fix(bindings): repair the two example command-line paths that #2546 revives

cuda_bindings/examples/0_Introduction/simple_zero_copy.py

@@ -69,7 +69,9 @@ def main():
 
     # Get the device selected by the user or default to 0, and then set it.
     if check_cmd_line_flag("device="):
-        device_count = cudart.cudaGetDeviceCount()
+        # cudaGetDeviceCount returns (cudaError_t, count); comparing the raw
+        # tuple against an int below raises TypeError.
+        device_count = check_cuda_errors(cudart.cudaGetDeviceCount())
         idev = int(get_cmd_line_argument_int("device="))
 
         if idev >= device_count or idev < 0:

cuda_bindings/examples/3_CUDA_Features/global_to_shmem_async_copy.py

@@ -799,6 +799,11 @@ def matrix_multiply(dims_a, dims_b, kernel_number):
     grid_shared_state_kernel = cudart.dim3()
     grid_shared_state_kernel.x = dims_b.x / threads_shared_state_kernel.x
     grid_shared_state_kernel.y = dims_a.y / threads_shared_state_kernel.x
+    # C++ dim3 defaults every component to 1; cudart.dim3 wraps a
+    # zero-initialised struct, so .z must be set explicitly or the two
+    # cuLaunchKernel calls below pass gridDimZ=0 and fail with
+    # CUDA_ERROR_INVALID_VALUE.
+    grid_shared_state_kernel.z = 1
 
     print(f"Running kernel = {kernel_number} - {kernel_names[kernel_number.value]}")
     # Create and start timer