new(8)
}
}
}
}
Instruction::ADD(lhs, rhs) => match (lhs, rhs) {
(MATHTarget::RegisterPair(RegisterPair::HL), MATHTarget::RegisterPair(pair)) => {
// ADD HL, rp[p] | add register pair to HL.
use RegisterPair::*;
let mut flags: Flags = *cpu.flags();
match pair {
BC | DE | HL | SP => {
let hl_value = cpu.register_pair(RegisterPair::HL);
let value = cpu.register_pair(pair);
let sum = Self::add_u16s(hl_value, value, &mut flags);
cpu.set_register_pair(RegisterPair::HL, sum);
}
_ => unreachable!("There is no \"ADD HL, {:?}\" instruction", pair),
}
cpu.set_flags(flags);
Cycle::new(8)
}
(MATHTarget::Register(InstrRegister::A), MATHTarget::Register(reg)) => {
// ADD A, r[z] | Add (A + r[z]) to register A
use InstrRegister::*;
let mut flags: Flags = *cpu.flags();
let a_value = cpu.register(Register::A);
let (cycles, sum) = match reg {
B | C | D | E | H | L | A => {
let value = cpu.register(reg.to_register());
let sum = Self::add_u8s(a_value, value, &mut flags);
(Cycle::new(4), sum)
}
IndirectHL => {
let addr = cpu.register_pair(RegisterPair::HL);
let value = cpu.read_byte(addr);
let sum = Self::add_u8s(a_value, value, &mut flags);
(Cycle::new(8), sum)
}
};
cpu.set_register(Register::A, sum);
cpu.set_flags(flags);
cycles
}
(MATHTarget::RegisterPair(RegisterPair::SP), MATHTarget::ImmediateByte(d)) => {
// ADD SP, d | Add d (is signed) to register pair SP.
let mut flags: Flags = *cpu.flags();
let d = d as i8;
let sum = Self::add_u16_i8(cpu.register_pair(RegisterPair::SP), d, &mut flags);
cpu.set_flags(flags);
cpu.set_register_pair(RegisterPair::SP, sum);
Cycle::new(16)
}
(MATHTarget::Register(InstrRegister::A), MATHTarget::ImmediateByte(n)) => {
// ADD A, n | Add n to register A
let mut flags: Flags = *cpu.flags();
let sum = Self::add_u8s(cpu.register(Register::A), n, &mut flags);
cpu.set_register(Register::A, sum);
cpu.set_flags(flags);
Cycle::new(8)
}
_ => unreachable!("There is no \"ADD {:?}, {:?}\" instruction", lhs, rhs),
},
Instruction::INC(registers) => {
match registers {
Registers::Byte(reg) => {
// INC r[y] | Increment Register
use InstrRegister::*;
let mut flags: Flags = *cpu.flags();
let cycles = match reg {
B | C | D | E | H | L | A => {
let reg = reg.to_register();
let value = cpu.register(reg);
cpu.set_register(reg, Self::inc_register(value, &mut flags));
Cycle::new(4)
}
IndirectHL => {
let addr = cpu.register_pair(RegisterPair::HL);
let byte = Self::inc_register(cpu.read_byte(addr), &mut flags);
cpu.write_byte(addr, byte);
Cycle::new(12)
}
};
cpu.set_flags(flags);
cycles
}
Registers::Word(pair) => {
// INC rp[p] | Increment Register Pair
// Note: According to RGBDS, no flags are set here.
use RegisterPair::*;
match pair {
BC | DE | HL | SP => {
let value = cpu.register_pair(pair);
cpu.set_register_pair(pair, value.wrapping_add(1));
}
_ => unreachable!("There is no \"INC {:?}\" instruction", pair),
}
Cycle::new(8)
}
}
}
Instruction::DEC(registers) => {
match registers {
Registers::Byte(reg) => {
// DEC r[y] | Decrement Register
use InstrRegister::*;
let mut flags: Flags = *cpu.flags();
let cycles = match reg {
B | C | D | E | H | L | A => {
let reg = reg.to_register();
let value = cpu.register(reg);
cpu.set_register(reg, Self::dec_register(value, &mut flags));
Cycle::new(4)
}
IndirectHL => {
let addr = cpu.register_pair(RegisterPair::HL);
let byte = cpu.read_byte(addr);
cpu.write_byte(addr, Self::dec_register(byte, &mut flags));
Cycle::new(12)
}
};
cpu.set_flags(flags);
cycles
}
Registers::Word(pair) => {
// DEC rp[p] | Decrement Register Pair
use RegisterPair::*;
match pair {
BC | DE | HL | SP => {
let value = cpu.register_pair(pair);
cpu.set_register_pair(pair, value.wrapping_sub(1));
}
_ => unreachable!("There is no \"DEC {:?}\" instruction", pair),
};
Cycle::new(8)
}
}
}
Instruction::RLCA => {
// Rotate Register A left
let mut flags: Flags = *cpu.flags();
let a = cpu.register(Register::A);
let msb = a >> 7;
let rot_a = a.rotate_left(1);
flags.update(false, false, false, msb == 0x01);
cpu.set_flags(flags);
cpu.set_register(Register::A, rot_a);
Cycle::new(4)
}
Instruction::RRCA => {
// Rotate Register A right
let mut flags: Flags = *cpu.flags();
let a = cpu.register(Register::A);
let lsb = a & 0x01;
let rot_a = a.rotate_right(1);
flags.update(false, false, false, lsb == 0x01);
cpu.set_flags(flags);
cpu.set_register(Register::A, rot_a);
Cycle::new(4)
}
Instruction::RLA => {
// Rotate register A left through carry
let mut flags: Flags = *cpu.flags();
let a = cpu.register(Register::A);
let (rot_a, carry) = Self::rl_thru_carry(a, flags.c());
flags.update(false, false, false, carry);
cpu.set_flags(flags);
cpu.set_register(Register::A, rot_a);
Cycle::new(4)
}
Instruction::RRA => {
// Rotate register A right through carry
let mut flags: Flags = *cpu.flags();
let a = cpu.register(Register::A);
let (rot_a, carry) = Self::rr_thru_carry(a, flags.c());
flags.update(false, false, false, carry);
cpu.set_flags(flags);
cpu.set_register(Register::A, rot_a);
Cycle::new(4)
}
Instruction::DAA => {
// source: https://ehaskins.com/2018-01-30%20Z80%20DAA/
// TODO: Maybe i16 isn't the right choice here?
