diff --git a/loopy/auto_test.py b/loopy/auto_test.py index 988e83f88..f69515133 100644 --- a/loopy/auto_test.py +++ b/loopy/auto_test.py @@ -529,7 +529,7 @@ def auto_test_vs_ref( test_knl = lp.preprocess_kernel(test_knl) - if not test_knl.schedule: + if not test_knl.linearization: test_kernels = lp.generate_loop_schedules(test_knl) else: test_kernels = [test_knl] diff --git a/loopy/check.py b/loopy/check.py index 133097073..b9610785b 100644 --- a/loopy/check.py +++ b/loopy/check.py @@ -906,18 +906,18 @@ def _check_for_unused_hw_axes_in_kernel_chunk(kernel, sched_index=None): local_axes = set() i = 0 - loop_end_i = past_end_i = len(kernel.schedule) + loop_end_i = past_end_i = len(kernel.linearization) else: - assert isinstance(kernel.schedule[sched_index], CallKernel) - _, past_end_i = gather_schedule_block(kernel.schedule, sched_index) + assert isinstance(kernel.linearization[sched_index], CallKernel) + _, past_end_i = gather_schedule_block(kernel.linearization, sched_index) group_size, local_size = kernel.get_grid_sizes_for_insn_ids_as_exprs( - get_insn_ids_for_block_at(kernel.schedule, sched_index)) + get_insn_ids_for_block_at(kernel.linearization, sched_index)) group_axes = {ax for ax, length in enumerate(group_size)} local_axes = {ax for ax, length in enumerate(local_size)} i = sched_index + 1 - assert isinstance(kernel.schedule[past_end_i - 1], ReturnFromKernel) + assert isinstance(kernel.linearization[past_end_i - 1], ReturnFromKernel) loop_end_i = past_end_i - 1 # alternative: just disregard length-1 dimensions? @@ -926,7 +926,7 @@ def _check_for_unused_hw_axes_in_kernel_chunk(kernel, sched_index=None): GroupIndexTag) while i < loop_end_i: - sched_item = kernel.schedule[i] + sched_item = kernel.linearization[i] if isinstance(sched_item, CallKernel): i = _check_for_unused_hw_axes_in_kernel_chunk(kernel, i) @@ -984,7 +984,7 @@ def _check_for_unused_hw_axes_in_kernel_chunk(kernel, sched_index=None): def check_for_unused_hw_axes_in_insns(kernel): - if kernel.schedule: + if kernel.linearization: _check_for_unused_hw_axes_in_kernel_chunk(kernel) # }}} @@ -1077,7 +1077,7 @@ def check_that_all_insns_are_scheduled(kernel): from loopy.schedule import sched_item_to_insn_id scheduled_insns = { insn_id - for sched_item in kernel.schedule + for sched_item in kernel.linearization for insn_id in sched_item_to_insn_id(sched_item)} assert scheduled_insns <= all_schedulable_insns diff --git a/loopy/codegen/__init__.py b/loopy/codegen/__init__.py index 694ebb07a..e30e1a73d 100644 --- a/loopy/codegen/__init__.py +++ b/loopy/codegen/__init__.py @@ -411,7 +411,7 @@ def generate_code_v2(kernel): from loopy.preprocess import preprocess_kernel kernel = preprocess_kernel(kernel) - if kernel.schedule is None: + if kernel.linearization is None: from loopy.schedule import get_one_linearized_kernel kernel = get_one_linearized_kernel(kernel) @@ -500,7 +500,7 @@ def generate_code_v2(kernel): kernel.target.host_program_name_prefix + kernel.name + kernel.target.host_program_name_suffix), - schedule_index_end=len(kernel.schedule), + schedule_index_end=len(kernel.linearization), codegen_cachemanager=CodegenOperationCacheManager.from_kernel(kernel), ) diff --git a/loopy/codegen/control.py b/loopy/codegen/control.py index f7e953d94..e60553f58 100644 --- a/loopy/codegen/control.py +++ b/loopy/codegen/control.py @@ -35,7 +35,7 @@ def synthesize_idis_for_extra_args(kernel, schedule_index): """ :returns: A list of :class:`loopy.codegen.ImplementedDataInfo` """ - sched_item = kernel.schedule[schedule_index] + sched_item = kernel.linearization[schedule_index] from loopy.codegen import ImplementedDataInfo from