compiler: Implement element type coercion for arrays

So far, this is not exposed to the user beyond implicit conversions.

Note that all the implicit conversions, such as triggered by adding
arrays of mismatching types, or dividing integer arrays, are currently
emitted in a maximally inefficient way, where a temporary copy is first
made for the type conversion. The conversions would more sensibly be
implemented during the per-element operations to save on the extra
copies, but the current behaviour fell out of the rest of the IR
generator structure without extra changes.

artiq/compiler/transforms/artiq_ir_generator.py

@@ -1415,7 +1415,19 @@ class ARTIQIRGenerator(algorithm.Visitor):
         if node.type.find() == value.type:
             return value
         else:
-            return self.append(ir.Coerce(value, node.type,
+            if builtins.is_array(node.type):
+                result_elt = node.type.find()["elt"]
+                shape = self.append(ir.GetAttr(value, "shape"))
+                result = self._alloate_new_array(result_elt, shape)
+                func = self._get_array_unaryop("Coerce",
+                                               lambda v: ir.Coerce(v, result_elt),
+                                               node.type, value.type)
+                self._invoke_arrayop(func, [result, value])
+                return result
+            else:
+                return self.append(
+                    ir.Coerce(value,
+                              node.type,
                               name="{}.{}".format(_readable_name(value),
                                                   node.type.name)))
 

artiq/test/lit/integration/array_binops.py

@@ -33,12 +33,15 @@ assert c[0] == 0
 assert c[1] == 1
 assert c[2] == 0
 
-# FIXME: Implement array coercion.
+cf = b / a
-# c = b / a
+assert cf[0] == 4.0
-# assert c[0] == 4.0
+assert cf[1] == 2.5
-# assert c[1] == 2.5
+assert cf[2] == 2.0
-# assert c[2] == 2.0
 
+cf2 = cf + a
+assert cf2[0] == 5.0
+assert cf2[1] == 4.5
+assert cf2[2] == 5.0
 
 d = array([[1, 2], [3, 4]])
 e = array([[5, 6], [7, 8]])