Benchmark Results
Single Thread Benchmark
This benchmark only include single thread benchmark, all the multithread features are disabled.
Machine Info
Julia & CPU Info
julia> versioninfo()
Julia Version 1.3.0
Commit 46ce4d7933 (2019-11-26 06:09 UTC)
Platform Info:
OS: Linux (x86_64-pc-linux-gnu)
CPU: Intel(R) Xeon(R) Gold 6230 CPU @ 2.10GHz
WORD_SIZE: 64
LIBM: libopenlibm
LLVM: libLLVM-6.0.1 (ORCJIT, skylake)
BLAS: intel MKL
Python version: 3.7.3
GPU: Tesla V100
Package Info
| Package | Version | Type of Simulator |
|---|---|---|
| Yao | v0.6.1 | full amplitudes |
| CuYao | v0.2.0 | full amplitudes |
| qiskit | 0.16.0 | full amplitudes |
| qiskit-aer | 0.4.0 | full amplitudes |
| qiskit-terra | 0.12.0 | full amplitudes |
| qulacs | 0.1.9 | full amplitudes |
| projectq | 0.4.2 | full amplitudes |
| Cirq | 0.7.0 | full amplitudes |
| PennyLane | 0.8.1 | full amplitudes |
| QuEST (pyquest-cffi) | 0.1.1 | full amplitudes |
| JKQ DDSIM¹ | v1.1 | decision diagrams |
¹ This benchmark uses the mean estimator for the timings. To get accurate timings when recreating the results, please ensure no other other applications run concurrently.
Single Gate Benchmark
Benchmarks of a) Pauli-X gate; b) Hadamard gate; c) CNOT gate; d) Toffolli gate.
Parameterized Circuit Benchmark
b) Benchmarks of parameterized circuit. c) Benchmarks of parametrized circuit with batched registers (batch size = 1000).
NOTE:
- qiskit state vector simulator does not support rotation x/z gate, thus there is no benchmark on the following circuits.
- PennyLane benchmark contains some overhead from error handling since we do not include measurement in this benchmark (#7)
- the performance of CUDA may vary on different machine (#6), although the difference is not very huge