let mut correction: i16 = 0;
let mut value = cpu.register(Register::A) as i16;
let mut flags = *cpu.flags();
if flags.h() || (!flags.n() && (value & 0xF) > 9) {
correction |= 0x06;
}
if flags.c() || (!flags.n() && value > 0x99) {
correction |= 0x60;
flags.set_c(true);
}
value += if flags.n() { -correction } else { correction };
let result = value as u8;
flags.set_z(result == 0);
flags.set_h(false);
cpu.set_flags(flags);
cpu.set_register(Register::A, value as u8);
Cycle::new(4)
}
Instruction::CPL => {
// Compliment A register (inverse)
let mut flags: Flags = *cpu.flags();
let a = cpu.register(Register::A);
flags.set_n(true);
flags.set_h(true);
cpu.set_flags(flags);
cpu.set_register(Register::A, !a); // Bitwise not is ! instead of ~
Cycle::new(4)
}
Instruction::SCF => {
// Set Carry Flag
let mut flags: Flags = *cpu.flags();
flags.set_n(false);
flags.set_h(false);
flags.set_c(true);
cpu.set_flags(flags);
Cycle::new(4)
}
Instruction::CCF => {
// Compliment Carry Flag (inverse)
let mut flags: Flags = *cpu.flags();
flags.set_n(false);
flags.set_h(false);
flags.set_c(!flags.c());
cpu.set_flags(flags);
Cycle::new(4)
}
Instruction::HALT => {
// Enter CPU low power consumption mode until interrupt occurs
use HaltState::*;
let req = cpu.read_byte(0xFF0F);
let enabled = cpu.read_byte(0xFFFF);
let halt_state = if let ImeState::Enabled = cpu.ime() {
ImeEnabled
} else if req & enabled != 0 {
SomePending
} else {
NonePending
};
cpu.halt(halt_state);
// Though this can actually last forever
Cycle::new(4)
}
Instruction::ADC(target) => match target {
MATHTarget::Register(reg) => {
// ADC A, r[z] | Add register r[z] plus the Carry flag to A
use InstrRegister::*;
let mut flags: Flags = *cpu.flags();
let a_value = cpu.register(Register::A);
let (cycles, sum) = match reg {
B | C | D | E | H | L | A => {
let value = cpu.register(reg.to_register()) + (flags.c() as u8);
let sum = Self::add_u8s(a_value, value, &mut flags);
(Cycle::new(4), sum)
}
IndirectHL => {
let value = cpu.read_byte(cpu.register_pair(RegisterPair::HL))
+ (flags.c() as u8);
let sum = Self::add_u8s(a_value, value, &mut flags);
(Cycle::new(8), sum)
}
};
cpu.set_register(Register::A, sum);
cpu.set_flags(flags);
cycles
}
MATHTarget::ImmediateByte(n) => {
// ADC A, n | Add immediate byte plus the carry flag to A
let mut flags: Flags = *cpu.flags();
let value = n + (flags.c() as u8);
let sum = Self::add_u8s(cpu.register(Register::A), value, &mut flags);
cpu.set_flags(flags);
cpu.set_register(Register::A, sum);
Cycle::new(8)
}
_ => unreachable!("There is no \"ADC {:?}\" instruction", target),
},
Instruction::SUB(target) => match target {
MATHTarget::Register(reg) => {
// SUB r[z] | Subtract the value in register r[z] from register A, then store in A
use InstrRegister::*;
let mut flags: Flags = *cpu.flags();
let a_value = cpu.register(Register::A);
let (cycles, diff) = match reg {
B | C | D | E | H | L | A => {
let value = cpu.register(reg.to_register());
let diff = Self::sub_u8s(a_value, value, &mut flags);
(Cycle::new(4), diff)
}
IndirectHL => {
let value = cpu.read_byte(cpu.register_pair(RegisterPair::HL));
let diff = Self::sub_u8s(a_value, value, &mut flags);
(Cycle::new(8), diff)
}
};
cpu.set_register(Register::A, diff);
cpu.set_flags(flags);
cycles
}
MATHTarget::ImmediateByte(n) => {
// SUB n | Subtract the immediate byte from register A, then store in A
let mut flags: Flags = *cpu.flags();
let diff = Self::sub_u8s(cpu.register(Register::A), n, &mut flags);
cpu.set_flags(flags);
cpu.set_register(Register::A, diff);
Cycle::new(8)
}
_ => unreachable!("There is no \"SUB {:?}\" instruction", target),
},
Instruction::SBC(target) => match target {
MATHTarget::Register(reg) => {
// SBC A, r[z] | Subtract the value from register r[z] from A, add the Carry flag and then store in A
// FIXME: See ADC, is this a correct understanding of this Instruction
use InstrRegister::*;
let mut flags: Flags = *cpu.flags();
let a_value = cpu.register(Register::A);
let (cycles, diff) = match reg {
B | C | D | E | H | L | A => {
let value = cpu.register(reg.to_register()) + (flags.c() as u8);
let diff = Self::sub_u8s(a_value, value, &mut flags);
(Cycle::new(4), diff)
}
IndirectHL => {
let value = cpu.read_byte(cpu.register_pair(RegisterPair::HL))
+ (flags.c() as u8);
let diff = Self::sub_u8s(a_value, value, &mut flags);
(Cycle::new(8), diff)
}
};
cpu.set_register(Register::A, diff);
cpu.set_flags(flags);
cycles
}
MATHTarget::ImmediateByte(n) => {
// SBC A, n | Subtract the value from immediate byte from A, add the carry flag and then store in A
// FIXME: The Fixme above applies to this variant as well
let mut flags: Flags = *cpu.flags();
let value = n + (flags.c() as u8);
let diff = Self::sub_u8s(cpu.register(Register::A), value, &mut flags);
cpu.set_flags(flags);
cpu.set_register(Register::A, diff);
Cycle::new(8)
}
_ => unreachable!("There is no \"SBC {:?}\" instruction", target),
},
Instruction::AND(target) => match target {
MATHTarget::Register(reg) => {
// AND r[z] | Bitwise AND register r[z] and register A, store in register A
use InstrRegister::*;
let mut flags: Flags = *cpu.flags();
let a_value = cpu.register(Register::A);
let (cycles, result) = match reg {
B | C | D | E | H | L | A => {
let value = cpu.register(reg.to_register());
(Cycle::new(4), a_value & value)
}
IndirectHL => {
let value = cpu.read_byte(cpu.register_pair(RegisterPair::HL));
(Cycle::new(8), a_value & value)
}
};
flags.update(result == 0, false, true, false);
cpu.set_register(Register::A, result);
cpu.set_flags(flags);
cycles
}
MATHTarget::ImmediateByte(n) => {
// AND n | Bitwise AND immediate byte and register A, store in register A
let mut flags: Flags = *cpu.flags();
let result = cpu.register(Register::A) & n;
flags.update(result == 0, false, true, false);
cpu.set_register(Register::A, result);
cpu.set_flags(flags);
Cycle::new(8)
}
_ => unreachable!("There is no \"AND {:?}\" instruction", target),
},
Instruction::XOR(target) => match target {
MATHTarget::Register(reg) => {
// XOR r[z] | Bitwise XOR register r[z] and register A, store in register A
use InstrRegister::*;
let mut flags: Flags = *cpu.flags();
let a_value = cpu.register(Register::A);
let (cycles, result) = match reg {
B | C | D | E | H | L | A => {
let value = cpu.register(reg.to_register());
(Cycle::new(4), a_value ^ value)
}
IndirectHL => {
let value = cpu.read_byte(cpu.register_pair(RegisterPair::HL));
(Cycle::new(8), a_value ^ value)
}
};
flags.update(result == 0, false, false, false);
cpu.set_flags(flags);
cpu.set_register(Register::A, result);
cycles
}
MATHTarget::ImmediateByte(n) => {
// XOR n | Bitwise XOR immediate byte and register A, store in register A
let mut flags: Flags = *cpu.flags();
let result = cpu.register(Register::A) ^ n;
flags.update(result == 0, false, false, false);
cpu.set_flags(flags);
cpu.set_register(Register::A, result);
Cycle::new(8)
}
_ => unreachable!("There is no \"XOR {:?}\" instruction", target),
},
Instruction::OR(target) => match target {
MATHTarget::Register(reg) => {
// OR r[z] | Bitwise OR register r[z] and register A, store in register A
use InstrRegister::*;
let mut flags: Flags = *cpu.flags();
let a_value = cpu.register(Register::A);
let (cycles, result) = match reg {
B | C | D | E | H | L | A => {
let value = cpu.register(reg.to_register());
(Cycle::new(4), a_value | value)
}
IndirectHL => {
let value = cpu.read_byte(cpu.register_pair(RegisterPair::HL));
(Cycle::new(8), a_value | value)
}
};
flags.update(result == 0, false, false, false);
cpu.set_flags(flags);
cpu.set_register(Register::A, result);
cycles
}
MATHTarget::ImmediateByte(n) => {
// OR n | Bitwise OR on immediate byte n and register A, store in register A
let mut flags: Flags = *cpu.flags();
let result = cpu.register(Register::A) | n;
flags.update(result == 0, false, false, false);
cpu.set_flags(flags);
cpu.set_register(Register::A, result);
Cycle::new(8)
}
_ => unreachable!("There is no \"OR {:?}\" instruction", target),
},
Instruction::CP(target) => match target {
MATHTarget::Register(reg) => {
// CP r[z] | Same behaviour as SUB, except the result is not stored.
