qibo-final/qibojit-benchmarks/scripts/quimb_mpi

Quimb MPI Benchmark Suite

Demonstrates MPI-enabled tensor network simulation using quimb with 2 MPI processes.

Files

• quimb_mpi.sh — Main benchmark sweep script using mpirun -np 2
• expectation_mps_mpi.json — Expectation mode with MPS (default for expectation)
• expectation_dense_mpi.json — Expectation mode, dense tensor network
• dense_vector_mps_mpi.json — State extraction with MPS (default for state)
• dense_vector_dense_mpi.json — State extraction, dense tensor network

Key Differences from /scripts/quimb/

• Each python compare.py call wrapped with mpirun -np 2 to enable MPI parallelization
• All JSON configs include "MPI_enabled": true
• Default file names include _mpi suffix to distinguish from single-process configs
• Filenames in logs clearly show MPI mode (e.g., expectation_mps_mpi.json)

Configuration Files

Expectation Modes (computing Hamiltonian expectation values)

• expectation_mps_mpi.json: MPI-enabled, uses Matrix Product State (MPS) tensor network ansatz; good for larger systems with limited bond dimension
• expectation_dense_mpi.json: MPI-enabled, uses dense tensor network; full contraction without MPS truncation

State Modes (full state vector computation)

• dense_vector_mps_mpi.json: MPI-enabled, extracts full state using MPS contraction (faster, bounded memory)
• dense_vector_dense_mpi.json: MPI-enabled, extracts full state using dense tensor (maximum memory usage)

Usage

cd /ssd_data/tankya2/code/ASC_2026_prism/ASC-2026/qibojit-benchmarks/scripts/quimb_mpi

# Run with defaults (all matching dense_vector*.json and expectation*.json configs)
./quimb_mpi.sh

# Test only expectation mode, custom qubit sizes
modes="expectation" nqubits_list="8 10 12" ./quimb_mpi.sh

# Test dense vector only with one explicit config
state_cfg=dense_vector_dense_mpi.json modes="dense_vector" ./quimb_mpi.sh

# Restrict the sweep to a chosen subset of configs
EXPECTATION_CONFIGS="expectation_mps_mpi.json expectation_dense_mpi.json" ./quimb_mpi.sh

# Use 4 MPI processes (if testing scalability)
np=4 ./quimb_mpi.sh

Supported Circuits

✓ Fully Supported (Numerically Validated)

Circuit	Alias	Parameters	Status	Notes
qft	—	none	✓ Works	Quantum Fourier Transform; aligned with qibojit
variational	—	nlayers	✓ Works	Parameterized circuit; aligned with qibojit
bv	bernstein-vazirani	none	✓ Works	Bernstein-Vazirani algorithm; aligned with qibojit
hs	hidden-shift	none	✓ Works	Hidden-shift problem; aligned with qibojit
qaoa	—	nlayers	✓ Works	QAOA with RZZ gate support; aligned with qibojit

❌ Not Supported

Circuit	Issue	Notes
qv	Qiskit API compatibility	quantum-volume uses deprecated qiskit .qasm() method (v3.0+); not specific to quimb

Environment Variables

Variable	Default	Description
nqubits_list	6 8 10	Space-separated qubit counts to benchmark
circuits	qaoa	Space-separated circuit names
modes	dense_vector expectation	Benchmark modes (dense_vector, expectation)
nlayers	2	Layers for layered circuits (variational, qaoa)
np	2	Number of MPI processes (mpirun -np)
nreps	1	Repetitions per benchmark
filename	quimb_benchmark_mpi.dat	Output log file
precision	complex128	NumPy dtype (complex128 or complex64)
exp_cfg	unset	Optional single expectation config; when unset, all expectation*.json files are swept
state_cfg	unset	Optional single dense-vector config; when unset, all dense_vector*.json files are swept
EXPECTATION_CONFIGS	unset	Optional space-separated subset of expectation configs to run
STATE_CONFIGS	unset	Optional space-separated subset of dense-vector configs to run

Results

Results are stored in /qibojit-benchmarks/benchmarks/results/ with filename prefix quimb_benchmark_mpi.dat by default (configurable via filename= env var).

Each benchmark includes:

• import_time — Backend initialization time
• creation_time — Circuit creation time
• dry_run_time — First execution time (warm-up)
• simulation_times — Execution times for nreps runs
• expectation_result — Observable expectation values (for expectation mode)

Observable

By default, expectation uses:

• X on qubit 0, identity on remaining qubits
• Configurable via pauli_pattern in JSON config (e.g., "XIIII" for 5 qubits)

Reference Validation

Each benchmark run automatically includes a qibojit (numba) reference with the same observable configuration, enabling direct numerical comparison for validation.

Example:

[expectation MPI]   quimb + variational @ 6q:    0.2019...
[qibojit ref]       numba + variational @ 6q:    0.2019...  ✓ aligned

Supported Circuits

✓ Fully Supported (Numerically Validated)

Circuit	Alias	Parameters	Status	Notes
qft	—	none	✓ Works	Quantum Fourier Transform; aligned with qibojit
variational	—	nlayers	✓ Works	Parameterized circuit; aligned with qibojit
bv	bernstein-vazirani	none	✓ Works	Bernstein-Vazirani algorithm; aligned with qibojit
hs	hidden-shift	none	✓ Works	Hidden-shift problem; aligned with qibojit
qaoa	—	nlayers	✓ Works	QAOA with RZZ gate support; aligned with qibojit

❌ Not Supported

Circuit	Issue	Notes
qv	Qiskit API compatibility	quantum-volume uses deprecated qiskit .qasm() method (v3.0+); not specific to quimb

Notes

• Intel MPI Fabric: On Prism, automatically sets I_MPI_FABRICS=shm (shared-memory fabric) for single-node local runs
• Observable: X on first qubit, identity on rest (configurable via pauli_pattern in JSON)
• Reference: qibojit (numba backend) included automatically with aligned observable for validation
• RZZ Gate: Full support for QAOA and algorithms using two-qubit Ising rotations
• cu1 Gate: Fully supported in expectation value mode (QFT compatible)
• Numerical Alignment: All benchmarks automatically include qibojit reference with identical observable for validation

MPI Enable/Disable

To test without MPI, use configs from /scripts/quimb/ instead:

# Compare: MPI vs no MPI
./quimb_mpi.sh                    # with MPI (2 processes)
cd ../quimb && ./quimb_local.sh   # without MPI (single process)

Tensor Network Options

Quimb supports both MPS (Matrix Product State) and dense tensor network contractions. By default the script sweeps every matching JSON file in this folder, or you can target individual configs:

# Compare MPS vs dense at same nqubits with MPI
exp_cfg=expectation_mps_mpi.json nqubits_list="12" ./quimb_mpi.sh
exp_cfg=expectation_dense_mpi.json nqubits_list="12" ./quimb_mpi.sh

MPS is typically faster and uses bounded memory, while dense provides full accuracy without truncation.

Additional Notes

• Bond Dimension: MPS modes use max_bond_dimension=20 by default; adjust in JSON for memory/accuracy tradeoff
• SVD Cutoff: Set to 1e-10 for high accuracy; increase for faster execution with acceptable loss tolerance
• Scaling: MPI parallelization enables larger systems; test with np=4, np=8, etc. for scaling studies
• Single-Node MPI: Default -np 2 demonstrates MPI capability on local machine; for cluster runs, adjust -np and node configuration