Add typos CI, fix typos

.github/workflows/ci.yml

@@ -20,6 +20,13 @@ jobs:
                 pipx install ruff
                 ruff check
 
+    typos:
+        name: Typos
+        runs-on: ubuntu-latest
+        steps:
+        -   uses: actions/checkout@v4
+        -   uses: crate-ci/typos@master
+
     pylint:
         name: Pylint
         runs-on: ubuntu-latest

MEMO

@@ -7,7 +7,7 @@ Documentation Notes
 Things to consider
 ^^^^^^^^^^^^^^^^^^
 
-- Depedencies are pointwise for shared loop dimensions
+- Dependencies are pointwise for shared loop dimensions
   and global over non-shared ones (between dependent and ancestor)
 
 - multiple insns could fight over which iname gets local axis 0

contrib/mem-pattern-explorer/pattern_vis.py

@@ -76,7 +76,7 @@ def tick(self):
 class Array:
     def __init__(self, ctx, name, shape, strides, elements_per_row=None):
         # Each array element stores a tuple:
-        # (timestamp, subgroup, g0, g1, g2, ) of last acccess
+        # (timestamp, subgroup, g0, g1, g2, ) of last access
 
         assert len(shape) == len(strides)
 

doc/misc.rst

@@ -158,7 +158,7 @@ In the meantime, you can generate code simply by saying::
     print(cg_result.host_code())
     print(cg_result.device_code())
 
-Additionally, for C-based languages, header defintions are available via::
+Additionally, for C-based languages, header definitions are available via::
 
     loopy.generate_header(knl)
 
@@ -338,8 +338,8 @@ This list is always growing, but here are a few pointers:
 
   Use :func:`loopy.join_inames`.
 
-In what sense does Loopy suport vectorization?
-----------------------------------------------
+In what sense does Loopy support vectorization?
+-----------------------------------------------
 
 There are really two ways in which the OpenCL/CUDA model of computation exposes
 vectorization:
@@ -352,7 +352,7 @@ vectorization:
   e.g. ``float4``, which support arithmetic with implicit vector semantics
   as well as a number of 'intrinsic' functions.
 
-Loopy suports both. The first one, SIMT, is accessible by tagging inames with,
+Loopy supports both. The first one, SIMT, is accessible by tagging inames with,
 e.g., ``l.0```. Accessing the second one requires using both execution- and
 data-reshaping capabilities in loopy. To start with, you need an array that
 has an axis with the length of the desired vector. If that's not yet available,

loopy/__init__.py

@@ -563,18 +563,18 @@ def make_copy_kernel(new_dim_tags, old_dim_tags=None):
 
     indices = ["i%d" % i for i in range(rank)]
     shape = ["n%d" % i for i in range(rank)]
-    commad_indices = ", ".join(indices)
+    command_indices = ", ".join(indices)
     bounds = " and ".join(
             f"0<={ind}<{shape_i}"
             for ind, shape_i in zip(indices, shape))
 
     set_str = "{{[{}]: {} }}".format(
-                commad_indices,
+                command_indices,
                 bounds
                 )
     result = make_kernel(set_str,
             "output[%s] = input[%s]"
-            % (commad_indices, commad_indices),
+            % (command_indices, command_indices),
             lang_version=MOST_RECENT_LANGUAGE_VERSION,
             default_offset=auto)
 

loopy/check.py

@@ -1123,7 +1123,7 @@ def satisfy_dep_reqs_in_order(dep_reqs_to_vars, edges, order):
             # for each *pred*, we will calculate all the direct/indirect
             # instructions that can be reached.
             seen_successors = set()
-            # first let us start with direct sucessors
+            # first let us start with direct successors
             to_check = edges[pred].copy()
             while to_check:
                 successor = to_check.pop()
@@ -1219,7 +1219,7 @@ def check_variable_access_ordered(kernel):
     """Checks that between each write to a variable and all other accesses to
     the variable there is either:
 
-    * a direct/indirect depdendency edge, or
+    * a direct/indirect dependency edge, or
     * an explicit statement that no ordering is necessary (expressed
       through a bi-directional :attr:`loopy.InstructionBase.no_sync_with`)
     """

loopy/frontend/fortran/translator.py

@@ -200,7 +200,7 @@ def get_type(self, name, none_ok=False):
                     return None
 
                 raise TranslationError(
-                        "no type for '%s' found in 'implict none' routine"
+                        "no type for '%s' found in 'implicit none' routine"
                         % name) from None
 
             return self.implicit_types.get(name[0], np.dtype(np.int32))
@@ -426,7 +426,7 @@ def map_Implicit(self, node):
             scope.implicit_types = None
 
