Avoid temp variable assignments

numba/core/interpreter.py

@@ -540,14 +540,69 @@ def _remove_unused_temporaries(self):
         current block.
         """
         new_body = []
+        replaced_var = {}
         for inst in self.current_block.body:
-            if (isinstance(inst, ir.Assign)
-                    and inst.target.is_temp
-                    and inst.target.name in self.assigner.unused_dests):
-                continue
+            # the same temporary is assigned to multiple variables in cases
+            # like a = b[i] = 1, so need to handle replaced temporaries in
+            # later setitem/setattr nodes
+            if (isinstance(inst, (ir.SetItem, ir.SetItem))
+                    and inst.value.name in replaced_var):
+                inst.value = replaced_var[inst.value.name]
+            elif isinstance(inst, ir.Assign):
+                if (inst.target.is_temp
+                        and inst.target.name in self.assigner.unused_dests):
+                    continue
+                # the same temporary is assigned to multiple variables in cases
+                # like a = b = 1, so need to handle replaced temporaries in
+                # later assignments
+                if (isinstance(inst.value, ir.Var)
+                        and inst.value.name in replaced_var):
+                    inst.value = replaced_var[inst.value.name]
+                    new_body.append(inst)
+                    continue
+                # chained unpack cases may reuse temporary
+                # e.g. a = (b, c) = (x, y)
+                if (isinstance(inst.value, ir.Expr)
+                        and inst.value.op == "exhaust_iter"
+                        and inst.value.value.name in replaced_var):
+                    inst.value.value = replaced_var[inst.value.value.name]
+                    new_body.append(inst)
+                    continue
+                # eliminate temporary variables that are assigned to user
+                # variables right after creation. E.g.:
+                # $1 = f(); a = $1 -> a = f()
+                # the temporary variable is not reused elsewhere since CPython
+                # bytecode is stack-based and this pattern corresponds to a pop
+                if (isinstance(inst.value, ir.Var) and inst.value.is_temp
+                        and new_body and isinstance(new_body[-1], ir.Assign)):
+                    prev_assign = new_body[-1]
+                    # _var_used_in_binop check makes sure we don't create a new
+                    # inplace binop operation which can fail
+                    # (see TestFunctionType.test_in_iter_func_call)
+                    if (prev_assign.target.name == inst.value.name
+                            and not self._var_used_in_binop(
+                                inst.target.name, prev_assign.value)):
+                        replaced_var[inst.value.name] = inst.target
+                        prev_assign.target = inst.target
+                        # replace temp var definition in target with proper defs
+                        self.definitions[inst.target.name].remove(inst.value)
+                        self.definitions[inst.target.name].extend(
+                            self.definitions.pop(inst.value.name)
+                        )
+                        continue
+
             new_body.append(inst)
+
         self.current_block.body = new_body
 
+    def _var_used_in_binop(self, varname, expr):
+        """return True if 'expr' is a binary expression and 'varname' is used
+        in it as an argument
+        """
+        return (isinstance(expr, ir.Expr)
+                and expr.op in ("binop", "inplace_binop")
+                and (varname == expr.lhs.name or varname == expr.rhs.name))
+
     def _insert_outgoing_phis(self):
         """
         Add assignments to forward requested outgoing values

numba/core/ir_utils.py

@@ -1739,23 +1739,25 @@ def replace_arg_nodes(block, args):
             stmt.value = args[idx]
     return
 
+
 def replace_returns(blocks, target, return_label):
     """
     Return return statement by assigning directly to target, and a jump.
     """
     for block in blocks.values():
-        casts = []
-        for i, stmt in enumerate(block.body):
-            if isinstance(stmt, ir.Return):
-                assert(i + 1 == len(block.body))
-                block.body[i] = ir.Assign(stmt.value, target, stmt.loc)
-                block.body.append(ir.Jump(return_label, stmt.loc))
-                # remove cast of the returned value
-                for cast in casts:
-                    if cast.target.name == stmt.value.name:
-                        cast.value = cast.value.value
-            elif isinstance(stmt, ir.Assign) and isinstance(stmt.value, ir.Expr) and stmt.value.op == 'cast':
-                casts.append(stmt)
+        # some blocks may be empty during transformations
+        if not block.body:
+            continue
+        stmt = block.terminator
+        if isinstance(stmt, ir.Return):
+            block.body.pop()  # remove return
+            cast_stmt = block.body.pop()
+            assert (isinstance(cast_stmt, ir.Assign)
+                and isinstance(cast_stmt.value, ir.Expr)
+                and cast_stmt.value.op == 'cast'), "invalid return cast"
+            block.body.append(ir.Assign(cast_stmt.value.value, target, stmt.loc))
+            block.body.append(ir.Jump(return_label, stmt.loc))
+
 