loopy.kernel.data import InameArg, AddressSpace @@ -66,13 +66,13 @@ def synthesize_idis_for_extra_args(kernel, schedule_index): def generate_code_for_sched_index(codegen_state, sched_index): kernel = codegen_state.kernel - sched_item = kernel.schedule[sched_index] + sched_item = kernel.linearization[sched_index] if isinstance(sched_item, CallKernel): assert not codegen_state.is_generating_device_code from loopy.schedule import (gather_schedule_block, get_insn_ids_for_block_at) - _, past_end_i = gather_schedule_block(kernel.schedule, sched_index) + _, past_end_i = gather_schedule_block(kernel.linearization, sched_index) assert past_end_i <= codegen_state.schedule_index_end extra_args = synthesize_idis_for_extra_args(kernel, sched_index) @@ -89,7 +89,7 @@ def generate_code_for_sched_index(codegen_state, sched_index): new_codegen_state, sched_index) glob_grid, loc_grid = kernel.get_grid_sizes_for_insn_ids_as_exprs( - get_insn_ids_for_block_at(kernel.schedule, sched_index)) + get_insn_ids_for_block_at(kernel.linearization, sched_index)) return merge_codegen_results(codegen_state, [ codegen_result, @@ -176,7 +176,7 @@ def generate_code_for_sched_index(codegen_state, sched_index): def get_required_predicates(kernel, sched_index): result = None - for _, sched_item in generate_sub_sched_items(kernel.schedule, sched_index): + for _, sched_item in generate_sub_sched_items(kernel.linearization, sched_index): if isinstance(sched_item, Barrier): my_preds = frozenset() elif isinstance(sched_item, RunInstruction): @@ -238,7 +238,7 @@ def build_loop_nest(codegen_state, schedule_index): i = schedule_index while i < codegen_state.schedule_index_end: - sched_item = kernel.schedule[i] + sched_item = kernel.linearization[i] if isinstance(sched_item, LeaveLoop): break @@ -246,7 +246,7 @@ def build_loop_nest(codegen_state, schedule_index): my_sched_indices.append(i) if isinstance(sched_item, (EnterLoop, CallKernel)): - _, i = gather_schedule_block(kernel.schedule, i) + _, i = gather_schedule_block(kernel.linearization, i) assert i <= codegen_state.schedule_index_end, \ "schedule block extends beyond schedule_index_end" diff --git a/loopy/codegen/loop.py b/loopy/codegen/loop.py index d149eb95a..277bc9087 100644 --- a/loopy/codegen/loop.py +++ b/loopy/codegen/loop.py @@ -121,7 +121,7 @@ def get_slab_decomposition(kernel, iname): def generate_unroll_loop(codegen_state, sched_index): kernel = codegen_state.kernel - iname = kernel.schedule[sched_index].iname + iname = kernel.linearization[sched_index].iname bounds = kernel.get_iname_bounds(iname, constants_only=True) @@ -163,7 +163,7 @@ def generate_unroll_loop(codegen_state, sched_index): def generate_vectorize_loop(codegen_state, sched_index): kernel = codegen_state.kernel - iname = kernel.schedule[sched_index].iname + iname = kernel.linearization[sched_index].iname bounds = kernel.get_iname_bounds(iname, constants_only=True) @@ -236,7 +236,8 @@ def set_up_hw_parallel_loops(codegen_state, schedule_index, next_func, LocalIndexTag, GroupIndexTag, VectorizeTag) from loopy.schedule import get_insn_ids_for_block_at - insn_ids_for_block = get_insn_ids_for_block_at(kernel.schedule, schedule_index) + insn_ids_for_block = get_insn_ids_for_block_at(kernel.linearization, + schedule_index) if hw_inames_left is None: all_inames_by_insns = set() @@ -348,7 +349,7 @@ def generate_sequential_loop_dim_code(codegen_state, sched_index): kernel = codegen_state.kernel ecm = codegen_state.expression_to_code_mapper - loop_iname = kernel.schedule[sched_index].iname + loop_iname = kernel.linearization[sched_index].iname slabs = get_slab_decomposition(kernel, loop_iname) diff --git