         for stmt, specs in node.items:
-            if scope.implict_types is None:
+            if scope.implict_types is None:  # spellchecker: disable-line
                 raise TranslationError("implicit decl not allowed after "
                         "'implicit none'")
             tp = self.dtype_from_stmt(stmt)

loopy/isl_helpers.py

@@ -186,7 +186,7 @@ def simplify_pw_aff(pw_aff, context=None):
                     continue
 
                 if aff_i.gist(dom_j).is_equal(aff_j):
-                    # aff_i is sufficient to conver aff_j, eliminate aff_j
+                    # aff_i is sufficient to cover aff_j, eliminate aff_j
                     new_pieces = pieces[:]
                     if i < j:
                         new_pieces.pop(j)

loopy/kernel/__init__.py

@@ -164,7 +164,7 @@ class LoopKernel(Taggable):
     .. attribute:: loop_priority
 
         A frozenset of priority constraints to the kernel. Each such constraint
-        is a tuple of inames. Inames occuring in such a tuple will be scheduled
+        is a tuple of inames. Inames occurring in such a tuple will be scheduled
         earlier than any iname following in the tuple. This applies only to inames
         with non-parallel implementation tags.
 
@@ -515,7 +515,7 @@ def get_leaf_domain_indices(self, inames):
         for iname in inames:
             home_domain_index = hdm[iname]
             if home_domain_index in domain_indices:
-                # nothin' new
+                # nothing new
                 continue
 
             domain_path_to_root = [home_domain_index] + ppd[home_domain_index]

loopy/kernel/creation.py

@@ -1884,7 +1884,7 @@ def add_inferred_inames(knl):
 # {{{ apply single-writer heuristic
 
 @for_each_kernel
-def apply_single_writer_depencency_heuristic(kernel, warn_if_used=True,
+def apply_single_writer_dependency_heuristic(kernel, warn_if_used=True,
         error_if_used=False):
     logger.debug("%s: default deps" % kernel.name)
 
@@ -2023,7 +2023,7 @@ class SliceToInameReplacer(IdentityMapper):
     .. attribute:: subarray_ref_bounds
 
         A :class:`list` (one entry for each :class:`SubArrayRef` to be created)
-        of :class:`dict` instances to store the slices enountered in the
+        of :class:`dict` instances to store the slices encountered in the
         expressions as a mapping from ``iname`` to a tuple of ``(start, stop,
         step)``, which describes the boxy (i.e. affine) constraints imposed on
         the ``iname`` by the corresponding slice notation its intended to
@@ -2574,7 +2574,7 @@ def make_function(domains, instructions, kernel_data=None, **kwargs):
     knl = guess_arg_shape_if_requested(knl, default_order)
     knl = apply_default_order_to_args(knl, default_order)
     knl = resolve_dependencies(knl)
-    knl = apply_single_writer_depencency_heuristic(knl, warn_if_used=False)
+    knl = apply_single_writer_dependency_heuristic(knl, warn_if_used=False)
 
     # -------------------------------------------------------------------------
     # Ordering dependency:

loopy/kernel/function_interface.py

@@ -672,7 +672,7 @@ def is_type_specialized(self):
 
 class CallableKernel(InKernelCallable):
     """
-    Records informations about a callee kernel. Also provides interface through
+    Records information about a callee kernel. Also provides interface through
     member methods to make the callee kernel compatible to be called from a
     caller kernel.
 

loopy/kernel/instruction.py

@@ -934,7 +934,7 @@ class CallInstruction(MultiAssignmentBase):
 
         A tuple of `:class:loopy.Optional`. If an entry is not empty, it
         contains the type that will be assigned to the new temporary variable
-        created from the assigment.
+        created from the assignment.
 
     .. automethod:: __init__
     """
@@ -1099,7 +1099,7 @@ def is_array_call(assignees, expression):
     Returns *True* is the instruction is an array call.
 
     An array call is a function call applied to array type objects. If any of
-    the arguemnts or assignees to the function is an array,
+    the arguments or assignees to the function is an array,
     :meth:`is_array_call` will return *True*.
     """
     from pymbolic.primitives import Call, Subscript
@@ -1460,7 +1460,7 @@ class BarrierInstruction(_DataObliviousInstruction):
     .. attribute:: mem_kind
 
         A string, ``"global"`` or ``"local"``. Chooses which memory type to
-        sychronize, for targets that require this (e.g. OpenCL)
+        synchronize, for targets that require this (e.g. OpenCL)
 
     The textual syntax in a :mod:`loopy` kernel is::
 

loopy/kernel/tools.py

@@ -263,7 +263,7 @@ def find_all_insn_inames(kernel):
             if insn.within_inames_is_final:
                 continue
 