use InstrRegister::*;
let mut flags: Flags = *cpu.flags();
let a_value = cpu.register(Register::A);
let cycles = match reg {
B | C | D | E | H | L | A => {
let value = cpu.register(reg.to_register());
let _ = Self::sub_u8s(a_value, value, &mut flags);
Cycle::new(4)
}
IndirectHL => {
let value = cpu.read_byte(cpu.register_pair(RegisterPair::HL));
let _ = Self::sub_u8s(a_value, value, &mut flags);
Cycle::new(8)
}
};
cpu.set_flags(flags);
cycles
}
MATHTarget::ImmediateByte(n) => {
// CP n | Same behaviour as SUB, except the result is not stored,
let mut flags: Flags = *cpu.flags();
let _ = Self::sub_u8s(cpu.register(Register::A), n, &mut flags);
cpu.set_flags(flags);
Cycle::new(8)
}
_ => unreachable!("There is no \"CP {:?}\" instruction", target),
},
Instruction::RET(cond) => {
// RET cc[y] | Essentially a POP PC, Return from Subroutine
// RET | Essentially a POP PC, Return from Subroutine
let flags: &Flags = cpu.flags();
match cond {
JumpCondition::NotZero => {
if !flags.z() {
let addr = Self::pop(cpu);
cpu.set_register_pair(RegisterPair::PC, addr);
Cycle::new(20)
} else {
Cycle::new(8)
}
}
JumpCondition::Zero => {
if flags.z() {
let addr = Self::pop(cpu);
cpu.set_register_pair(RegisterPair::PC, addr);
Cycle::new(20)
} else {
Cycle::new(8)
}
}
JumpCondition::NotCarry => {
if !flags.c() {
let addr = Self::pop(cpu);
cpu.set_register_pair(RegisterPair::PC, addr);
Cycle::new(20)
} else {
Cycle::new(8)
}
}
JumpCondition::Carry => {
if flags.c() {
let addr = Self::pop(cpu);
cpu.set_register_pair(RegisterPair::PC, addr);
Cycle::new(20)
} else {
Cycle::new(8)
}
}
JumpCondition::Always => {
let addr = Self::pop(cpu);
cpu.set_register_pair(RegisterPair::PC, addr);
Cycle::new(16)
}
}
}
Instruction::LDHL(d) => {
// LDHL SP + d | Add SP + d to register HL
// LD HL, SP + d | Add SP + d to register HL
let mut flags: Flags = *cpu.flags();
let sum = Self::add_u16_i8(cpu.register_pair(RegisterPair::SP), d, &mut flags);
cpu.set_register_pair(RegisterPair::HL, sum);
cpu.set_flags(flags);
Cycle::new(12)
}
Instruction::POP(pair) => {
// POP rp2[p] | Pop from stack into register pair rp2[]
// Flags are set when we call cpu.set_register_pair(RegisterPair::AF, value);
use RegisterPair::*;
match pair {
BC | DE | HL | AF => {
let value = Self::pop(cpu);
cpu.set_register_pair(pair, value);
}
_ => unreachable!("There is no \"POP {:?}\" instruction", pair),
}
Cycle::new(12)
}
Instruction::RETI => {
// Same as RET, after which interrupts are enabled.
let addr = Self::pop(cpu);
cpu.set_register_pair(RegisterPair::PC, addr);
cpu.set_ime(ImeState::Enabled);
Cycle::new(16)
}
Instruction::JP(cond, target) => match target {
JPTarget::RegisterPair(RegisterPair::HL) => {
// JP HL | Load register pair HL into program counter
let addr = cpu.register_pair(RegisterPair::HL);
cpu.set_register_pair(RegisterPair::PC, addr);
Cycle::new(4)
}
JPTarget::ImmediateWord(nn) => {
// JP cc[y], nn | Store Immediate Word in the Program Counter if cond is met
// JP nn | Store Immediate Word in the Program Counter
let flags: &Flags = cpu.flags();
match cond {
JumpCondition::NotZero => {
if !flags.z() {
cpu.set_register_pair(RegisterPair::PC, nn);
Cycle::new(16)
} else {
Cycle::new(12)
}
}
JumpCondition::Zero => {
if flags.z() {
cpu.set_register_pair(RegisterPair::PC, nn);
Cycle::new(16)
} else {
Cycle::new(12)
}
}
JumpCondition::NotCarry => {
if !flags.c() {
cpu.set_register_pair(RegisterPair::PC, nn);
Cycle::new(16)
} else {
Cycle::new(12)
}
}
JumpCondition::Carry => {
if flags.c() {
cpu.set_register_pair(RegisterPair::PC, nn);
Cycle::new(16)
} else {
Cycle::new(12)
}
}
JumpCondition::Always => {
cpu.set_register_pair(RegisterPair::PC, nn);
Cycle::new(16)
}
}
}
_ => unreachable!("There is no \"JP {:?}\" instruction", target),
},
Instruction::DI => {
// Disable IME
cpu.set_ime(ImeState::Disabled);
Cycle::new(4)
}
Instruction::EI => {
// Enable IME (After the next instruction)
// FIXME: IME is set after the next instruction, this currently is not represented in this emulator.