a/loopy/codegen/result.py b/loopy/codegen/result.py index 685df8fde..f06b17f17 100644 --- a/loopy/codegen/result.py +++ b/loopy/codegen/result.py @@ -292,7 +292,8 @@ def generate_host_or_device_program(codegen_state, schedule_index): from loopy.codegen.control import build_loop_nest if codegen_state.is_generating_device_code: from loopy.schedule import CallKernel - assert isinstance(codegen_state.kernel.schedule[schedule_index], CallKernel) + assert isinstance(codegen_state.kernel.linearization[schedule_index], + CallKernel) from loopy.codegen.loop import set_up_hw_parallel_loops codegen_result = set_up_hw_parallel_loops( diff --git a/loopy/codegen/tools.py b/loopy/codegen/tools.py index 6451d6054..72cfe17c6 100644 --- a/loopy/codegen/tools.py +++ b/loopy/codegen/tools.py @@ -73,7 +73,7 @@ def __eq__(self, other): # relevant to CodegenOperationCacheManager return (self.inames == other.inames and self.instructions == other.instructions - and self.schedule == other.schedule) + and self.schedule == other.linearization) class CodegenOperationCacheManager: @@ -96,7 +96,7 @@ def from_kernel(kernel): assert isinstance(kernel, LoopKernel) return CodegenOperationCacheManager( KernelProxyForCodegenOperationCacheManager(kernel.instructions, - kernel.schedule, + kernel.linearization, kernel.inames)) def with_kernel(self, kernel): diff --git a/loopy/kernel/__init__.py b/loopy/kernel/__init__.py index 6885e1ab5..a8ef3e4e8 100644 --- a/loopy/kernel/__init__.py +++ b/loopy/kernel/__init__.py @@ -394,11 +394,11 @@ def __init__(self, domains, instructions, args=None, # these should not both be present raise ValueError( "received both `schedule` and `linearization` args, " - "'LoopKernel.schedule' is deprecated. " + "'LoopKernel.linearization' is deprecated. " "Use 'LoopKernel.linearization'.") elif schedule is not None: warn( - "'LoopKernel.schedule' is deprecated. " + "'LoopKernel.linearization' is deprecated. " "Use 'LoopKernel.linearization'.", DeprecationWarning, stacklevel=2) linearization = schedule @@ -771,7 +771,7 @@ def _get_inames_domain_backend(self, inames): @property def schedule(self): warn( - "LoopKernel.schedule is deprecated. " + "LoopKernel.linearization is deprecated. " "Call LoopKernel.linearization instead, " "will be unsupported in 2022.", DeprecationWarning, stacklevel=2) @@ -1410,12 +1410,12 @@ def stringify(self, what=None, with_dependencies=False, use_separators=True, "(use loopy.show_dependency_graph to visualize)") lines.extend(dep_lines) - if "schedule" in what and kernel.schedule is not None: + if "schedule" in what and kernel.linearization is not None: lines.extend(sep) if show_labels: lines.append("LINEARIZATION:") from loopy.schedule import dump_schedule - lines.append(dump_schedule(kernel, kernel.schedule)) + lines.append(dump_schedule(kernel, kernel.linearization)) lines.extend(sep) @@ -1541,7 +1541,7 @@ def __setstate__(self, state): "domains", "instructions", "args", - "schedule", + "linearization", "name", "preambles", "assumptions", diff --git a/loopy/kernel/tools.py b/loopy/kernel/tools.py index f660ed4e2..18d012b8a 100644 --- a/loopy/kernel/tools.py +++ b/loopy/kernel/tools.py @@ -473,7 +473,7 @@ def get_dot_dependency_graph(kernel, iname_cluster=True, use_insn_id=False): from loopy.kernel.creation import apply_single_writer_depencency_heuristic kernel = apply_single_writer_depencency_heuristic(kernel, warn_if_used=False) - if iname_cluster and not kernel.schedule: + if iname_cluster and not kernel.linearization: try: from loopy.schedule import get_one_linearized_kernel kernel = get_one_linearized_kernel(kernel) @@ -550,7 +550,7 @@ def