-            # {{{ depdency-based propagation
+            # {{{ dependency-based propagation
 
             inames_old = insn_id_to_inames[insn.id]
             inames_new = inames_old | guess_iname_deps_based_on_var_use(
@@ -513,8 +513,8 @@ def get_dot_dependency_graph(kernel, callables_table, iname_cluster=True,
     """
 
     # make sure all automatically added stuff shows up
-    from loopy.kernel.creation import apply_single_writer_depencency_heuristic
-    kernel = apply_single_writer_depencency_heuristic(kernel, warn_if_used=False)
+    from loopy.kernel.creation import apply_single_writer_dependency_heuristic
+    kernel = apply_single_writer_dependency_heuristic(kernel, warn_if_used=False)
 
     if iname_cluster and not kernel.linearization:
         try:
@@ -1252,9 +1252,9 @@ def find_recursive_dependencies(kernel, insn_ids):
 
         for insn_id in queue:
             insn = kernel.id_to_insn[insn_id]
-            additionals = insn.depends_on - result
-            result.update(additionals)
-            new_queue.extend(additionals)
+            additional = insn.depends_on - result
+            result.update(additional)
+            new_queue.extend(additional)
 
         queue = new_queue
 
@@ -1735,7 +1735,7 @@ def get_global_barrier_order(kernel):
 
 @memoize_on_first_arg
 def find_most_recent_global_barrier(kernel, insn_id):
-    """Return the id of the latest occuring global barrier which the
+    """Return the id of the latest occurring global barrier which the
     given instruction (indirectly or directly) depends on, or *None* if this
     instruction does not depend on a global barrier.
 
@@ -1995,7 +1995,7 @@ def infer_args_are_input_output(kernel):
         elif isinstance(arg, (ConstantArg, ImageArg, ValueArg)):
             pass
         else:
-            raise NotImplementedError("Unkonwn argument type %s." % type(arg))
+            raise NotImplementedError("Unknown argument type %s." % type(arg))
 
         if not (arg.is_input or arg.is_output):
             raise LoopyError("Kernel argument must be either input or output."

loopy/match.py

@@ -1,4 +1,4 @@
-"""Matching functionality for instruction ids and subsitution
+"""Matching functionality for instruction ids and substitution
 rule invocations stacks."""
 
 

loopy/options.py

@@ -118,7 +118,7 @@ class Options(ImmutableRecord):
 
     .. attribute:: cl_exec_manage_array_events
 
-        Within the PyOpenCL executor, respect and udpate
+        Within the PyOpenCL executor, respect and update
         :attr:`pyopencl.array.Array.events`.
 
         Defaults to *True*.
@@ -156,7 +156,7 @@ class Options(ImmutableRecord):
 
         Allow re-ordering of floating point arithmetic. Re-ordering may
         give different results as floating point arithmetic is not
-        associative in addition and mulitplication. Default is *True*.
+        associative in addition and multiplication. Default is *True*.
         Note that the implementation of this option is currently incomplete.
 
     .. attribute:: build_options

loopy/preprocess.py

@@ -825,8 +825,8 @@ def preprocess_program(t_unit: TranslationUnit) -> TranslationUnit:
     from loopy.transform.subst import expand_subst
     t_unit = expand_subst(t_unit)
 
-    from loopy.kernel.creation import apply_single_writer_depencency_heuristic
-    t_unit = apply_single_writer_depencency_heuristic(t_unit)
+    from loopy.kernel.creation import apply_single_writer_dependency_heuristic
+    t_unit = apply_single_writer_dependency_heuristic(t_unit)
 
     # Ordering restrictions:
     #

loopy/schedule/__init__.py

@@ -718,7 +718,7 @@ def get_insns_in_topologically_sorted_order(
         for dep in insn.depends_on:
             rev_dep_map[dep].add(insn.id)
 
-    # For breaking ties, we compare the features of an intruction
+    # For breaking ties, we compare the features of an instruction
     # so that instructions with the same set of features are lumped
     # together. This helps in :method:`schedule_as_many_run_insns_as_possible`
     # which bails after 5 insns that don't have the same feature.
@@ -1196,7 +1196,7 @@ def insn_sort_key(insn_id):
                                 print(
                                     "%(warn)swarning:%(reset_all)s '%(iname)s', "
                                     "which the schedule is "
-                                    "currently stuck inside of, seems mis-nested. "
+                                    "currently stuck inside of, seems misnested. "
                                     "'%(subdep)s' must occur " "before '%(dep)s', "
                                     "but '%(subdep)s must be outside "
                                     "'%(iname)s', whereas '%(dep)s' must be back "
@@ -1404,7 +1404,7 @@ def insn_sort_key(insn_id):
                     get_priority_tiers(wanted, sched_state.kernel.loop_priority))
 
             # Update the loop priority set, because some constraints may have
-            # have been contradictary.
+            # have been contradictory.
             loop_priority_set = set().union(*[set(t) for t in priority_tiers])
 
             priority_tiers.append(

loopy/statistics.py

@@ -629,7 +629,7 @@ class Op(ImmutableRecord):
        work-group executes on a single compute unit with all work-items within
        the work-group sharing local memory. A sub-group is an
        implementation-dependent grouping of work-items within a work-group,
-       analagous to an NVIDIA CUDA warp.
+       analogous to an NVIDIA CUDA warp.
 