cpu.set_ime(ImeState::Pending);
Cycle::new(4)
}
Instruction::CALL(cond, nn) => {
// CALL cc[y], nn | Store pc on the stack, then store nn in the program counter if cond is met
// CALL nn | Store nn on the stack, then store nn in the program counter
let flags: &Flags = cpu.flags();
let pc = cpu.register_pair(RegisterPair::PC);
match cond {
JumpCondition::NotZero => {
if !flags.z() {
Self::push(cpu, pc);
cpu.set_register_pair(RegisterPair::PC, nn);
Cycle::new(24)
} else {
Cycle::new(12)
}
}
JumpCondition::Zero => {
if flags.z() {
Self::push(cpu, pc);
cpu.set_register_pair(RegisterPair::PC, nn);
Cycle::new(24)
} else {
Cycle::new(12)
}
}
JumpCondition::NotCarry => {
if !flags.c() {
Self::push(cpu, pc);
cpu.set_register_pair(RegisterPair::PC, nn);
Cycle::new(24)
} else {
Cycle::new(12)
}
}
JumpCondition::Carry => {
if flags.c() {
Self::push(cpu, pc);
cpu.set_register_pair(RegisterPair::PC, nn);
Cycle::new(24)
} else {
Cycle::new(12)
}
}
JumpCondition::Always => {
Self::push(cpu, pc);
cpu.set_register_pair(RegisterPair::PC, nn);
Cycle::new(24)
}
}
}
Instruction::PUSH(pair) => {
// PUSH rp2[p] | Push register pair onto the stack
use RegisterPair::*;
match pair {
BC | DE | HL | AF => {
let value = cpu.register_pair(pair);
Self::push(cpu, value);
}
_ => unreachable!("There is no \"PUSH {:?}\" instruction", pair),
}
Cycle::new(16)
}
Instruction::RST(n) => {
// RST n | Push current address onto the stack, jump to 0x0000 + n
let addr = cpu.register_pair(RegisterPair::PC);
Self::push(cpu, addr);
cpu.set_register_pair(RegisterPair::PC, n as u16);
Cycle::new(16)
}
Instruction::RLC(reg) => {
// RLC r[z] | Rotate register r[z] left
use InstrRegister::*;
let mut flags: Flags = *cpu.flags();
let (cycles, msb, rotated) = match reg {
B | C | D | E | H | L | A => {
let register = reg.to_register();
let value = cpu.register(register);
let rotated = value.rotate_left(1);
cpu.set_register(register, rotated);
(Cycle::new(8), value >> 7, rotated)
}
IndirectHL => {
let addr = cpu.register_pair(RegisterPair::HL);
let value = cpu.read_byte(addr);
let rotated = value.rotate_left(1);
cpu.write_byte(addr, rotated);
(Cycle::new(16), value >> 7, rotated)
}
};
flags.update(rotated == 0, false, false, msb == 0x01);
cpu.set_flags(flags);
cycles
}
Instruction::RRC(reg) => {
// RRC r[z] | Rotate Register r[z] right
use InstrRegister::*;
let mut flags: Flags = *cpu.flags();
let (cycles, lsb, rotated) = match reg {
B | C | D | E | H | L | A => {
let register = reg.to_register();
let value = cpu.register(register);
let rotated = value.rotate_right(1);
cpu.set_register(register, rotated);
(Cycle::new(8), value & 0x01, rotated)
}
IndirectHL => {
let addr = cpu.register_pair(RegisterPair::HL);
let value = cpu.read_byte(addr);
let rotated = value.rotate_right(1);
cpu.write_byte(addr, rotated);
(Cycle::new(16), value & 0x01, rotated)
}
};
flags.update(rotated == 0, false, false, lsb == 0x01);
cpu.set_flags(flags);
cycles
}
Instruction::RL(reg) => {
// RL r[z] | Rotate register r[z] left through carry
use InstrRegister::*;
let mut flags: Flags = *cpu.flags();
let (cycles, rotated, carry) = match reg {
B | C | D | E | H | L | A => {
let register = reg.to_register();
let value = cpu.register(register);
let (rotated, carry) = Self::rl_thru_carry(value, flags.c());
cpu.set_register(register, rotated);
(Cycle::new(8), rotated, carry)
}
IndirectHL => {
let addr = cpu.register_pair(RegisterPair::HL);
let value = cpu.read_byte(addr);
let (rotated, carry) = Self::rl_thru_carry(value, flags.c());
cpu.write_byte(addr, rotated);
(Cycle::new(16), rotated, carry)
}
};
flags.update(rotated == 0, false, false, carry);
cpu.set_flags(flags);
cycles
}
Instruction::RR(reg) => {
// RR r[z] | Rotate register r[z] right through carry
use InstrRegister::*;
let mut flags: Flags = *cpu.flags();
let (cycles, rotated, carry) = match reg {
B | C | D | E | H | L | A => {
let register = reg.to_register();
let value = cpu.register(register);
let (rotated, carry) = Self::rr_thru_carry(value, flags.c());
cpu.set_register(register, rotated);
(Cycle::new(8), rotated, carry)
}
IndirectHL => {
let addr = cpu.register_pair(RegisterPair::HL);
let value = cpu.read_byte(addr);
let (rotated, carry) = Self::rr_thru_carry(value, flags.c());
cpu.write_byte(addr, rotated);
(Cycle::new(16), rotated, carry)
}
};
flags.update(rotated == 0, false, false, carry);
cpu.set_flags(flags);
cycles
}
Instruction::SLA(reg) => {
// SLA r[z] | Shift left arithmetic register r[z]
use InstrRegister::*;
let mut flags: Flags = *cpu.flags();
let (cycles, msb, shifted) = match reg {
B | C | D | E | H | L | A => {
let register = reg.to_register();
let value = cpu.register(register);
let shifted = value << 1;
cpu.set_register(register, shifted);
(Cycle::new(8), (value >> 7) & 0x01, shifted)
}
IndirectHL => {
let addr = cpu.register_pair(RegisterPair::HL);
let value = cpu.read_byte(addr);
let shifted = value << 1;
cpu.write_byte(addr, value);
(Cycle::new(16), (value >> 7) & 0x01, shifted)
}
};
flags.update(shifted == 0, false, false, msb == 0x01);
cpu.set_flags(flags);
cycles
}
Instruction::SRA(reg) => {
// SRA r[z] | Shift right arithmetic register r[z]
use InstrRegister::*;
let mut flags: Flags = *cpu.flags();
let (cycles, lsb, shifted) = match reg {
B | C | D | E | H | L | A => {
let register = reg.to_register();
let value = cpu.register(register);
let msb = (value >> 7) & 0x01;
let shifted = msb << 7 | value >> 1;
cpu.set_register(register, shifted);
(Cycle::new(8), value & 0x01, shifted)
}
IndirectHL => {
let addr = cpu.register_pair(RegisterPair::HL);
let value = cpu.read_byte(addr);
let msb = (value >> 7) & 0x01;
let shifted = msb << 7 | value >> 1;
cpu.write_byte(addr, value);
(Cycle::new(16), value & 0x01, shifted)
}
};
flags.update(shifted == 0, false, false, lsb == 0x01);
cpu.set_flags(flags);
cycles
}
Instruction::SWAP(reg) => {
// SWAP r[z] | Swap the 4 highest and lowest bits in a byte
use InstrRegister::*;
let mut flags: Flags = *cpu.flags();
let (cycles, swapped) = match reg {