get_dot_dependency_graph(kernel, iname_cluster=True, use_insn_id=False): EnterLoop, LeaveLoop, RunInstruction, Barrier, CallKernel, ReturnFromKernel) - for sched_item in kernel.schedule: + for sched_item in kernel.linearization: if isinstance(sched_item, EnterLoop): lines.append('subgraph cluster_%s { label="%s"' % (sched_item.iname, sched_item.iname)) @@ -1736,7 +1736,7 @@ def get_subkernels(kernel): from loopy.schedule import CallKernel return tuple(sched_item.kernel_name - for sched_item in kernel.schedule + for sched_item in kernel.linearization if isinstance(sched_item, CallKernel)) @@ -1756,7 +1756,7 @@ def get_subkernel_to_insn_id_map(kernel): subkernel = None result = {} - for sched_item in kernel.schedule: + for sched_item in kernel.linearization: if isinstance(sched_item, CallKernel): subkernel = sched_item.kernel_name result[subkernel] = set() diff --git a/loopy/schedule/__init__.py b/loopy/schedule/__init__.py index a854ce04e..86bd005f6 100644 --- a/loopy/schedule/__init__.py +++ b/loopy/schedule/__init__.py @@ -169,11 +169,11 @@ def get_insn_ids_for_block_at(schedule, start_idx): def find_used_inames_within(kernel, sched_index): - sched_item = kernel.schedule[sched_index] + sched_item = kernel.linearization[sched_index] if isinstance(sched_item, BeginBlockItem): loop_contents, _ = gather_schedule_block( - kernel.schedule, sched_index) + kernel.linearization, sched_index) run_insns = [subsched_item for subsched_item in loop_contents if isinstance(subsched_item, RunInstruction)] @@ -1955,7 +1955,9 @@ def generate_loop_schedules_inner(kernel, debug_args={}): debug = ScheduleDebugger(**debug_args) - preschedule = kernel.schedule if kernel.state == KernelState.LINEARIZED else () + preschedule = (kernel.linearization + if kernel.state == KernelState.LINEARIZED + else ()) prescheduled_inames = { insn.iname diff --git a/loopy/schedule/device_mapping.py b/loopy/schedule/device_mapping.py index 6ad83456e..35c73b775 100644 --- a/loopy/schedule/device_mapping.py +++ b/loopy/schedule/device_mapping.py @@ -43,7 +43,7 @@ def map_schedule_onto_host_or_device(kernel): [CallKernel(kernel_name=device_prog_name_gen(), extra_args=[], extra_inames=[])] + - list(kernel.schedule) + + list(kernel.linearization) + [ReturnFromKernel(kernel_name=kernel.name)]) kernel = kernel.copy(linearization=new_schedule) else: @@ -54,7 +54,7 @@ def map_schedule_onto_host_or_device(kernel): def map_schedule_onto_host_or_device_impl(kernel, device_prog_name_gen): - schedule = kernel.schedule + schedule = kernel.linearization loop_bounds = get_block_boundaries(schedule) # {{{ inner mapper function diff --git a/loopy/schedule/tools.py b/loopy/schedule/tools.py index 796d16381..a94586ab8 100644 --- a/loopy/schedule/tools.py +++ b/loopy/schedule/tools.py @@ -78,7 +78,7 @@ def add_extra_args_to_schedule(kernel): new_schedule = [] from loopy.schedule import CallKernel - for sched_item in kernel.schedule: + for sched_item in kernel.linearization: if isinstance(sched_item, CallKernel): subkernel = sched_item.kernel_name diff --git a/loopy/statistics.py b/loopy/statistics.py index ef335abce..c4c43b290 100755 --- a/loopy/statistics.py +++ b/loopy/statistics.py @@ -1787,7 +1787,7 @@ def get_count_poly(iname_list): else: return one - for sched_item in knl.schedule: + for sched_item in knl.linearization: if isinstance(sched_item, EnterLoop): if sched_item.iname: # (if not empty) iname_list.append(sched_item.iname) diff --git a/loopy/target/c/__init__.py b/loopy/target/c/__init__.py index 92b94d658..51cbc1506 100644 --- a/loopy/target/c/__init__.py +++ b/loopy/target/c/__init__.py @@ -618,7 +618,7 @@ def get_function_definition(self, codegen_state, codegen_result, # whether