     .. attribute:: kernel_name
 
@@ -723,7 +723,7 @@ class MemAccess(ImmutableRecord):
        work-group executes on a single compute unit with all work-items within
        the work-group sharing local memory. A sub-group is an
        implementation-dependent grouping of work-items within a work-group,
-       analagous to an NVIDIA CUDA warp.
+       analogous to an NVIDIA CUDA warp.
 
     .. attribute:: kernel_name
 
@@ -1109,7 +1109,7 @@ def _get_lid_and_gid_strides(knl, array, index):
 
     # create lid_strides and gid_strides dicts
 
-    # strides are coefficents in flattened index, i.e., we want
+    # strides are coefficients in flattened index, i.e., we want
     # lid_strides = {0:l0, 1:l1, 2:l2, ...} and
     # gid_strides = {0:g0, 1:g1, 2:g2, ...},
     # where l0, l1, l2, g0, g1, and g2 come from flattened index
@@ -1723,7 +1723,7 @@ def get_op_map(program, count_redundant_work=False,
     :arg subgroup_size: (currently unused) An :class:`int`, :class:`str`
         ``"guess"``, or *None* that specifies the sub-group size. An OpenCL
         sub-group is an implementation-dependent grouping of work-items within
-        a work-group, analagous to an NVIDIA CUDA warp. subgroup_size is used,
+        a work-group, analogous to an NVIDIA CUDA warp. subgroup_size is used,
         e.g., when counting a :class:`MemAccess` whose count_granularity
         specifies that it should only be counted once per sub-group. If set to
         *None* an attempt to find the sub-group size using the device will be
@@ -1921,7 +1921,7 @@ def get_mem_access_map(program, count_redundant_work=False,
     :arg subgroup_size: An :class:`int`, :class:`str` ``"guess"``, or
         *None* that specifies the sub-group size. An OpenCL sub-group is an
         implementation-dependent grouping of work-items within a work-group,
-        analagous to an NVIDIA CUDA warp. subgroup_size is used, e.g., when
+        analogous to an NVIDIA CUDA warp. subgroup_size is used, e.g., when
         counting a :class:`MemAccess` whose count_granularity specifies that it
         should only be counted once per sub-group. If set to *None* an attempt
         to find the sub-group size using the device will be made, if this fails
@@ -2085,7 +2085,7 @@ def get_synchronization_map(program, subgroup_size=None, entrypoint=None):
     :arg subgroup_size: (currently unused) An :class:`int`, :class:`str`
         ``"guess"``, or *None* that specifies the sub-group size. An OpenCL
         sub-group is an implementation-dependent grouping of work-items within
-        a work-group, analagous to an NVIDIA CUDA warp. subgroup_size is used,
+        a work-group, analogous to an NVIDIA CUDA warp. subgroup_size is used,
         e.g., when counting a :class:`MemAccess` whose count_granularity
         specifies that it should only be counted once per sub-group. If set to
         *None* an attempt to find the sub-group size using the device will be

loopy/symbolic.py

@@ -683,7 +683,7 @@ class TaggedVariable(LoopyExpressionBase, p.Variable, Taggable):
         A :class:`frozenset` of subclasses of :class:`pytools.tag.Tag` used to
         provide metadata on this object. Legacy string tags are converted to
         :class:`~loopy.LegacyStringInstructionTag` or, if they used to carry
-        a functional meaning, the tag carrying that same fucntional meaning
+        a functional meaning, the tag carrying that same functional meaning
         (e.g. :class:`~loopy.UseStreamingStoreTag`).
 
     Inherits from :class:`pymbolic.primitives.Variable`
@@ -737,7 +737,7 @@ class Reduction(LoopyExpressionBase):
     .. attribute:: allow_simultaneous
 
         A :class:`bool`. If not *True*, an iname is allowed to be used
-        in precisely one reduction, to avoid mis-nesting errors.
+        in precisely one reduction, to avoid misnesting errors.
     """
 
     init_arg_names = ("operation", "inames", "expr", "allow_simultaneous")