Remove SVDynamic (#999)

**Context:** When the LQ plugin was migrated from Catalyst to Lightning,
the `StateVectorLQDynamic` class was also migrated. Given the fact that
we have `StateVectorLQManaged`, we can get rid of
`StateVectorLQDynamic`.

**Description of the Change:** Replace the uses of
`StateVectorLQDynamic` with `StateVectorLQManaged`. I basically reused
the code from the LK simulator and removed the calls to the
`StateVectorLQDynamic` related methods. Of course, I had to use
`std::complex` instead of `kokkos::complex`.

**Benefits:** Using the same underlying class for LG, LK and LQ.

**NOTE:** I have deleted the only two tests that were failing:

1. **Qubit allocatation and deallocation**
(https://github.com/PennyLaneAI/pennylane-lightning/pull/999/files#diff-4782c8d1aa88463716419ff1ef4f9f551d6241e43f5bc107d27ff0ed4299a3ebL78):
the second call to `State(...)` showed a size of 2 for the given state
and a size of 4 for the device state, which made the assertion fail.
2. **QuantumDevice object test**
(https://github.com/PennyLaneAI/pennylane-lightning/pull/999/files#diff-4782c8d1aa88463716419ff1ef4f9f551d6241e43f5bc107d27ff0ed4299a3ebL193)
The test was killed after 1 minute of running, probably because an
infinite loop/recursion?

It seems to me these two tests do not comply with the way
`StateVectorLQManaged` is structured but let the discussion resolve it.

[sc-77689]

---------

Co-authored-by: ringo-but-quantum <[email protected]>
Co-authored-by: Ali Asadi <[email protected]>

.github/CHANGELOG.md

@@ -5,8 +5,9 @@
 * Add native N-controlled generators and adjoint support to `lightning.gpu`'s single-GPU backend.
   [(#970)](https://github.com/PennyLaneAI/pennylane-lightning/pull/970)
 
-* Add a Catalyst-specific wrapping class for Lightning Qubit.
+* Add a Catalyst-specific wrapping class for Lightning-Qubit.
   [(#960)](https://github.com/PennyLaneAI/pennylane-lightning/pull/960)
+  [(#999)](https://github.com/PennyLaneAI/pennylane-lightning/pull/999)
 
 * Add native N-controlled gates support to `lightning.gpu`'s single-GPU backend.
   [(#938)](https://github.com/PennyLaneAI/pennylane-lightning/pull/938)

pennylane_lightning/core/_version.py

@@ -16,4 +16,4 @@
    Version number (major.minor.patch[-label])
 """
 
-__version__ = "0.40.0-dev11"
+__version__ = "0.40.0-dev12"

pennylane_lightning/core/src/simulators/lightning_qubit/catalyst/LightningObsManager.hpp

@@ -24,10 +24,10 @@
 #include "Utils.hpp"
 
 #include "ObservablesLQubit.hpp"
-#include "StateVectorLQubitDynamic.hpp"
+#include "StateVectorLQubitManaged.hpp"
 
 namespace {
-using Pennylane::LightningQubit::StateVectorLQubitDynamic;
+using Pennylane::LightningQubit::StateVectorLQubitManaged;
 using Pennylane::LightningQubit::Observables::HermitianObs;
 using Pennylane::LightningQubit::Observables::NamedObs;
 using Pennylane::LightningQubit::Observables::TensorProdObs;
@@ -43,7 +43,7 @@ namespace Catalyst::Runtime::Simulator {
  */
 template <typename PrecisionT> class LightningObsManager {
   private:
-    using VectorStateT = StateVectorLQubitDynamic<PrecisionT>;
+    using VectorStateT = StateVectorLQubitManaged<PrecisionT>;
     using ComplexT = typename VectorStateT::ComplexT;
     using ObservablePairType =
         std::pair<std::shared_ptr<Observable<VectorStateT>>, ObsType>;

pennylane_lightning/core/src/simulators/lightning_qubit/catalyst/LightningSimulator.cpp

@@ -22,8 +22,29 @@
 namespace Catalyst::Runtime::Simulator {
 
 auto LightningSimulator::AllocateQubit() -> QubitIdType {
-    size_t sv_id = this->device_sv->allocateWire();
-    return this->qubit_manager.Allocate(sv_id);
+    const std::size_t num_qubits = this->device_sv->getNumQubits();
+
+    if (num_qubits == 0U) {
+        this->device_sv = std::make_unique<StateVectorT>(1);
+        return this->qubit_manager.Allocate(num_qubits);
+    }
+
+    std::vector<std::complex<double>, AlignedAllocator<std::complex<double>>>
+        data = this->device_sv->getDataVector();
+    const std::size_t dsize = data.size();
+    data.resize(dsize << 1UL);
+
+    auto src = data.begin();
+    std::advance(src, dsize - 1);
+
+    for (auto dst = data.end() - 2; src != data.begin();
+         std::advance(src, -1), std::advance(dst, -2)) {
+        *dst = *src;
+        *src = std::complex<double>(.0, .0);
+    }
+
+    this->device_sv = std::make_unique<StateVectorT>(data);
+    return this->qubit_manager.Allocate(num_qubits);
 }
 
 auto LightningSimulator::AllocateQubits(size_t num_qubits)
@@ -45,14 +66,11 @@ auto LightningSimulator::AllocateQubits(size_t num_qubits)
 }
 
 void LightningSimulator::ReleaseAllQubits() {
-    this->device_sv->clearData();
     this->qubit_manager.ReleaseAll();
+    this->device_sv = std::make_unique<StateVectorT>(0); // reset the device
 }
 
 void LightningSimulator::ReleaseQubit(QubitIdType q) {
-    if (this->qubit_manager.isValidQubitId(q)) {
-        this->device_sv->releaseWire(this->qubit_manager.getDeviceId(q));
-    }
     this->qubit_manager.Release(q);
 }
 

pennylane_lightning/core/src/simulators/lightning_qubit/catalyst/LightningSimulator.hpp

@@ -22,7 +22,7 @@
 #include <numeric>
 #include <span>
 
-#include "StateVectorLQubitDynamic.hpp"
+#include "StateVectorLQubitManaged.hpp"
 
 #include "CacheManager.hpp"
 #include "Exception.hpp"
@@ -35,7 +35,7 @@ namespace Catalyst::Runtime::Simulator {
 class LightningSimulator final : public Catalyst::Runtime::QuantumDevice {
   private:
     using StateVectorT =
-        Pennylane::LightningQubit::StateVectorLQubitDynamic<double>;
+        Pennylane::LightningQubit::StateVectorLQubitManaged<double>;
 
     // static constants for RESULT values
     static constexpr bool GLOBAL_RESULT_TRUE_CONST = true;