BitBOINC

🌊 Potamos -- what it is and how it works 🌊

Potamos (Greek: "river" -- steady, systematic, reaches everything downstream) is BitBOINC's sequential range scan search 🔎. It's the simpler of our two search strategies, but it's the ONLY one that can work a puzzle whose public key hasn't been exposed on-chain yet.

⚙️ How it actually searches

Each Bitcoin puzzle address hides a private key inside a known numeric range (e.g. puzzle #71's key is somewhere in a ~2^71-wide range). Potamos walks that range forward, chunk by chunk:

📦 The full range is split into large fixed-size chunks -- 2^32 keys per CPU chunk, 2^38 keys per GPU chunk (GPUs are so much faster per-task that a CPU-sized chunk would be pure scheduling overhead for them).
🎟️ Every host pulls the NEXT unclaimed chunk from one shared counter for that puzzle -- CPU and GPU, every platform, all drawing from the same counter, so nobody ever scans a range someone else already covered.
🧮 For every candidate private key in its chunk, the app computes the public key, hashes it (SHA-256 + RIPEMD-160), and compares against the puzzle's target address -- using Jacobian coordinates + batched modular inversion so consecutive keys are cheap.
✅ If nothing hits, the chunk is done and the host reports back; if something hits, that's the key. 🔑

⭐ Why we need it at all

Our OTHER search method, Keraunos (kangaroo), is mathematically far more efficient -- but it fundamentally requires the puzzle's public key to already be visible on the blockchain. A lot of puzzles have never had their pubkey exposed (that only happens once someone spends FROM the address). Potamos doesn't care -- it only needs the address, so it's the only one of our two apps that can touch those puzzles at all. Right now Potamos is working puzzle #71 for exactly that reason. 🧭

🖥️ Platforms supported

CPU:
🪟 Windows -- x86, x64, ARM64
🐧 Linux -- x86_64, i686, aarch64, armhf, ppc64le, riscv64
🍎 macOS -- Apple Silicon
🤖 Android -- aarch64

GPU-accelerated 🚀 (Windows/Linux/macOS):
🟩 NVIDIA -- CUDA, Kepler/Volta-class and Turing-or-newer (separately tuned)
🟥 AMD -- OpenCL
🟦 Intel -- OpenCL
⬛ Apple GPUs -- OpenCL

🛠️ Coming soon: Moore Threads MUSA GPU support -- the port is in progress, not yet built/released.

💰 A recent fix worth mentioning

Potamos's declared "how much work is this chunk" estimate had been set artificially low as a side effect of an unrelated scheduling tweak, which meant every completed Potamos chunk was earning noticeably less credit than the real work performed. That's fixed now -- new chunks are credited fairly. (Nothing retroactive -- work already completed keeps whatever credit it already got.)

🐛 Found a bug?

Reply below -- crashes, stuck/stalled tasks, wrong chunk sizes for your hardware, credit that still looks off, whatever. Include your OS/GPU and, if you can, the task's stderr output. 🙏

---