[XLA:TPU] Add a bunch of F64 Convert ops. F64 <-> C64/U64/U32.

PiperOrigin-RevId: 306344563
Change-Id: Ia24263b9d51be5e057a1d3d9314a6c5fa647038b

tensorflow/compiler/tests/unary_ops_test.py

@@ -808,19 +808,46 @@ class UnaryOpsTest(xla_test.XLATestCase):
 
   def testCast(self):
     shapes = [[], [4], [2, 3], [2, 0, 4]]
-    types = (
+    types = {
-        set([dtypes.bool, dtypes.int32, dtypes.float32])
+        dtypes.bool, dtypes.float32, dtypes.float64, dtypes.complex64,
-        | self.complex_tf_types)
+        dtypes.int32, dtypes.int64, dtypes.uint32, dtypes.uint64
-    for shape in shapes:
+    }
     for src_type in types:
       for dst_type in types:
-          src = np.arange(np.prod(shape)).astype(src_type.as_numpy_dtype)
+        src_np_dtype = src_type.as_numpy_dtype
-          if src_type in self.complex_tf_types:
+        dst_np_dtype = dst_type.as_numpy_dtype
-            src += (np.arange(np.prod(shape)) * 2j).astype(
-                src_type.as_numpy_dtype)
-          src = src.reshape(shape)
 
-          dst = src.astype(dst_type.as_numpy_dtype)
+        for shape in shapes:
+          src = np.arange(np.prod(shape)).astype(src_np_dtype)
+
+          if src_type in self.complex_tf_types:
+            src += (np.arange(np.prod(shape)) * 2j).astype(src_np_dtype)
+          src = src.reshape(shape)
+          dst = src.astype(dst_np_dtype)
+          self._assertOpOutputMatchesExpected(
+              lambda x, dst_type=dst_type: math_ops.cast(x, dst_type),
+              src,
+              expected=dst)
+
+        # Check special values.
+        if src_type.is_integer:
+          imin = np.iinfo(src_np_dtype).min
+          imax = np.iinfo(src_np_dtype).max
+          src = np.array([imin, imax, 0, 1, -1], dtype=src_np_dtype)
+        elif src_type in self.float_tf_types:
+          if dst_type.is_integer:
+            imin = np.iinfo(dst_np_dtype).min
+            imax = np.iinfo(dst_np_dtype).max // 2
+            src = np.array([imin, imax, 0, 1], dtype=src_np_dtype)
+          elif dst_type in self.float_tf_types:
+            fmin = np.finfo(dst_np_dtype).min
+            fmax = np.finfo(dst_np_dtype).max
+            tiny = np.finfo(dst_np_dtype).tiny
+            eps = np.finfo(dst_np_dtype).eps
+            src = np.array(
+                [fmin, fmax, np.nan, eps, -eps, tiny, -tiny, np.inf, -np.inf],
+                dtype=src_np_dtype)
+        dst = src.astype(dst_np_dtype)
         self._assertOpOutputMatchesExpected(
             lambda x, dst_type=dst_type: math_ops.cast(x, dst_type),
             src,