This is an automated email from the ASF dual-hosted git repository.
github-bot pushed a change to branch nightly
in repository https://gitbox.apache.org/repos/asf/tvm.git
from 176c77388b [ci][docker] Nightly Docker image update (#18710)
add 198df475fa [REFACTOR][TEST] Migrate all codegen test to tvmscript
(#18719)
add 49b6d7b1f7 Remove cron schedule from nightly Docker update workflow
(#18724)
add 3dce4aedb5 [REFACTOR][S-TIR] Lift transform passes to s_tir namespace
(#18722)
add f7948c5952 [REFACTOR][S-TIR] Migrate tir/schedule to s_tir (#18728)
add d23d1dbc24 fix: correct typos in Python docstrings (#18727)
add 9eac0e1635 [Relax][PyTorch] Fix scalar parameter inputs in Dynamo
(#18725)
No new revisions were added by this update.
Summary of changes:
.github/workflows/nightly_docker_update.yml | 2 -
CMakeLists.txt | 1 +
include/tvm/meta_schedule/cost_model.h | 2 +-
include/tvm/meta_schedule/database.h | 20 +-
include/tvm/meta_schedule/measure_candidate.h | 6 +-
include/tvm/meta_schedule/mutator.h | 16 +-
include/tvm/meta_schedule/postproc.h | 8 +-
.../tvm/meta_schedule/schedule/cuda/thread_bind.h | 17 +-
.../tvm/meta_schedule/schedule/generic/winograd.h | 4 +-
include/tvm/meta_schedule/schedule_rule.h | 11 +-
include/tvm/meta_schedule/search_strategy.h | 8 +-
include/tvm/meta_schedule/space_generator.h | 8 +-
include/tvm/{tir => s_tir}/schedule/instruction.h | 19 +-
include/tvm/{tir => s_tir}/schedule/schedule.h | 21 +-
include/tvm/{tir => s_tir}/schedule/state.h | 15 +-
include/tvm/{tir => s_tir}/schedule/trace.h | 15 +-
include/tvm/s_tir/transform.h | 237 ++++
include/tvm/tir/transform.h | 245 -----
python/tvm/contrib/mrvl.py | 4 +-
python/tvm/dlight/analysis/common_analysis.py | 8 +-
.../frontend/torch/base_fx_graph_translator.py | 4 +-
python/tvm/relax/frontend/torch/dynamo.py | 7 +-
python/tvm/relax/transform/transform.py | 4 +-
python/tvm/s_tir/__init__.py | 1 +
python/tvm/s_tir/backend/adreno/pipeline.py | 40 +-
python/tvm/s_tir/pipeline.py | 40 +-
python/tvm/s_tir/schedule/_ffi_api.py | 2 +-
python/tvm/s_tir/schedule/analysis.py | 4 +-
python/tvm/s_tir/schedule/instruction.py | 4 +-
python/tvm/s_tir/schedule/schedule.py | 6 +-
python/tvm/s_tir/schedule/state.py | 2 +-
python/tvm/s_tir/schedule/trace.py | 2 +-
.../{tir/analysis => s_tir/transform}/__init__.py | 4 +-
python/tvm/{tir => s_tir}/transform/_ffi_api.py | 2 +-
python/tvm/s_tir/transform/transform.py | 255 +++++
python/tvm/tir/analysis/analysis.py | 2 +-
python/tvm/tir/transform/transform.py | 277 -----
src/meta_schedule/database/database.cc | 31 +-
src/meta_schedule/database/schedule_fn_database.cc | 20 +-
.../feature_extractor/per_store_feature.cc | 52 +-
.../mutator/mutate_compute_location.cc | 22 +-
src/meta_schedule/mutator/mutate_parallel.cc | 39 +-
src/meta_schedule/mutator/mutate_thread_binding.cc | 22 +-
src/meta_schedule/mutator/mutate_tile_size.cc | 20 +-
src/meta_schedule/mutator/mutate_unroll.cc | 21 +-
src/meta_schedule/mutator/mutator.cc | 6 +-
.../postproc/disallow_async_strided_mem_copy.cc | 28 +-
.../postproc/disallow_dynamic_loop.cc | 9 +-
src/meta_schedule/postproc/postproc.cc | 2 +-
.../postproc/rewrite_cooperative_fetch.cc | 53 +-
src/meta_schedule/postproc/rewrite_layout.cc | 13 +-