this is the first device program in the schedule. is_first_dev_prog = codegen_state.is_generating_device_code for i in range(schedule_index): - if isinstance(kernel.schedule[i], CallKernel): + if isinstance(kernel.linearization[i], CallKernel): is_first_dev_prog = False break if is_first_dev_prog: @@ -707,7 +707,7 @@ def get_temporary_decls(self, codegen_state, schedule_index): from loopy.schedule.tools import ( temporaries_read_in_subkernel, temporaries_written_in_subkernel) - subkernel = kernel.schedule[schedule_index].kernel_name + subkernel = kernel.linearization[schedule_index].kernel_name sub_knl_temps = ( temporaries_read_in_subkernel(kernel, subkernel) | temporaries_written_in_subkernel(kernel, subkernel)) diff --git a/loopy/target/cuda.py b/loopy/target/cuda.py index 67dc1fe24..56a1d2040 100644 --- a/loopy/target/cuda.py +++ b/loopy/target/cuda.py @@ -260,7 +260,7 @@ def get_function_declaration(self, codegen_state, codegen_result, _, local_grid_size = \ codegen_state.kernel.get_grid_sizes_for_insn_ids_as_exprs( get_insn_ids_for_block_at( - codegen_state.kernel.schedule, schedule_index)) + codegen_state.kernel.linearization, schedule_index)) from loopy.symbolic import get_dependencies if not get_dependencies(local_grid_size): diff --git a/loopy/target/execution.py b/loopy/target/execution.py index 222839e56..2c31f389f 100644 --- a/loopy/target/execution.py +++ b/loopy/target/execution.py @@ -769,7 +769,7 @@ def get_typed_and_scheduled_kernel_uncached(self, arg_to_dtype_set): from loopy.type_inference import infer_unknown_types kernel = infer_unknown_types(kernel, expect_completion=True) - if kernel.schedule is None: + if kernel.linearization is None: from loopy.preprocess import preprocess_kernel kernel = preprocess_kernel(kernel) diff --git a/loopy/target/opencl.py b/loopy/target/opencl.py index d9afebea9..d5e99c5bc 100644 --- a/loopy/target/opencl.py +++ b/loopy/target/opencl.py @@ -468,7 +468,7 @@ def get_function_declaration(self, codegen_state, codegen_result, from loopy.schedule import get_insn_ids_for_block_at _, local_sizes = codegen_state.kernel.get_grid_sizes_for_insn_ids_as_exprs( get_insn_ids_for_block_at( - codegen_state.kernel.schedule, schedule_index)) + codegen_state.kernel.linearization, schedule_index)) from loopy.symbolic import get_dependencies if not get_dependencies(local_sizes): diff --git a/loopy/transform/save.py b/loopy/transform/save.py index 0a9cfb7bc..4872d59b7 100644 --- a/loopy/transform/save.py +++ b/loopy/transform/save.py @@ -61,12 +61,12 @@ class LivenessAnalysis: def __init__(self, kernel): self.kernel = kernel - self.schedule = self.kernel.schedule + self.schedule = self.kernel.linearization @memoize_method def get_successor_relation(self): successors = {} - block_bounds = get_block_boundaries(self.kernel.schedule) + block_bounds = get_block_boundaries(self.kernel.linearization) for idx, (item, next_item) in enumerate(zip( reversed(self.schedule), @@ -313,7 +313,7 @@ def base_storage_to_temporary_map(self): def subkernel_to_slice_indices(self): result = {} - for sched_item_idx, sched_item in enumerate(self.kernel.schedule): + for sched_item_idx, sched_item in enumerate(self.kernel.linearization): if isinstance(sched_item, CallKernel): start_idx = sched_item_idx elif isinstance(sched_item, ReturnFromKernel): @@ -328,7 +328,7 @@ def subkernel_to_surrounding_inames(self): within_subkernel = False result = {} - for sched_item_idx, sched_item in enumerate(self.kernel.schedule): + for sched_item_idx, sched_item in enumerate(self.kernel.linearization): if isinstance(sched_item, CallKernel): within_subkernel = True