 def gen_np_call(func_as_str, func, lhs, args, typingctx, typemap, calltypes):
     scope = args[0].scope

numba/core/typeinfer.py

@@ -1601,7 +1601,14 @@ def typeof_global(self, inst, target, gvar):
         # Setting literal_value for globals because they are handled
         # like const value in numba
         lit = types.maybe_literal(gvar.value)
-        self.lock_type(target.name, lit or typ, loc=inst.loc)
+        # The user may have provided the type for this variable already.
+        # In this case, call add_type() to make sure the value type is
+        # consistent. See numba.tests.test_array_reductions for examples.
+        tv = self.typevars[target.name]
+        if tv.locked:
+            tv.add_type(lit or typ, loc=inst.loc)
+        else:
+            self.lock_type(target.name, lit or typ, loc=inst.loc)
         self.assumed_immutables.add(inst)
 
     def typeof_expr(self, inst, target, expr):

numba/parfors/parfor.py

@@ -3522,13 +3522,17 @@ def check_conflicting_reduction_operators(param, nodes):
 def get_reduction_init(nodes):
     """
     Get initial value for known reductions.
-    Currently, only += and *= are supported. We assume the inplace_binop node
-    is followed by an assignment.
+    Currently, only += and *= are supported.
     """
-    require(len(nodes) >=2)
-    require(isinstance(nodes[-1].value, ir.Var))
-    require(nodes[-2].target.name == nodes[-1].value.name)
-    acc_expr = nodes[-2].value
+    require(len(nodes) >=1)
+    # there could be an extra assignment after the reduce node
+    # See: test_reduction_var_reuse
+    if isinstance(nodes[-1].value, ir.Var):
+        require(len(nodes) >=2)
+        require(nodes[-2].target.name == nodes[-1].value.name)
+        acc_expr = nodes[-2].value
+    else:
+        acc_expr = nodes[-1].value
     require(isinstance(acc_expr, ir.Expr) and acc_expr.op=='inplace_binop')
     if acc_expr.fn == operator.iadd or acc_expr.fn == operator.isub:
         return 0, acc_expr.fn
@@ -3573,9 +3577,13 @@ def lookup(var, varonly=True):
         if isinstance(rhs, ir.Expr):
             in_vars = set(lookup(v, True).name for v in rhs.list_vars())
             if name in in_vars:
-                next_node = nodes[i+1]
-                target_name = next_node.target.unversioned_name
-                if not (isinstance(next_node, ir.Assign) and target_name == unversioned_name):
+                # reductions like sum have an assignment afterwards
+                # e.g. $2 = a + $1; a = $2
+                # reductions that are functions calls like max() don't have an
+                # extra assignment afterwards
+                if (not (i+1 < len(nodes) and isinstance(nodes[i+1], ir.Assign)
+                        and nodes[i+1].target.unversioned_name == unversioned_name)
+                        and lhs.unversioned_name != unversioned_name):
                     raise ValueError(
                         f"Use of reduction variable {unversioned_name!r} other "
                         "than in a supported reduction function is not "
@@ -3593,7 +3601,7 @@ def lookup(var, varonly=True):
                 replace_dict[non_red_args[0]] = ir.Var(lhs.scope, name+"#init", lhs.loc)
                 replace_vars_inner(rhs, replace_dict)
                 reduce_nodes = nodes[i:]
-                break;
+                break
     return reduce_nodes
 
 def get_expr_args(expr):

numba/parfors/parfor_lowering.py

@@ -1385,7 +1385,10 @@ def replace_var_with_array_in_block(vars, block, typemap, calltypes):
             const_assign = ir.Assign(const_node, const_var, inst.loc)
             new_block.append(const_assign)
 
-            setitem_node = ir.SetItem(inst.target, const_var, inst.value, inst.loc)
+            val_var = ir.Var(inst.target.scope, mk_unique_var("$val"), inst.loc)
+            typemap[val_var.name] = typemap[inst.target.name]
+            new_block.append(ir.Assign(inst.value, val_var, inst.loc))
+            setitem_node = ir.SetItem(inst.target, const_var, val_var, inst.loc)
             calltypes[setitem_node] = signature(
                 types.none, types.npytypes.Array(typemap[inst.target.name], 1, "C"), types.intp, typemap[inst.target.name])
             new_block.append(setitem_node)

