rsagg

RuSt Algorand GPU vanity address Generator

Usage

Optimize batch size

cargo run --release optimize

Note down the best performance batch value for use in the generator.

Run the generator

cargo run --release generate --batch BATCH_SIZE PREFIX

Passing extra options to the OpenCL compiler

Arbitrary options can be appended verbatim to the clBuildProgram call with the --cl-build-options flag. optimize persists the value into bacon.json so generate reuses it.

cargo run --release generate --cl-build-options "-cl-nv-maxrregcount=96" --batch BATCH_SIZE PREFIX
# tune & persist into bacon.json
cargo run --release optimize --cl-build-options "-cl-nv-maxrregcount=96" PREFIX

Note (NVIDIA): this kernel is register-heavy and occupancy-bound, so capping registers-per-thread with -cl-nv-maxrregcount=<N> can boost throughput a lot (~+30% on a Maxwell GTX 950 at N=96). The optimum is hardware-specific with a sharp cliff, so tune per GPU (try 80/88/96/104). NVIDIA-only — do not pass it on other vendors.

Provider-specific instructions

vast.ai

1. Go to the vast.ai template: https://cloud.vast.ai/?template_id=85b0923f64acdb8521825c54e369fac3
2. Select a machine of your choice and click "Rent"
3. Connect to the machine
4. Run ./bacon optimize / generate as described above

Security notes

The utility hasn't been audited so please make sure to rekey any accounts for the addresses generated with it.