bgourlie/rs-nes

A work-in-progress NES emulator written in Rust

rs-nes

A work-in-progress NES emulator written in rust.

Debugger

The old debugger has been removed as I refactor the code. Checkout this commit to use the debugger. The demo for the debugger can be seen on youtube.

Running

cd rs-nes
cargo run --bin native_client --release --features="native_client"  -- path/to/rom.nes

Current Status

• The CPU is fully-implemented and well-tested.
• The PPU is fairly accurately emulated but has a few minor bugs.
• Audio is not implemented yet.
• Mappers
  • NROM (Mario Bros., Super Mario Bros., Excite Bike, etc)
  • UxROM is partially implemented (Mega Man, Castlevania, Contra, etc)

Controls

W: Up
A: Left
S: Down
D: Right
J: B
K: A
Shift: Select
Enter: Start

Attribution

Much of the inline documentation for NES specific components (PPU, APU) are taken directly from The NES dev wiki.

rs-nes's Issues

Implement background scrolling

Latch used to determine address state is shared between ppu scroll and address register

Currently it's implemented as both registers having their own latch. Refactor so that they share the latch to emulate the hardware more accurately.

Fix code coverage results

Tracking issue: huonw/travis-cargo#11

Write NesMemory tests

In particular, memory mapped reads/writes. Will need to introduce traits for PPU, APU, and Input so that they can be mocked.

Vulkan renderer segfaults on close

Something I introduced in a4307ad caused the vulkan renderer to segfault while the program is exiting.

Create separate Absolute/ZeroPage addressing mode implementations for store instructions

Current implementations are ugly and hacky because they have to accommodate store instruction special needs

Memory trait: Rename load/store to read write

Capture crash errors in debugger and report them

Implement setting sprite overflow flag

Make Ppu.step() return result or make Memory.tick() not return result

If Ppu.step() could trigger some illegal state then we will want to make it return a result so the crash can be reported to the debugger. If not, then there is no reason for Memory.tick() to return a result.

OAM Dma timing

OAM DMA write suspends the cpu for 513 CPU cycles (+1 on odd CPU cycles), during which I assume other components should continue to tick.

Make cpu.nmi() private

Implement correct pixel multiplexing

Redo terrible pixel abstraction

The current pixel abstraction for bg and sprite pixels is bad. Make it better.

Debugger can't accommodate self-modifying code or bank-switched PRG

Disassembly is done once and results are cached for the life of the debugger. To accommodate self-modifying code/bank-switched PRG, we would need to hash PRG address space and re-disassemble when hash changes.