B | C | D | E | H | L | A => {
let register = reg.to_register();
let value = cpu.register(register);
let swapped = Self::swap_bits(value);
cpu.set_register(register, swapped);
(Cycle::new(8), swapped)
}
IndirectHL => {
let addr = cpu.register_pair(RegisterPair::HL);
let value = cpu.read_byte(addr);
let swapped = Self::swap_bits(value);
cpu.write_byte(addr, swapped);
(Cycle::new(16), swapped)
}
};
flags.update(swapped == 0, false, false, false);
cpu.set_flags(flags);
cycles
}
Instruction::SRL(reg) => {
// SRL r[z] | Shift right logic register r[z]
use InstrRegister::*;
let mut flags: Flags = *cpu.flags();
let (cycles, lsb, shift_reg) = match reg {
B | C | D | E | H | L | A => {
let register = reg.to_register();
let value = cpu.register(register);
let shifted = value >> 1;
cpu.set_register(register, shifted);
(Cycle::new(8), value & 0x01, shifted)
}
IndirectHL => {
let addr = cpu.register_pair(RegisterPair::HL);
let value = cpu.read_byte(addr);
let shifted = value >> 1;
cpu.write_byte(addr, shifted);
(Cycle::new(16), value & 0x01, shifted)
}
};
flags.update(shift_reg == 0, false, false, lsb == 0x01);
cpu.set_flags(flags);
cycles
}
Instruction::BIT(y, reg) => {
// BIT y, r[z] | Test y is in register r[z]
use InstrRegister::*;
let mut flags: Flags = *cpu.flags();
let (cycles, is_bit_set) = match reg {
B | C | D | E | H | L | A => {
let register = reg.to_register();
let value = cpu.register(register);
let is_bit_set = ((value >> y) & 0x01) == 0x01;
(Cycle::new(8), is_bit_set)
}
IndirectHL => {
let addr = cpu.register_pair(RegisterPair::HL);
let value = cpu.read_byte(addr);
let is_bit_set = ((value >> y) & 0x01) == 0x01;
(Cycle::new(12), is_bit_set)
}
};
flags.update(!is_bit_set, false, true, flags.c());
cpu.set_flags(flags);
cycles
}
Instruction::RES(y, reg) => {
// RES y, r[z] | Reset bit y to zero
//
// 00000001 << 3 = 00001000
// ~00001000 = 11110111
// value & 11110111 means that only a specific bit will be reset
use InstrRegister::*;
match reg {
B | C | D | E | H | L | A => {
let register = reg.to_register();
let value = cpu.register(register);
cpu.set_register(register, value & !(1u8 << y));
Cycle::new(8)
}
IndirectHL => {
let addr = cpu.register_pair(RegisterPair::HL);
let value = cpu.read_byte(addr);
cpu.write_byte(addr, value & !(1u8 << y));
Cycle::new(16)
}
}
}
Instruction::SET(y, reg) => {
// BIT y, r[z] | Set bit y to one
//
// 00000001 << 3 = 00001000
// value | 00001000 means that only a specific bit will be set
use InstrRegister::*;
match reg {
B | C | D | E | H | L | A => {
let register = reg.to_register();
let value = cpu.register(register);
cpu.set_register(register, value | (1u8 << y));
Cycle::new(8)
}
IndirectHL => {
let addr = cpu.register_pair(RegisterPair::HL);
let value = cpu.read_byte(addr);
cpu.write_byte(addr, value | (1u8 << y));
Cycle::new(16)
}
}
}
}
}
/// PUSHes a u16 onto the stack
///
/// Mutates the stack pointer and the stack
fn push(cpu: &mut Cpu, value: u16) {
let mut sp = cpu.register_pair(RegisterPair::SP);
sp -= 1;
cpu.write_byte(sp, (value >> 8) as u8);
sp -= 1;
cpu.write_byte(sp, value as u8);
cpu.set_register_pair(RegisterPair::SP, sp);
}
/// POPs a u16 from the stack
///
/// Mutates the stack pointer and returns the u16 which was popped from the stack
fn pop(cpu: &mut Cpu) -> u16 {
let mut sp = cpu.register_pair(RegisterPair::SP);
let low = cpu.read_byte(sp);
sp += 1;
let high = cpu.read_byte(sp);
sp += 1;
cpu.set_register_pair(RegisterPair::SP, sp);
(high as u16) << 8 | low as u16
}
fn dec_register(byte: u8, flags: &mut Flags) -> u8 {
Self::sub_u8s_no_carry(byte, 1, flags)
}
fn inc_register(byte: u8, flags: &mut Flags) -> u8 {
Self::add_u8s_no_carry(byte, 1, flags)
}
fn sub_u8s_no_carry(left: u8, right: u8, flags: &mut Flags) -> u8 {
let diff = left.wrapping_sub(right);
flags.set_z(diff == 0);
flags.set_n(true);
flags.set_h(Self::bit_4_borrow(left, right));
diff
}
fn sub_u8s(left: u8, right: u8, flags: &mut Flags) -> u8 {
let (diff, did_overflow) = left.overflowing_sub(right);
flags.update(
diff == 0,
true,
Self::bit_4_borrow(left, right),
did_overflow,
);
diff
}
fn add_u16_i8_no_flags(left: u16, right: i8) -> u16 {
if right < 0 {
left.wrapping_sub((!right + 1) as u16)
} else {
left.wrapping_add(right as u16)
}
}
fn add_u16_i8(left: u16, right: i8, flags: &mut Flags) -> u16 {
let (_, did_overflow) = (left as u8).overflowing_add(right as u8);
let sum = Self::add_u16_i8_no_flags(left, right);
let half_carry = Self::bit_3_overflow(left as u8, right as u8);
flags.update(false, false, half_carry, did_overflow);
sum
}
fn add_u8s_no_carry(left: u8, right: u8, flags: &mut Flags) -> u8 {
let sum = left.wrapping_add(right);
flags.set_z(sum == 0);
flags.set_n(false);
flags.set_h(Self::bit_3_overflow(left, right));
sum
}
fn add_u8s(left: u8, right: u8, flags: &mut Flags) -> u8 {
let (sum, did_overflow) = left.overflowing_add(right);
flags.update(
sum == 0,
false,
Self::bit_3_overflow(left, right),
did_overflow,
);
sum
}
fn add_u16s(left: u16, right: u16, flags: &mut Flags) -> u16 {
let (sum, did_overflow) = left.overflowing_add(right);
flags.set_n(false);
flags.set_h(Self::bit_11_overflow(left, right));
flags.set_c(did_overflow);
sum
}
fn bit_11_overflow(left: u16, right: u16) -> bool {
(((left & 0x0FFF) + (right & 0x0FFF)) & 0x1000) == 0x1000
}
fn bit_3_overflow(left: u8, right: u8) -> bool {
(((left & 0xF) + (right & 0xF)) & 0x10) == 0x10
}
fn bit_4_borrow(left: u8, right: u8) -> bool {
(left & 0x0F) < (right & 0x0F)
}
fn rl_thru_carry(byte: u8, carry: bool) -> (u8, bool) {
let carry_flag = (byte >> 7) & 0x01; // get the MSB of the u8 (which will rotate into the carry bit)
let new_byte = (byte << 1) | carry as u8; // shift the bit left, and then OR the carry bit in.
(new_byte, carry_flag == 0x01)
}
fn rr_thru_carry(byte: u8, carry: bool) -> (u8, bool) {
let carry_flag = byte & 0x01; // get the LSB of the u8 (which will rotate into the carry bit)
let new_byte = ((carry as u8) << 7) | (byte >> 1); // shift the bit right, and then OR the carry bit in.
