Thread safety

This MR introduces a simple check for thread safety in the kernel constrains check. If the assignments are not thread safe, optimizations like OpenMP or GPU should fail.

tests/test_vectorization_specific.py

@@ -226,14 +226,13 @@ def test_issue62(dtype, instruction_set, padding):
     dy = sp.Symbol("dy")
     src, dst, rhs = ps.fields(f"src, src_tmp, rhs: {dtype}[2D]", layout='fzyx')
 
-    up = ps.Assignment(src[0, 0], ((dy ** 2 * (src[1, 0] + src[-1, 0]))
+    up = ps.Assignment(dst[0, 0], ((dy ** 2 * (src[1, 0] + src[-1, 0]))
                                    + (dx ** 2 * (src[0, 1] + src[0, -1]))
                                    - (rhs[0, 0] * dx ** 2 * dy ** 2)) / (2 * (dx ** 2 + dy ** 2)))
 
     config = ps.CreateKernelConfig(data_type=dtype,
                                    default_number_float=dtype,
-                                   cpu_vectorize_info=opt,
-                                   skip_independence_check=True)
+                                   cpu_vectorize_info=opt)
 
     ast = ps.create_kernel(up, config=config)
     ast.compile()