result[sched_item.kernel_name] = frozenset(current_outer_inames) @@ -353,14 +353,14 @@ def is_global_barrier(item): try: pre_barrier = next(item for item in - self.kernel.schedule[subkernel_start::-1] + self.kernel.linearization[subkernel_start::-1] if is_global_barrier(item)).originating_insn_id except StopIteration: pre_barrier = None try: post_barrier = next(item for item in - self.kernel.schedule[subkernel_end:] + self.kernel.linearization[subkernel_end:] if is_global_barrier(item)).originating_insn_id except StopIteration: post_barrier = None @@ -746,7 +746,7 @@ def save_and_reload_temporaries(kernel): from loopy.schedule.tools import ( temporaries_read_in_subkernel, temporaries_written_in_subkernel) - for sched_idx, sched_item in enumerate(kernel.schedule): + for sched_idx, sched_item in enumerate(kernel.linearization): if isinstance(sched_item, CallKernel): # Any written temporary that is live-out needs to be read into @@ -766,7 +766,7 @@ def save_and_reload_temporaries(kernel): saver.reload(temporary, sched_item.kernel_name) elif isinstance(sched_item, ReturnFromKernel): - if sched_idx == len(kernel.schedule) - 1: + if sched_idx == len(kernel.linearization) - 1: # Kernel exit: nothing live interesting_temporaries = set() else: diff --git a/test/test_loopy.py b/test/test_loopy.py index 1afdeb9c3..9700270bf 100644 --- a/test/test_loopy.py +++ b/test/test_loopy.py @@ -1060,8 +1060,8 @@ def test_kernel_splitting(ctx_factory): from loopy.preprocess import preprocess_kernel knl = preprocess_kernel(knl) - from loopy.schedule import get_one_scheduled_kernel - knl = get_one_scheduled_kernel(knl) + from loopy.schedule import get_one_linearized_kernel + knl = get_one_linearized_kernel(knl) # map schedule onto host or device print(knl) @@ -1101,8 +1101,8 @@ def test_kernel_splitting_with_loop(ctx_factory): from loopy.preprocess import preprocess_kernel knl = preprocess_kernel(knl) - from loopy.schedule import get_one_scheduled_kernel - knl = get_one_scheduled_kernel(knl) + from loopy.schedule import get_one_linearized_kernel + knl = get_one_linearized_kernel(knl) # map schedule onto host or device print(knl) @@ -1119,14 +1119,14 @@ def test_kernel_splitting_with_loop(ctx_factory): def save_and_reload_temporaries_test(queue, knl, out_expect, debug=False): from loopy.preprocess import preprocess_kernel - from loopy.schedule import get_one_scheduled_kernel + from loopy.schedule import get_one_linearized_kernel knl = preprocess_kernel(knl) - knl = get_one_scheduled_kernel(knl) + knl = get_one_linearized_kernel(knl) from loopy.transform.save import save_and_reload_temporaries knl = save_and_reload_temporaries(knl) - knl = get_one_scheduled_kernel(knl) + knl = get_one_linearized_kernel(knl) if debug: print(knl) @@ -1390,7 +1390,7 @@ def test_save_ambiguous_storage_requirements(): knl = lp.set_temporary_scope(knl, "a", "local") knl = lp.preprocess_kernel(knl) - knl = lp.get_one_scheduled_kernel(knl) + knl = lp.get_one_linearized_kernel(knl) from loopy.diagnostic import LoopyError with pytest.raises(LoopyError): @@ -1747,7 +1747,7 @@ def test_missing_global_barrier(): from loopy.diagnostic import MissingBarrierError with pytest.raises(MissingBarrierError): - lp.get_one_scheduled_kernel(knl) + lp.get_one_linearized_kernel(knl) def test_index_cse(ctx_factory): @@ -1879,7 +1879,7 @@ def test_const_temp_with_initializer_not_saved(): seq_dependencies=True) knl = lp.preprocess_kernel(knl) - knl = lp.get_one_scheduled_kernel(knl) + knl = lp.get_one_linearized_kernel(knl) knl = lp.save_and_reload_temporaries(knl) # This ensures no save slot was added. @@ -2085,7 +2085,7 @@ def test_unscheduled_insn_detection(): """, "...") - knl = lp.get_one_scheduled_kernel(lp.preprocess_kernel(knl)) + knl = lp.get_one_linearized_kernel(lp.preprocess_kernel(knl)) insn1, = lp.find_instructions(knl, "id:insn1") knl.instructions.append(insn1.copy(id="insn2")) @@ -2210,7 +2210,7 @@ def barrier_between(knl, id1, id2, ignore_barriers_in_levels=()): seen_barrier = False loop_level = 0 - for sched_item in knl.schedule: + for sched_item in knl.linearization: if isinstance(sched_item, RunInstruction): if sched_item.insn_id == id1: watch_for_barrier = True @@ -2250,7 +2250,7 @@ def test_barrier_insertion_near_top_of_loop(): knl = lp.tag_inames(knl, dict(i="l.0")) knl = lp.set_temporary_scope(knl, "a", "local") knl = lp.set_temporary_scope(knl, "b", "local") - knl = lp.get_one_scheduled_kernel(lp.preprocess_kernel(knl)) + knl = lp.get_one_linearized_kernel(lp.preprocess_kernel(knl)) print(knl) @@ -2277,7 +2277,7 @@ def test_barrier_insertion_near_bottom_of_loop(): knl = lp.tag_inames(knl, dict(i="l.0")) knl = lp.set_temporary_scope(knl, "a", "local") knl = lp.set_temporary_scope(knl, "b", "local") - knl = lp.get_one_scheduled_kernel(lp.preprocess_kernel(knl)) + knl = lp.get_one_linearized_kernel(lp.preprocess_kernel(knl)) print(knl) @@ -2646,7 +2646,7 @@ def test_check_for_variable_access_ordering(): from loopy.diagnostic import VariableAccessNotOrdered with pytest.raises(VariableAccessNotOrdered): - lp.get_one_scheduled_kernel(knl) + lp.get_one_linearized_kernel(knl) def test_check_for_variable_access_ordering_with_aliasing(): @@ -2665,7 +2665,7 @@ def test_check_for_variable_access_ordering_with_aliasing(): from loopy.diagnostic import VariableAccessNotOrdered with pytest.raises(VariableAccessNotOrdered): - lp.get_one_scheduled_kernel(knl) + lp.get_one_linearized_kernel(knl) @pytest.mark.parametrize(("second_index", "expect_barrier"), @@ -2688,7 +2688,7 @@ def test_no_barriers_for_nonoverlapping_access(second_index, expect_barrier): knl = lp.tag_inames(knl, "i:l.0") knl = lp.preprocess_kernel(knl) - knl = lp.get_one_scheduled_kernel(knl) + knl = lp.get_one_linearized_kernel(knl) assert barrier_between(knl, "first", "second") == expect_barrier diff --git a/test/test_target.py b/test/test_target.py index 3bf625c3a..25f3e1dc9 100644 --- a/test/test_target.py +++ b/test/test_target.py @@ -71,7 +71,7 @@ def test_ispc_target(occa_mode=False): default_tag="l.auto") codegen_result = lp.generate_code_v2( - lp.get_one_scheduled_kernel( + lp.get_one_linearized_kernel( lp.preprocess_kernel(knl))) print(codegen_result.device_code()) @@ -97,7 +97,7 @@ def test_cuda_target(): print( lp.generate_code( - lp.get_one_scheduled_kernel( + lp.get_one_linearized_kernel( lp.preprocess_kernel(knl)))[0]) @@ -140,7 +140,7 @@ def test_generate_c_snippet(): knl = lp.prioritize_loops(knl, "I,k_outer,k_inner") knl = lp.preprocess_kernel(knl) - knl = lp.get_one_scheduled_kernel(knl) + knl = lp.get_one_linearized_kernel(knl) print(lp.generate_body(knl)) @@ -355,7 +355,7 @@ def test_ispc_streaming_stores(): knl = lp.set_argument_order(knl, vars + ["n"]) knl = lp.preprocess_kernel(knl) - knl = lp.get_one_scheduled_kernel(knl) + knl = lp.get_one_linearized_kernel(knl) assert "streaming_store(" in lp.generate_code_v2(knl).all_code() diff --git a/test/test_transform.py b/test/test_transform.py index b154a9468..50b283795 100644 --- a/test/test_transform.py +++ b/test/test_transform.py @@ -271,7 +271,7 @@ def test_vectorize(ctx_factory): knl = lp.tag_inames(knl, {"i_inner": "vec"}) knl = lp.preprocess_kernel(knl) - knl = lp.get_one_scheduled_kernel(knl) + knl = lp.get_one_linearized_kernel(knl) code, inf = lp.generate_code(knl) lp.auto_test_vs_ref(