(new_byte, carry_flag == 0x01)
}
fn swap_bits(byte: u8) -> u8 {
let upper = byte >> 4;
let lower = byte & 0x0F;
(lower << 4) | upper
}
}
impl Instruction {
pub fn from_byte(cpu: &mut Cpu, byte: u8) -> Self {
if byte == 0xCB {
Self::from_prefixed_byte(cpu)
} else {
Self::from_unprefixed_byte(cpu, byte)
}
}
fn from_unprefixed_byte(cpu: &mut Cpu, opcode: u8) -> Self {
// https://gb-archive.github.io/salvage/decoding_gbz80_opcodes/Decoding%20Gamboy%20Z80%20Opcodes.html
#![allow(clippy::many_single_char_names)]
let x = (opcode >> 6) & 0x03;
let y = (opcode >> 3) & 0x07;
let z = opcode & 0x07;
let p = y >> 1;
let q = y & 0x01;
let pc = cpu.register_pair(RegisterPair::PC);
match (x, z, q, y, p) {
(0, 0, _, 0, _) => Self::NOP, // NOP
(0, 0, _, 1, _) => Self::LD(
// LD (nn), SP
LDTarget::ByteAtAddress(cpu.read_imm_word(pc)),
LDTarget::RegisterPair(RegisterPair::SP),
),
(0, 0, _, 2, _) => Self::STOP, // STOP
(0, 0, _, 3, _) => Self::JR(JumpCondition::Always, cpu.read_imm_byte(pc) as i8), // JR d
(0, 0, _, 4..=7, _) => Self::JR(Table::cc(y - 4), cpu.read_imm_byte(pc) as i8), // JR cc[y - 4], d
(0, 1, 0, _, _) => Self::LD(
// LD rp[p], nn
LDTarget::RegisterPair(Table::rp(p)),
LDTarget::ImmediateWord(cpu.read_imm_word(pc)),
),
(0, 1, 1, _, _) => Self::ADD(
// ADD HL, rp[p]
MATHTarget::RegisterPair(RegisterPair::HL),
MATHTarget::RegisterPair(Table::rp(p)),
),
(0, 2, 0, _, 0) => Self::LD(
// LD (BC), A
LDTarget::IndirectRegister(InstrRegisterPair::BC),
LDTarget::Register(InstrRegister::A),
),
(0, 2, 0, _, 1) => Self::LD(
// LD (DE), A
LDTarget::IndirectRegister(InstrRegisterPair::DE),
LDTarget::Register(InstrRegister::A),
),
(0, 2, 1, _, 0) => Self::LD(
// LD A, (BC)
LDTarget::Register(InstrRegister::A),
LDTarget::IndirectRegister(InstrRegisterPair::BC),
),
(0, 2, 1, _, 1) => Self::LD(
// LD A, (DE)
LDTarget::Register(InstrRegister::A),
LDTarget::IndirectRegister(InstrRegisterPair::DE),
),
(0, 2, 0, _, 2) => Self::LD(
// LD (HL+), A
LDTarget::IndirectRegister(InstrRegisterPair::IncrementHL),
LDTarget::Register(InstrRegister::A),
),
(0, 2, 0, _, 3) => Self::LD(
// LD (HL-), A
LDTarget::IndirectRegister(InstrRegisterPair::DecrementHL),
LDTarget::Register(InstrRegister::A),
),
(0, 2, 1, _, 2) => Self::LD(
// LD A, (HL+)
LDTarget::Register(InstrRegister::A),
LDTarget::IndirectRegister(InstrRegisterPair::IncrementHL),
),
(0, 2, 1, _, 3) => Self::LD(
// LD A, (HL-)
LDTarget::Register(InstrRegister::A),
LDTarget::IndirectRegister(InstrRegisterPair::DecrementHL),
),
(0, 3, 0, _, _) => Self::INC(
// INC rp[p]
Registers::Word(Table::rp(p)),
),
(0, 3, 1, _, _) => Self::DEC(
// DEC rp[p]
Registers::Word(Table::rp(p)),
),
(0, 4, _, _, _) => Self::INC(
// INC r[y]
Registers::Byte(Table::r(y)),
),
(0, 5, _, _, _) => Self::DEC(
// DEC r[y]
Registers::Byte(Table::r(y)),
),
(0, 6, _, _, _) => Self::LD(
// LD r[y], n
LDTarget::Register(Table::r(y)),
LDTarget::ImmediateByte(cpu.read_imm_byte(pc)),
),
(0, 7, _, 0, _) => Self::RLCA,
(0, 7, _, 1, _) => Self::RRCA,
(0, 7, _, 2, _) => Self::RLA,
(0, 7, _, 3, _) => Self::RRA,
(0, 7, _, 4, _) => Self::DAA,
(0, 7, _, 5, _) => Self::CPL,
(0, 7, _, 6, _) => Self::SCF,
(0, 7, _, 7, _) => Self::CCF,
(1, 6, _, 6, _) => Self::HALT,
(1, _, _, _, _) => Self::LD(
// LD r[y], r[z]
LDTarget::Register(Table::r(y)),
LDTarget::Register(Table::r(z)),
),
(2, _, _, _, _) => Table::x2_alu(y, z), // alu[y] r[z]
(3, 0, _, 0..=3, _) => Self::RET(Table::cc(y)), // RET cc[y]
(3, 0, _, 4, _) => Self::LD(
// LD (0xFF00 + n), A
LDTarget::ByteAtAddressWithOffset(cpu.read_imm_byte(pc)),
LDTarget::Register(InstrRegister::A),
),
(3, 0, _, 5, _) => Self::ADD(
// ADD SP, d
MATHTarget::RegisterPair(RegisterPair::SP),
MATHTarget::ImmediateByte(cpu.read_imm_byte(pc)),
),
(3, 0, _, 6, _) => Self::LD(
// LD A, (0xFF00 + n)
LDTarget::Register(InstrRegister::A),
LDTarget::ByteAtAddressWithOffset(cpu.read_imm_byte(pc)),
),
(3, 0, _, 7, _) => Self::LDHL(cpu.read_imm_byte(pc) as i8), // LD HL, SP + d
(3, 1, 0, _, _) => Self::POP(Table::rp2(p)), // POP rp2[p]
(3, 1, 1, _, 0) => Self::RET(JumpCondition::Always), // RET
(3, 1, 1, _, 1) => Self::RETI,
(3, 1, 1, _, 2) => Self::JP(
// JP HL
JumpCondition::Always,
JPTarget::RegisterPair(RegisterPair::HL),
),
(3, 1, 1, _, 3) => Self::LD(
// LD SP, HL
LDTarget::RegisterPair(RegisterPair::SP),
LDTarget::RegisterPair(RegisterPair::HL),
),
(3, 2, _, 0..=3, _) => Self::JP(
// JP cc[y], nn
Table::cc(y),
JPTarget::ImmediateWord(cpu.read_imm_word(pc)),
),
(3, 2, _, 4, _) => Self::LD(
// LD (0xFF00 + C) ,A
LDTarget::IndirectC,
LDTarget::Register(InstrRegister::A),
),
(3, 2, _, 5, _) => Self::LD(
// LD (nn), A
LDTarget::ByteAtAddress(cpu.read_imm_word(pc)),
LDTarget::Register(InstrRegister::A),
),
(3, 2, _, 6, _) => Self::LD(
// LD A, (0xFF00 + C)
LDTarget::Register(InstrRegister::A),
LDTarget::IndirectC,
),
(3, 2, _, 7, _) => Self::LD(
// LD A, (nn)
LDTarget::Register(InstrRegister::A),
LDTarget::ByteAtAddress(cpu.read_imm_word(pc)),
),
(3, 3, _, 0, _) => Self::JP(
// JP nn
JumpCondition::Always,
JPTarget::ImmediateWord(cpu.read_imm_word(pc)),
),
(3, 3, _, 1, _) => unreachable!("0xCB is handled by Instruction::from_prefixed_byte"),
// (3, 3, _, 2, _) => unreachable!("\"removed\" in documentation"),
// (3, 3, _, 3, _) => unreachable!("\"removed\" in documentation"),
// (3, 3, _, 4, _) => unreachable!("\"removed\" in documentation"),
// (3, 3, _, 5, _) => unreachable!("\"removed\" in documentation"),
(3, 3, _, 6, _) => Self::DI,
(3, 3, _, 7, _) => Self::EI,
(3, 4, _, 0..=3, _) => Self::CALL(Table::cc(y), cpu.read_imm_word(pc)), // CALL cc[y], nn
// (3, 4, _, 4..=7, _) => unreachable!("\"removed\" in documentation"),
(3, 5, 0, _, _) => Self::PUSH(Table::rp2(p)), // PUSH rp2[p]
(3, 5, 1, _, 0) => Self::CALL(JumpCondition::Always, cpu.read_imm_word(pc)), // CALL nn
// (3, 5, 1, _, 1..=3) => unreachable!("\"removed\" in documentation"),
(3, 6, _, _, _) => Table::x3_alu(y, cpu.read_imm_byte(pc)),
(3, 7, _, _, _) => Self::RST(y * 8), // RST n
_ => unreachable!(
"Unknown Opcode: {:#04X}\n x: {}, z: {}, q: {}, y: {}, p: {}",
opcode, x, z, q, y, p
),
}
}
fn from_prefixed_byte(cpu: &mut Cpu) -> Self {
#![allow(clippy::many_single_char_names)]
let pc = cpu.register_pair(RegisterPair::PC);
let opcode = cpu.read_imm_byte(pc); // FIXME: Should the PC be incremented here?
