* Added destructor for PyKernelDecorator which in turn calss the

  destructor for CapturedDataStorage
* Renamed duplicate kernels - need to investigate if this points to an
  underlying bug. Without this, when running the full test suite, we get
  error like -
  `RuntimeError: python alt_launch_kernel failed to get set cudaq state
function for kernel: kernel`
  All tests run successfully standalone.

Signed-off-by: Pradnya Khalate <[email protected]>

Author: khalatepradnya

python/cudaq/kernel/kernel_decorator.py

@@ -524,6 +524,11 @@ def __call__(self, *args):
             # self.capturedDataStorage = None
             return result
 
+    def __del__(self):
+        if self.capturedDataStorage != None:
+            self.capturedDataStorage.__del__()
+        self.capturedDataStorage = None
+
 
 def kernel(function=None, **kwargs):
     """

python/tests/builder/test_qalloc_init_state.py

@@ -166,7 +166,6 @@ def test_kernel_complex_capture_f64():
     def kernel():
         q = cudaq.qvector(state)
 
-    # kernel.compile()
     counts = cudaq.sample(kernel)
     print(counts)
     assert '11' in counts
@@ -215,11 +214,10 @@ def test_kernel_complex64_capture_f32():
     state = cudaq.State.from_data(c)
 
     @cudaq.kernel
-    def kernel():
+    def kernel_foo():
         q = cudaq.qvector(state)
 
-    # kernel.compile()
-    counts = cudaq.sample(kernel)
+    counts = cudaq.sample(kernel_foo)
     print(counts)
     assert '11' in counts
     assert '00' in counts

python/tests/kernel/test_kernel_qvector_state_init.py

@@ -18,11 +18,19 @@
     not (cudaq.num_available_gpus() > 0 and cudaq.has_target('nvidia')),
     reason='Could not find nvidia in installation')
 
+
+@pytest.fixture(autouse=True)
+def do_something():
+    cudaq.reset_target()
+    yield
+    cudaq.reset_target()
+    cudaq.__clearKernelRegistries()
+
+
 # synthesis
 
 
 def test_kernel_synthesis_complex():
-    cudaq.reset_target()
 
     c = np.array([1. / np.sqrt(2.) + 0j, 1. / np.sqrt(2.), 0., 0.],
                  dtype=cudaq.complex())
@@ -50,7 +58,6 @@ def kernel(vec: cudaq.State):
 @skipIfNvidiaFP64NotInstalled
 def test_kernel_float_params_f64():
 
-    cudaq.reset_target()
     cudaq.set_target('nvidia', option='fp64')
 
     f = np.array([1. / np.sqrt(2.), 0., 0., 1. / np.sqrt(2.)], dtype=float)
@@ -61,10 +68,9 @@ def test_kernel_float_params_f64():
         e)
 
 
-@skipIfNvidiaFP64NotInstalled
+@skipIfNvidiaNotInstalled
 def test_kernel_float_params_f32():
 
-    cudaq.reset_target()
     cudaq.set_target('nvidia')
 
     f = np.array([1. / np.sqrt(2.), 0., 0., 1. / np.sqrt(2.)], dtype=np.float32)
@@ -80,7 +86,7 @@ def test_kernel_float_params_f32():
 
 @skipIfNvidiaFP64NotInstalled
 def test_kernel_complex_params_f64():
-    cudaq.reset_target()
+
     cudaq.set_target('nvidia', option='fp64')
 
     c = np.array([1. / np.sqrt(2.) + 0j, 0., 0., 1. / np.sqrt(2.)],
@@ -99,7 +105,7 @@ def kernel(vec: cudaq.State):
 
 @skipIfNvidiaFP64NotInstalled
 def test_kernel_complex128_params_f64():
-    cudaq.reset_target()
+
     cudaq.set_target('nvidia', option='fp64')
 
     c = np.array([1. / np.sqrt(2.) + 0j, 0., 0., 1. / np.sqrt(2.)],
@@ -118,7 +124,7 @@ def kernel(vec: cudaq.State):
 
 @skipIfNvidiaFP64NotInstalled
 def test_kernel_complex64_params_f64():
-    cudaq.reset_target()
+
     cudaq.set_target('nvidia', option='fp64')
 
     c = np.array([1. / np.sqrt(2.) + 0j, 0., 0., 1. / np.sqrt(2.)],
@@ -131,7 +137,7 @@ def test_kernel_complex64_params_f64():
 