numba/tests/test_copy_propagate.py

@@ -13,11 +13,12 @@
 
 def test_will_propagate(b, z, w):
     x = 3
+    x1 = x
     if b > 0:
         y = z + w
     else:
         y = 0
-    a = 2 * x
+    a = 2 * x1
     return a < b
 
 def test_wont_propagate(b, z, w):
@@ -54,15 +55,11 @@ def test1(self):
         typingctx = typing.Context()
         targetctx = cpu.CPUContext(typingctx)
         test_ir = compiler.run_frontend(test_will_propagate)
-        #print("Num blocks = ", len(test_ir.blocks))
-        #print(test_ir.dump())
         with cpu_target.nested_context(typingctx, targetctx):
             typingctx.refresh()
             targetctx.refresh()
             args = (types.int64, types.int64, types.int64)
             typemap, return_type, calltypes, _ = type_inference_stage(typingctx, test_ir, args, None)
-            #print("typemap = ", typemap)
-            #print("return_type = ", return_type)
             type_annotation = type_annotations.TypeAnnotation(
                 func_ir=test_ir,
                 typemap=typemap,
@@ -75,14 +72,12 @@ def test1(self):
             in_cps, out_cps = copy_propagate(test_ir.blocks, typemap)
             apply_copy_propagate(test_ir.blocks, in_cps, get_name_var_table(test_ir.blocks), typemap, calltypes)
 
-            self.assertFalse(findAssign(test_ir, "x"))
+            self.assertFalse(findAssign(test_ir, "x1"))
 
     def test2(self):
         typingctx = typing.Context()
         targetctx = cpu.CPUContext(typingctx)
         test_ir = compiler.run_frontend(test_wont_propagate)
-        #print("Num blocks = ", len(test_ir.blocks))
-        #print(test_ir.dump())
         with cpu_target.nested_context(typingctx, targetctx):
             typingctx.refresh()
             targetctx.refresh()
@@ -102,5 +97,6 @@ def test2(self):
 
             self.assertTrue(findAssign(test_ir, "x"))
 
+
 if __name__ == "__main__":
     unittest.main()

numba/tests/test_function_type.py

@@ -1,3 +1,4 @@
+import unittest
 import types as pytypes
 from numba import jit, njit, cfunc, types, int64, float64, float32, errors
 from numba import literal_unroll
@@ -1252,3 +1253,7 @@ def bar(fcs, ffs):
         got = bar(tup, tup_bar)
         expected = foo1(a) + foo2(a) + bar1(a) + bar2(a)
         self.assertEqual(got, expected)
+
+
+if __name__ == '__main__':
+    unittest.main()

numba/tests/test_ir_utils.py

@@ -39,7 +39,7 @@ def test_func():
         typing_res = type_inference_stage(
             typingctx, test_ir, (), None)
         matched_call = ir_utils.find_callname(
-            test_ir, test_ir.blocks[0].body[8].value, typing_res.typemap)
+            test_ir, test_ir.blocks[0].body[7].value, typing_res.typemap)
         self.assertTrue(isinstance(matched_call, tuple) and
                         len(matched_call) == 2 and
                         matched_call[0] == 'append')
@@ -89,7 +89,7 @@ def check_initial_ir(the_ir):
             # an assign of above into to variable `dead`
             # a const int above 0xdeaddead
             # an assign of said int to variable `deaddead`
-            # this is 4 things to remove
+            # this is 2 statements to remove
 
             self.assertEqual(len(the_ir.blocks), 1)
             block = the_ir.blocks[0]
@@ -99,18 +99,14 @@ def check_initial_ir(the_ir):
                     if 'dead' in getattr(x.target, 'name', ''):
                         deads.append(x)
 
-            expect_removed = []
             self.assertEqual(len(deads), 2)
-            expect_removed.extend(deads)
             for d in deads:
                 # check the ir.Const is the definition and the value is expected
                 const_val = the_ir.get_definition(d.value)
                 self.assertTrue(int('0x%s' % d.target.name, 16),
                                 const_val.value)
-                expect_removed.append(const_val)
 
-            self.assertEqual(len(expect_removed), 4)
-            return expect_removed
+            return deads
 
         def check_dce_ir(the_ir):
             self.assertEqual(len(the_ir.blocks), 1)

numba/tests/test_remove_dead.py

@@ -30,7 +30,8 @@
 
 
 def test_will_propagate(b, z, w):
-    x = 3
+    x1 = 3
+    x = x1
     if b > 0:
         y = z + w
     else:

numba/tests/test_unpack_sequence.py

@@ -235,4 +235,3 @@ def test_invalid_unpack(self):
 
 if __name__ == '__main__':
     unittest.main()
-