Add a 'strong' validation for qubit allocation batching to match the …

…assumption of the underlying simulator backends (#1797)

runtime/cudaq/qis/managers/default/DefaultExecutionManager.cpp

@@ -56,16 +56,32 @@ class DefaultExecutionManager : public cudaq::BasicExecutionManager {
  const void *state,
  cudaq::simulation_precision precision) override {
  // Here we have qubits in requestedAllocations
- // want to allocate and set state
+ // want to allocate and set state.
  // There could be previous 'default' allocations whereby we just cached them
  // in requestedAllocations.
  // These default allocations need to be dispatched separately.
- // FIXME: this assumes no qubit reuse, aka the qubits in targets are the
- // last ones to be allocated. This is consistent with the Kronecker product
- // assumption in CircuitSimulator.
  if (!requestedAllocations.empty() &&
  targets.size() != requestedAllocations.size()) {
  assert(targets.size() < requestedAllocations.size());
+ // This assumes no qubit reuse, aka the qubits are allocated in order.
+ // This is consistent with the Kronecker product assumption in
+ // CircuitSimulator.
+ for (std::size_t i = 0; i < requestedAllocations.size() - 1; ++i) {
+ // Verify this assumption to make sure the simulator set
+ // the state of appropriate qubits.
+ const auto &thisAlloc = requestedAllocations[i];
+ const auto &nextAlloc = requestedAllocations[i + 1];
+ if (nextAlloc.id != (thisAlloc.id + 1)) {
+ std::stringstream errorMsg;
+ errorMsg << "Out of order allocation detected. This is not supported "
+ "by simulator backends. Qubit allocations: [ ";
+ for (const auto &alloc : requestedAllocations) {
+ errorMsg << alloc.id << " ";
+ }
+ errorMsg << "]";
+ throw std::logic_error(errorMsg.str());
+ }
+ }
  const auto numDefaultAllocs =
  requestedAllocations.size() - targets.size();
  simulator()->allocateQubits(numDefaultAllocs);