dqmc-dev
Workflow for modifying DQMC C code, adding new measurements, and running tests. Use when editing source code, implementing new observables,
Open source Open in the app JSON README (API)
About
Workflow for modifying DQMC C code, adding new measurements, and running tests. Use when editing source code, implementing new observables, debugging, or validating code changes.
Details
- Kind
- Agent skills
- Topic
- Developer tools
- Publisher
- edwnh
- Origin
- majiayu
- Category
- ferramentas
- Stars
- 18
- Forks
- 4
- Last push
- 2026-04-07T22:10:41Z
- Repository state
- ativo
- Language
- C
- License
- MIT
- Added
- 2026-09-02 18:14:51
- Updated
- 2026-09-02 18:14:51
- Origin id
edwnh/dqmc/development/dqmc-dev@main
README
# Determinantal quantum Monte Carlo for the Hubbard model High-performance DQMC (C) for the Hubbard model, plus Python utilities (`dqmc_util`) for generating inputs and analyzing outputs. ## Supported platforms - Linux and macOS - Windows is not supported natively; use WSL2 or GitHub Codespaces if needed. ## Dependencies **Build (Linux):** - Intel compiler `icx` - Intel MKL headers and libraries - HDF5 headers and libraries (downloaded via `make deps`) **Build (macOS):** - `clang` (Xcode Command Line Tools) - HDF5 headers and libraries (downloaded via `make deps`) **Python (`dqmc_util`):** - Installed via `pip install -e .` (included in `environment-*.yml`) - Runtime deps: `numpy`, `scipy`, `h5py` ## Install and build 1. Create the conda environment: ```bash # macOS conda env create -f environment-macos.yml # Linux conda env create -f environment-linux.yml ``` 2. Activate it: ```bash conda activate dqmc ``` 3. Download/extract the bundled HDF5 dependency (first time only): ```bash make deps ``` 4. Build: ```bash make ``` The binary is `build/dqmc`. ## Quickstart (smoke run) ```bash conda activate dqmc make deps # first time only make dqmc-util gen Nx=4 Ny=4 U=4 dt=0.1 L=20 n_sweep_warm=50 n_sweep_meas=100 build/dqmc sim_0.h5 dqmc-util summary sim_0.h5 ``` ## Workflow 1. Generate HDF5 simulation files: `dqmc-util gen` (alias: `dqmc-util gen-1band-hub`) 2. Run Monte Carlo sweeps: `build/dqmc sim_0.h5` (optionally with checkpointing) 3. Analyze results: `dqmc-util summary`, `dqmc-util print-n`, or `dqmc_util.analyze_hub` in Python ### Run one file ```bash build/dqmc sim_0.h5 dqmc-util summary sim_0.h5 ``` ### Run many files (sharded queue) For 2+ files, prefer the queue/worker system (`dqmc-util enqueue` + `dqmc-util worker`). ```bash dqmc-util gen Nx=6 Ny=6 U=4 mu=0 dt=0.1 L=50 Nfiles=160 prefix=runs/U4_T0.25/bin dqmc-util enqueue queue 'runs/U4_T0.25/bin_*.h5' dqmc-util worker queue ./build/dqmc -n 8 -s 300 -t 3600 dqmc-util queue-status queue dqmc-util print-n runs/U4_T0.25/ ``` ## Examples See `examples/README.md` and: - `examples/mz2_vs_T/` - magnetic moment vs temperature - `examples/n_vs_mu/` - density vs chemical potential ## HDF5 layout One Markov chain = one HDF5 file = one “bin”. - `/metadata`: model parameters for analysis (e.g., `beta`, `mu`, `Nx`, `Ny`) - `/params`: simulation parameters and matrices (often stored in a shared `*.h5.params`) - `/state`: sweep counter, RNG state, auxiliary field configuration - `/meas_eqlt`: equal-time measurements - `/meas_uneqlt`: unequal-time measurements (optional) For details and authoritative defaults, see `dqmc_util/gen_1band_hub.py` (`SimParams`). ## Parameter notes - Trotter error rule of thumb: ensure `U * dt^2 <= 0.05` - `L` must be divisible by `n_matmul` and `period_eqlt` ## AI agent docs - `AGENTS.md`: working on this codebase (build, architecture, guardrails) - `.claude/skills/`: task runbooks (generate, run, analyze, scans, dev, advanced)