HPC Backend

The HPC module provides access to high-performance computing clusters for quantum simulation workloads that exceed the capacity of local or cloud simulators. Jobs are dispatched to multi-node CPU/GPU clusters, enabling simulation of circuits up to ~40 qubits with full statevector and matrix product state (MPS) methods.

HPC vs. Cloud Simulator vs. QPU

Feature	Cloud Simulator	HPC	QPU
Max qubits (exact)	~30	~40	50–433 (noisy)
Noise model	Optional	Optional	Real hardware noise
Throughput	Moderate	Very high	Low (queue)
Cost	Low	Medium	High
Latency	Seconds	Minutes	Minutes–hours
Best for	Development	Large simulations	Real quantum experiments

Available HPC Providers

GET /api/v1/hpc/providers

{
  "providers": [
    {
      "id": "azure-hpc-eastus",
      "name": "Azure HPC — East US",
      "max_qubits": 40,
      "simulators": ["statevector", "mps", "extended_stabilizer"],
      "gpu_accelerated": true,
      "nodes": 64
    },
    {
      "id": "lumi-csc-fi",
      "name": "LUMI Supercomputer (CSC, Finland)",
      "max_qubits": 38,
      "simulators": ["statevector", "mps"],
      "gpu_accelerated": true,
      "nodes": 512
    },
    {
      "id": "aws-hpc-us-east-1",
      "name": "AWS HPC — US East",
      "max_qubits": 36,
      "simulators": ["statevector", "density_matrix"],
      "gpu_accelerated": false,
      "nodes": 32
    }
  ]
}

Submit an HPC Job

POST /api/v1/hpc/jobs

{
  "job_type": "execute",
  "payload": {
    "circuit": "OPENQASM 2.0; ...",
    "shots": 8192,
    "simulator": "mps"
  },
  "resources": {
    "nodes": 8,
    "gpus": 0,
    "cpus": 32,
    "mem_gb": 128,
    "walltime": "01:00:00",
    "partition": "standard-g"
  },
  "priority": "normal"
}

Field	Description
`job_type`	`hpc`, `execute`, or `algorithm`
`script` / `command`	Raw SLURM batch script, or a shell command using the default template
`payload`	Structured payload for QCOS job types (circuit, shots, simulator)
`resources`	SLURM resources: `nodes` (1–512), `gpus` per node, `cpus`, `mem_gb`, `walltime`, `partition`
`priority`	`low`, `normal`, or `high`

Response

{
  "job_id": "hpc_a1b2c3d4",
  "status": "queued",
  "provider": "lumi",
  "region": "fi-01-lumi",
  "submitted_at": "2025-01-15T10:12:00Z",
  "estimated_wait_s": 60
}

Check Job Status

GET /api/v1/hpc/jobs/{job_id}

curl -H "Authorization: Bearer $QCOS_API_KEY" \
  https://api.softquantus.com/api/v1/hpc/jobs/hpc_a1b2c3d4

{
  "job_id": "hpc_a1b2c3d4",
  "status": "running",
  "provider": "lumi",
  "slurm_job_id": "8841723",
  "submitted_at": "2025-01-15T10:12:00Z",
  "started_at": "2025-01-15T10:14:00Z"
}

Retrieve HPC Results

Results arrive on the same status endpoint once the job completes — the result field carries the structured output, alongside stdout, stderr, and exit_code:

GET /api/v1/hpc/jobs/{job_id}

{
  "job_id": "hpc_a1b2c3d4",
  "status": "completed",
  "provider": "lumi",
  "completed_at": "2025-01-15T10:22:18Z",
  "exit_code": 0,
  "result": {
    "counts": {"0000...": 1024, "1111...": 1024},
    "shots": 8192
  }
}

Cancel a queued or running job with DELETE /api/v1/hpc/jobs/{job_id}.

Python Example — 36-Qubit GHZ State

import os
import time
import requests

API_KEY = os.environ["QCOS_API_KEY"]
BASE_URL = "https://api.softquantus.com"
HEADERS = {"Authorization": f"Bearer {API_KEY}", "Content-Type": "application/json"}

N_QUBITS = 36

# Build large GHZ circuit
gates = [f"h q[0];"] + [f"cx q[{i}],q[{i+1}];" for i in range(N_QUBITS - 1)]
measurements = " ".join(f"measure q[{i}] -> c[{i}];" for i in range(N_QUBITS))
qasm = (
    f'OPENQASM 2.0; include "qelib1.inc"; '
    f'qreg q[{N_QUBITS}]; creg c[{N_QUBITS}]; '
    + " ".join(gates) + " " + measurements
)

# Submit HPC job
resp = requests.post(
    f"{BASE_URL}/api/v1/hpc/jobs",
    headers=HEADERS,
    json={
        "job_type": "execute",
        "payload": {"circuit": qasm, "shots": 4096, "simulator": "mps"},
        "resources": {"nodes": 4, "cpus": 32, "mem_gb": 128, "walltime": "01:00:00"}
    }
)
resp.raise_for_status()
job = resp.json()
print(f"HPC job submitted: {job['job_id']}")
print(f"Estimated wait: {job['estimated_wait_s']} s")

# Poll — the completed status carries the result
while True:
    s = requests.get(f"{BASE_URL}/api/v1/hpc/jobs/{job['job_id']}", headers=HEADERS).json()
    print(f"  Status: {s['status']}")
    if s["status"] in ("completed", "failed", "cancelled"):
        break
    time.sleep(30)

counts = (s.get("result") or {}).get("counts", {})
top = sorted(counts.items(), key=lambda x: -x[1])[:5]
for state, count in top:
    print(f"  |{state}⟩  {count}")

Queue Management

GET /api/v1/hpc/queue

Returns your organisation’s current HPC queue with estimated wait times per provider. Use this before submitting to select the provider with the shortest wait.

Pricing

HPC execution is billed per node-minute actually consumed — idle time during queue wait is not charged. QCOS reports technical metering only; prices are set and billed by the SoftQuantus Platform, and the current per-provider rates are in the console’s pricing catalog.

Backends Overview
Batch Execution — For many smaller circuits
Bench — Benchmark HPC vs. QPU performance

Core

Modules

Certification

HPC Backend

HPC Backend

HPC vs. Cloud Simulator vs. QPU

Available HPC Providers

Submit an HPC Job

Response

Check Job Status

Retrieve HPC Results

Python Example — 36-Qubit GHZ State

Queue Management

Pricing

​HPC Backend

​HPC vs. Cloud Simulator vs. QPU

​Available HPC Providers

​Submit an HPC Job

​Response

​Check Job Status

​Retrieve HPC Results

​Python Example — 36-Qubit GHZ State

​Queue Management

​Pricing

​Related

HPC Backend

HPC vs. Cloud Simulator vs. QPU

Available HPC Providers

Submit an HPC Job

Response

Check Job Status

Retrieve HPC Results

Python Example — 36-Qubit GHZ State

Queue Management

Pricing

Related