 @skipIfNvidiaNotInstalled
 def test_kernel_complex64_params_f32():
-    cudaq.reset_target()
+
     cudaq.set_target('nvidia')
 
     c = np.array([1. / np.sqrt(2.) + 0j, 0., 0., 1. / np.sqrt(2.)],
@@ -150,7 +156,7 @@ def kernel(vec: cudaq.State):
 
 @skipIfNvidiaNotInstalled
 def test_kernel_complex128_params_f32():
-    cudaq.reset_target()
+
     cudaq.set_target('nvidia')
 
     c = np.array([1. / np.sqrt(2.) + 0j, 0., 0., 1. / np.sqrt(2.)],
@@ -163,7 +169,7 @@ def test_kernel_complex128_params_f32():
 
 @skipIfNvidiaNotInstalled
 def test_kernel_complex_params_f32():
-    cudaq.reset_target()
+
     cudaq.set_target('nvidia')
 
     c = np.array([1. / np.sqrt(2.) + 0j, 0., 0., 1. / np.sqrt(2.)],
@@ -176,7 +182,7 @@ def test_kernel_complex_params_f32():
 
 @skipIfNvidiaFP64NotInstalled
 def test_kernel_complex_capture_f64():
-    cudaq.reset_target()
+
     cudaq.set_target('nvidia', option='fp64')
 
     c = np.array([1. / np.sqrt(2.) + 0j, 0., 0., 1. / np.sqrt(2.)],
@@ -187,7 +193,6 @@ def test_kernel_complex_capture_f64():
     def kernel():
         q = cudaq.qvector(state)
 
-    # kernel.compile()
     counts = cudaq.sample(kernel)
     print(counts)
     assert '11' in counts
@@ -196,7 +201,7 @@ def kernel():
 
 @skipIfNvidiaFP64NotInstalled
 def test_kernel_complex128_capture_f64():
-    cudaq.reset_target()
+
     cudaq.set_target('nvidia', option='fp64')
 
     c = np.array([1. / np.sqrt(2.) + 0j, 0., 0., 1. / np.sqrt(2.)],
@@ -215,7 +220,7 @@ def kernel():
 
 @skipIfNvidiaFP64NotInstalled
 def test_kernel_complex128_capture_f64():
-    cudaq.reset_target()
+
     cudaq.set_target('nvidia', option='fp64')
 
     c = np.array([1. / np.sqrt(2.) + 0j, 0., 0., 1. / np.sqrt(2.)],
@@ -228,27 +233,26 @@ def test_kernel_complex128_capture_f64():
 
 @skipIfNvidiaNotInstalled
 def test_kernel_complex64_capture_f32():
-    cudaq.reset_target()
+
     cudaq.set_target('nvidia')
 
     c = np.array([1. / np.sqrt(2.) + 0j, 0., 0., 1. / np.sqrt(2.)],
                  dtype=np.complex64)
     state = cudaq.State.from_data(c)
 
     @cudaq.kernel
-    def kernel():
+    def kernel_foo():
         q = cudaq.qvector(state)
 
-    # kernel.compile()
-    counts = cudaq.sample(kernel)
+    counts = cudaq.sample(kernel_foo)
     print(counts)
     assert '11' in counts
     assert '00' in counts
 
 
 @skipIfNvidiaNotInstalled
 def test_kernel_complex128_capture_f32():
-    cudaq.reset_target()
+
     cudaq.set_target('nvidia')
 
     c = np.array([1. / np.sqrt(2.) + 0j, 0., 0., 1. / np.sqrt(2.)],
@@ -261,7 +265,7 @@ def test_kernel_complex128_capture_f32():
 
 @skipIfNvidiaNotInstalled
 def test_kernel_complex_capture_f32():
-    cudaq.reset_target()
+
     cudaq.set_target('nvidia')
 
     c = np.array([1. / np.sqrt(2.) + 0j, 0., 0., 1. / np.sqrt(2.)],
@@ -277,77 +281,75 @@ def test_kernel_complex_capture_f32():
 
 @skipIfNvidiaFP64NotInstalled
 def test_kernel_simulation_dtype_complex_params_f64():
-    cudaq.reset_target()
+
     cudaq.set_target('nvidia', option='fp64')
 
     c = np.array([1. / np.sqrt(2.) + 0j, 0., 0., 1. / np.sqrt(2.)],
                  dtype=cudaq.complex())
     state = cudaq.State.from_data(c)
 
     @cudaq.kernel
-    def kernel(vec: cudaq.State):
+    def kernel_bar(vec: cudaq.State):
         q = cudaq.qvector(vec)
 
