Merge branch 'fix-temp-default-shape' into 'master'

Fix the default shape of loopy.TemporaryVariable

See merge request inducer/loopy!354

Author: inducer

loopy/kernel/array.py

@@ -934,7 +934,8 @@ def num_target_axes(self):
         return len(target_axes)
 
     def num_user_axes(self, require_answer=True):
-        if self.shape is not None:
+        from loopy import auto
+        if self.shape not in (None, auto):
             return len(self.shape)
         if self.dim_tags is not None:
             return len(self.dim_tags)

loopy/kernel/data.py

@@ -525,7 +525,7 @@ class TemporaryVariable(ArrayBase):
             "_base_storage_access_may_be_aliasing",
             ]
 
-    def __init__(self, name, dtype=None, shape=(), address_space=None,
+    def __init__(self, name, dtype=None, shape=auto, address_space=None,
             dim_tags=None, offset=0, dim_names=None, strides=None, order=None,
             base_indices=None, storage_shape=None,
             base_storage=None, initializer=None, read_only=False,
@@ -579,7 +579,10 @@ def __init__(self, name, dtype=None, shape=(), address_space=None,
 
             if shape is auto:
                 shape = initializer.shape
-
+            else:
+                if shape != initializer.shape:
+                    raise LoopyError("Shape of '{}' does not match that of the"
+                            " initializer.".format(name))
         else:
             raise LoopyError(
                     "temporary variable '%s': "
@@ -589,7 +592,7 @@ def __init__(self, name, dtype=None, shape=(), address_space=None,
         if order is None:
             order = "C"
 
-        if base_indices is None:
+        if base_indices is None and shape is not auto:
             base_indices = (0,) * len(shape)
 
         if not read_only and initializer is not None:

loopy/target/opencl.py

@@ -551,8 +551,10 @@ def emit_atomic_update(self, codegen_state, lhs_atomicity, lhs_var,
             from loopy.kernel.data import TemporaryVariable, AddressSpace
             ecm = codegen_state.expression_to_code_mapper.with_assignments(
                     {
-                        old_val_var: TemporaryVariable(old_val_var, lhs_dtype),
-                        new_val_var: TemporaryVariable(new_val_var, lhs_dtype),
+                        old_val_var: TemporaryVariable(old_val_var, lhs_dtype,
+                            shape=()),
+                        new_val_var: TemporaryVariable(new_val_var, lhs_dtype,
+                            shape=()),
                         })
 
             lhs_expr_code = ecm(lhs_expr, prec=PREC_NONE, type_context=None)

test/test_loopy.py

@@ -68,8 +68,8 @@ def test_globals_decl_once_with_multi_subprogram(ctx_factory):
             out[ii] = 2*out[ii]+cnst[ii]{id=second}
             """,
             [lp.TemporaryVariable(
-                'cnst', shape=('n'), initializer=cnst,
-                address_space=lp.AddressSpace.GLOBAL,
+                'cnst', initializer=cnst,
+                scope=lp.AddressSpace.GLOBAL,
                 read_only=True), '...'])
     knl = lp.fix_parameters(knl, n=16)
     knl = lp.add_barrier(knl, "id:first", "id:second")