// https://gb-archive.github.io/salvage/decoding_gbz80_opcodes/Decoding%20Gamboy%20Z80%20Opcodes.html
let x = (opcode >> 6) & 0b00000011;
let y = (opcode >> 3) & 0b00000111;
let z = opcode & 0b00000111;
let p = y >> 1;
let q = y & 0b00000001;
match x {
0 => Table::rot(y, z),
1 => Self::BIT(y, Table::r(z)), // BIT y, r[z]
2 => Self::RES(y, Table::r(z)), // RES y, r[z]
3 => Self::SET(y, Table::r(z)), // SET y, r[z]
_ => unreachable!(
"Unknown Prefixed Opcode: 0xCB {:#04X}\n x: {}, z: {}, q: {}, y: {}, p: {}",
opcode, x, z, q, y, p
),
}
}
}
impl From<RegisterPair> for InstrRegisterPair {
fn from(pair: RegisterPair) -> Self {
match pair {
RegisterPair::AF => Self::AF,
RegisterPair::BC => Self::BC,
RegisterPair::DE => Self::DE,
RegisterPair::HL => Self::HL,
RegisterPair::SP => Self::SP,
RegisterPair::PC => Self::PC,
}
}
}
impl TryFrom<InstrRegisterPair> for RegisterPair {
type Error = &'static str; // FIXME: Proper error type goes here.
fn try_from(pair: InstrRegisterPair) -> Result<Self, Self::Error> {
match pair {
InstrRegisterPair::AF => Ok(Self::AF),
InstrRegisterPair::BC => Ok(Self::BC),
InstrRegisterPair::DE => Ok(Self::DE),
InstrRegisterPair::HL => Ok(Self::HL),
InstrRegisterPair::SP => Ok(Self::SP),
InstrRegisterPair::PC => Ok(Self::PC),
InstrRegisterPair::IncrementHL => {
Err("Can not convert InstrRegisterPair::IncrementHL to RegisterPair")
}
InstrRegisterPair::DecrementHL => {
Err("Can not convert InstrRegisterPair::DecrementHL to RegisterPair")
}
}
}
}
impl TryFrom<Register> for InstrRegister {
type Error = &'static str; // FIXME: Proper error type goes here
fn try_from(register: Register) -> Result<Self, Self::Error> {
match register {
Register::A => Ok(Self::A),
Register::B => Ok(Self::B),
Register::C => Ok(Self::C),
Register::D => Ok(Self::D),
Register::E => Ok(Self::E),
Register::H => Ok(Self::H),
Register::L => Ok(Self::L),
Register::Flag => Err("Can not convert Register::Flag to InstrRegister"),
}
}
}
impl TryFrom<InstrRegister> for Register {
type Error = String; // FIXME: Proper error type goes here.
fn try_from(register: InstrRegister) -> Result<Self, Self::Error> {
match register {
InstrRegister::A => Ok(Self::A),
InstrRegister::B => Ok(Self::B),
InstrRegister::C => Ok(Self::C),
InstrRegister::D => Ok(Self::D),
InstrRegister::E => Ok(Self::E),
InstrRegister::H => Ok(Self::H),
InstrRegister::L => Ok(Self::L),
InstrRegister::IndirectHL => {
Err("Can not convert InstrRegister::IndirectHL to Register".to_string())
}
}
}
}
impl Table {
pub fn r(index: u8) -> InstrRegister {
match index {
0 => InstrRegister::B,
1 => InstrRegister::C,
2 => InstrRegister::D,
3 => InstrRegister::E,
4 => InstrRegister::H,
5 => InstrRegister::L,
6 => InstrRegister::IndirectHL,
7 => InstrRegister::A,
_ => unreachable!("Index {} is out of bounds in r[]", index),
}
}
pub fn rp2(index: u8) -> RegisterPair {
match index {
0 => RegisterPair::BC,
1 => RegisterPair::DE,
2 => RegisterPair::HL,
3 => RegisterPair::AF,
_ => unreachable!("Index {} out of bounds in rp2[]", index),
}
}
pub fn rp(index: u8) -> RegisterPair {
match index {
0 => RegisterPair::BC,
1 => RegisterPair::DE,
2 => RegisterPair::HL,
3 => RegisterPair::SP,
_ => unreachable!("Index {} out of bounds in rp[]", index),
}
}
pub fn cc(index: u8) -> JumpCondition {
match index {
0 => JumpCondition::NotZero,
1 => JumpCondition::Zero,
2 => JumpCondition::NotCarry,
3 => JumpCondition::Carry,
_ => unreachable!("Index {} out of bounds in cc[]", index),
}
}
pub fn x2_alu(index: u8, r_index: u8) -> Instruction {
match index {
0 => Instruction::ADD(
// ADD A, r[z]
MATHTarget::Register(InstrRegister::A),
MATHTarget::Register(Self::r(r_index)),
),
1 => Instruction::ADC(MATHTarget::Register(Self::r(r_index))), // ADC A, r[z]
2 => Instruction::SUB(MATHTarget::Register(Self::r(r_index))), // SUB r[z]
3 => Instruction::SBC(MATHTarget::Register(Self::r(r_index))), // SBC A, r[z]
4 => Instruction::AND(MATHTarget::Register(Self::r(r_index))), // AND r[z]
5 => Instruction::XOR(MATHTarget::Register(Self::r(r_index))), // XOR r[z]
6 => Instruction::OR(MATHTarget::Register(Self::r(r_index))), // OR r[z]
7 => Instruction::CP(MATHTarget::Register(Self::r(r_index))), // CP r[z]
_ => unreachable!("Index {} is out of bounds in alu[]"),
}
}
pub fn x3_alu(index: u8, n: u8) -> Instruction {
match index {
0 => Instruction::ADD(
// ADD A, n
MATHTarget::Register(InstrRegister::A),
MATHTarget::ImmediateByte(n),
),
1 => Instruction::ADC(MATHTarget::ImmediateByte(n)), // ADC A, n
2 => Instruction::SUB(MATHTarget::ImmediateByte(n)), // SUB n
3 => Instruction::SBC(MATHTarget::ImmediateByte(n)), // SBC A, n
4 => Instruction::AND(MATHTarget::ImmediateByte(n)), // AND n
5 => Instruction::XOR(MATHTarget::ImmediateByte(n)), // XOR n