-    counts = cudaq.sample(kernel, state)
+    counts = cudaq.sample(kernel_bar, state)
     print(counts)
     assert '11' in counts
     assert '00' in counts
 
 
 @skipIfNvidiaNotInstalled
 def test_kernel_simulation_dtype_complex_params_f32():
-    cudaq.reset_target()
+
     cudaq.set_target('nvidia')
 
     c = np.array([1. / np.sqrt(2.) + 0j, 0., 0., 1. / np.sqrt(2.)],
                  dtype=cudaq.complex())
     state = cudaq.State.from_data(c)
 
     @cudaq.kernel
-    def kernel(vec: cudaq.State):
+    def kernel_baz(vec: cudaq.State):
         q = cudaq.qvector(vec)
 
-    counts = cudaq.sample(kernel, state)
+    counts = cudaq.sample(kernel_baz, state)
     print(counts)
     assert '11' in counts
     assert '00' in counts
 
 
 @skipIfNvidiaFP64NotInstalled
 def test_kernel_simulation_dtype_capture_f64():
-    cudaq.reset_target()
+
     cudaq.set_target('nvidia', option='fp64')
 
     c = np.array([1. / np.sqrt(2.) + 0j, 0., 0., 1. / np.sqrt(2.)],
                  dtype=cudaq.complex())
     state = cudaq.State.from_data(c)
 
     @cudaq.kernel
-    def kernel():
+    def kernel_qux():
         q = cudaq.qvector(state)
 
-    # kernel.compile()
-    counts = cudaq.sample(kernel)
+    counts = cudaq.sample(kernel_qux)
     print(counts)
     assert '11' in counts
     assert '00' in counts
 
 
 @skipIfNvidiaNotInstalled
 def test_kernel_simulation_dtype_capture_f32():
-    cudaq.reset_target()
+
     cudaq.set_target('nvidia')
 
     c = np.array([1. / np.sqrt(2.) + 0j, 0., 0., 1. / np.sqrt(2.)],
                  dtype=cudaq.complex())
     state = cudaq.State.from_data(c)
 
     @cudaq.kernel
-    def kernel():
+    def kernel_corge():
         q = cudaq.qvector(state)
 
-    # kernel.compile()
-    counts = cudaq.sample(kernel)
+    counts = cudaq.sample(kernel_corge)
     print(counts)
     assert '11' in counts
     assert '00' in counts
@@ -358,7 +360,7 @@ def kernel():
 
 @skipIfNvidiaFP64NotInstalled
 def test_init_from_other_kernel_state_f64():
-    cudaq.reset_target()
+
     cudaq.set_target('nvidia', option='fp64')
 
     @cudaq.kernel
@@ -371,13 +373,13 @@ def bell():
     state.dump()
 
     @cudaq.kernel
-    def kernel(initialState: cudaq.State):
+    def kernel_waldo(initialState: cudaq.State):
         qubits = cudaq.qvector(initialState)
 
-    state2 = cudaq.get_state(kernel, state)
+    state2 = cudaq.get_state(kernel_waldo, state)
     state2.dump()
 
-    counts = cudaq.sample(kernel, state)
+    counts = cudaq.sample(kernel_waldo, state)
     print(counts)
     assert '11' in counts
     assert '00' in counts
@@ -387,7 +389,7 @@ def kernel(initialState: cudaq.State):
 
 @skipIfNvidiaFP64NotInstalled
 def test_init_from_other_kernel_state_f32():
-    cudaq.reset_target()
+
     cudaq.set_target('nvidia')
 
     @cudaq.kernel
@@ -400,13 +402,13 @@ def bell():
     state.dump()
 
     @cudaq.kernel
-    def kernel(initialState: cudaq.State):
+    def kernel_thud(initialState: cudaq.State):
         qubits = cudaq.qvector(initialState)
 
-    state2 = cudaq.get_state(kernel, state)
+    state2 = cudaq.get_state(kernel_thud, state)
     state2.dump()
 
-    counts = cudaq.sample(kernel, state)
+    counts = cudaq.sample(kernel_thud, state)
     print(counts)
     assert '11' in counts
     assert '00' in counts
@@ -434,7 +436,6 @@ def kernel2():
 
         kernel2()
 
-    # kernel0.compile()
     counts = cudaq.sample(kernel0)
     print(counts)
     assert '1111' in counts
@@ -444,7 +445,6 @@ def kernel2():
 
 
 def test_invalid_arg_error_msg():
-    cudaq.reset_target()
 
     c = np.array([1. / np.sqrt(2.) + 0j, 0., 0., 1. / np.sqrt(2.)],
                  dtype=complex)