6 => Instruction::OR(MATHTarget::ImmediateByte(n)), // OR n
7 => Instruction::CP(MATHTarget::ImmediateByte(n)), // CP n
_ => unreachable!("Index {} is out of bounds in alu[]"),
}
}
pub fn rot(index: u8, r_index: u8) -> Instruction {
match index {
0 => Instruction::RLC(Self::r(r_index)), // RLC r[z]
1 => Instruction::RRC(Self::r(r_index)), // RRC r[z]
2 => Instruction::RL(Self::r(r_index)), // RL r[z]
3 => Instruction::RR(Self::r(r_index)), // RR r[z]
4 => Instruction::SLA(Self::r(r_index)), // SLA r[z]
5 => Instruction::SRA(Self::r(r_index)), // SRA r[z]
6 => Instruction::SWAP(Self::r(r_index)), // SWAP r[z]
7 => Instruction::SRL(Self::r(r_index)), // SRL r[z]
_ => unreachable!("Index {} is out of bounds in rot[]"),
}
}
}
impl std::fmt::Debug for JPTarget {
fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
match *self {
JPTarget::RegisterPair(pair) => write!(f, "{:?}", pair),
JPTarget::ImmediateWord(word) => write!(f, "{:#06X}", word),
}
}
}
impl std::fmt::Debug for LDTarget {
fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
match *self {
LDTarget::IndirectC => f.write_str("IndirectC"),
LDTarget::Register(reg) => write!(f, "{:?}", reg),
LDTarget::IndirectRegister(pair) => write!(f, "[{:?}]", pair),
LDTarget::ByteAtAddress(addr) => write!(f, "[{:#06X}]", addr),
LDTarget::ImmediateWord(word) => write!(f, "{:#06X}", word),
LDTarget::ImmediateByte(byte) => write!(f, "{:#04X}", byte),
LDTarget::RegisterPair(pair) => write!(f, "{:?}", pair),
LDTarget::ByteAtAddressWithOffset(byte) => {
write!(f, "[0xFF00 + {:#04X}]", byte)
}
}
}
}
impl std::fmt::Debug for MATHTarget {
fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
match *self {
MATHTarget::Register(reg) => write!(f, "{:?}", reg),
MATHTarget::RegisterPair(pair) => write!(f, "{:?}", pair),
MATHTarget::ImmediateByte(byte) => write!(f, "{:#04X}", byte),
}
}
}
impl std::fmt::Debug for InstrRegisterPair {
fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
match *self {
InstrRegisterPair::AF => f.write_str("AF"),
InstrRegisterPair::BC => f.write_str("BC"),
InstrRegisterPair::DE => f.write_str("DE"),
InstrRegisterPair::HL => f.write_str("HL"),
InstrRegisterPair::SP => f.write_str("SP"),
InstrRegisterPair::PC => f.write_str("PC"),
InstrRegisterPair::IncrementHL => f.write_str("HL+"),
InstrRegisterPair::DecrementHL => f.write_str("HL-"),
}
}
}
impl std::fmt::Debug for Registers {
fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
match *self {
Registers::Byte(reg) => write!(f, "{:?}", reg),
Registers::Word(pair) => write!(f, "{:?}", pair),
}
}
}
impl Cycle {
pub const fn new(num: u32) -> Self {
Self(num)
}
}
impl std::ops::Add for Cycle {
type Output = Self;
fn add(self, rhs: Self) -> Self::Output {
Self(self.0 + rhs.0)
}
}
impl std::ops::Add<u32> for Cycle {
type Output = Self;
fn add(self, rhs: u32) -> Self::Output {
Self(self.0 + rhs)
}
}
impl std::ops::AddAssign for Cycle {
fn add_assign(&mut self, rhs: Self) {
*self = Self(self.0 + rhs.0);
}
}
impl std::ops::AddAssign<u32> for Cycle {
fn add_assign(&mut self, rhs: u32) {
*self = Self(self.0 + rhs);
}
}
impl std::ops::Rem for Cycle {
type Output = Self;
fn rem(self, rhs: Self) -> Self::Output {
Self(self.0 % rhs.0)
}
}
impl std::ops::Rem<u32> for Cycle {
type Output = Self;
fn rem(self, rhs: u32) -> Self::Output {
Self(self.0 % rhs)
}
}
impl std::ops::RemAssign for Cycle {
fn rem_assign(&mut self, rhs: Self) {
*self = Self(self.0 % rhs.0);
}
}
impl std::ops::RemAssign<u32> for Cycle {
fn rem_assign(&mut self, rhs: u32) {
*self = Self(self.0 % rhs);
}
}
impl std::ops::Sub for Cycle {
type Output = Cycle;
fn sub(self, rhs: Self) -> Self::Output {
Self(self.0 - rhs.0)
}
}
impl std::ops::Sub<u32> for Cycle {
type Output = Cycle;
fn sub(self, rhs: u32) -> Self::Output {
Self(self.0 - rhs)
}
}
impl std::ops::SubAssign for Cycle {
fn sub_assign(&mut self, rhs: Self) {
*self = Self(self.0 - rhs.0);
}
}
impl std::ops::SubAssign<u32> for Cycle {
fn sub_assign(&mut self, rhs: u32) {
*self = Self(self.0 - rhs);
}
}
impl From<u32> for Cycle {
fn from(num: u32) -> Self {
Self(num)
}
}
impl From<Cycle> for u32 {
fn from(cycles: Cycle) -> Self {
cycles.0
}
}
impl InstrRegisterPair {
pub fn to_register_pair(self) -> RegisterPair {
RegisterPair::try_from(self).expect("Failed to convert InstrRegisterPair to RegisterPair")
}
}
impl InstrRegister {
pub fn to_register(self) -> Register {
Register::try_from(self).expect("Failed to convert from InstrRegister to Register")
}
}
#[cfg(test)]
mod tests {
use super::Cycle;
#[test]
fn cycle_add_works() {
let lhs: Cycle = Cycle::new(5);
let rhs: Cycle = Cycle::new(4);
assert_eq!(Cycle::new(9), rhs + lhs);
}
#[test]
fn cycle_add_assign_works() {
let mut cycles: Cycle = Cycle::new(5);
cycles += 5;
assert_eq!(Cycle::new(10), cycles);
}
}