.../postproc/rewrite_parallel_vectorize_unroll.cc | 52 +-
.../postproc/rewrite_reduction_block.cc | 23 +-
src/meta_schedule/postproc/rewrite_tensorize.cc | 20 +-
.../postproc/rewrite_unbound_block.cc | 19 +-
src/meta_schedule/postproc/verify_gpu_code.cc | 44 +-
src/meta_schedule/postproc/verify_vtcm_limit.cc | 2 +-
src/meta_schedule/schedule/cpu/winograd.cc | 9 +-
src/meta_schedule/schedule/cuda/thread_bind.cc | 22 +-
src/meta_schedule/schedule/cuda/winograd.cc | 9 +-
src/meta_schedule/schedule/generic/winograd.cc | 4 +
src/meta_schedule/schedule_rule/add_rfactor.cc | 24 +-
.../schedule_rule/apply_custom_rule.cc | 9 +-
src/meta_schedule/schedule_rule/auto_bind.cc | 7 +-
src/meta_schedule/schedule_rule/auto_inline.cc | 38 +-
.../schedule_rule/cross_thread_reduction.cc | 74 +-
.../schedule_rule/multi_level_tiling.cc | 36 +-
.../schedule_rule/multi_level_tiling.h | 32 +-
.../multi_level_tiling_tensor_core.cc | 96 +-
.../multi_level_tiling_wide_vector.cc | 35 +-
.../multi_level_tiling_with_intrin.cc | 15 +-
.../schedule_rule/parallel_vectorize_unroll.cc | 11 +-
.../schedule_rule/random_compute_location.cc | 29 +-
src/meta_schedule/schedule_rule/schedule_rule.cc | 4 +-
.../search_strategy/evolutionary_search.cc | 22 +-
src/meta_schedule/search_strategy/replay_func.cc | 8 +-
src/meta_schedule/search_strategy/replay_trace.cc | 20 +-
.../search_strategy/search_strategy.cc | 9 +-
.../space_generator/post_order_apply.cc | 22 +-
src/meta_schedule/space_generator/schedule_fn.cc | 12 +-
.../space_generator/space_generator.cc | 2 +-
.../space_generator/space_generator_union.cc | 6 +-
src/meta_schedule/task_scheduler/gradient_based.cc | 2 +-
src/meta_schedule/task_scheduler/task_scheduler.cc | 8 +-
src/meta_schedule/trace_apply.cc | 14 +-
src/meta_schedule/trace_apply.h | 6 +-
src/meta_schedule/utils.h | 38 +-
.../distributed/transform/legalize_redistribute.cc | 2 +-
src/relax/distributed/transform/lower_distir.cc | 2 +-
.../transform/lower_global_view_to_local_view.cc | 3 +-
src/relax/transform/meta_schedule.cc | 10 +-
src/relax/transform/split_call_tir_by_pattern.cc | 4 +-
src/{tir => s_tir}/schedule/analysis.h | 29 +-
src/{tir => s_tir}/schedule/analysis/analysis.cc | 56 +-
src/{tir => s_tir}/schedule/analysis/layout.cc | 7 +-
src/{tir => s_tir}/schedule/analysis/reducer.cc | 5 +-
src/{tir => s_tir}/schedule/analysis/verify.cc | 5 +-
src/{tir => s_tir}/schedule/concrete_schedule.cc | 156 +--
src/{tir => s_tir}/schedule/concrete_schedule.h | 11 +-
src/{tir => s_tir}/schedule/error.cc | 5 +-
src/{tir => s_tir}/schedule/error.h | 13 +-
src/{tir => s_tir}/schedule/instruction.cc | 9 +-
src/{tir => s_tir}/schedule/instruction_traits.h | 17 +-
src/{tir => s_tir}/schedule/ir_comparator.cc | 5 +-
src/{tir => s_tir}/schedule/ir_comparator.h | 11 +-
src/{tir => s_tir}/schedule/primitive.h | 17 +-
src/{tir => s_tir}/schedule/primitive/annotate.cc | 9 +-
.../schedule/primitive/annotate_buffer_access.cc | 11 +-
.../schedule/primitive/block_annotate.cc | 22 +-
.../schedule/primitive/blockize_tensorize.cc | 25 +-
.../schedule/primitive/cache_index.cc | 11 +-
.../schedule/primitive/cache_read_write.cc | 27 +-
.../schedule/primitive/compute_at.cc | 9 +-
.../schedule/primitive/compute_inline.cc | 11 +-
.../schedule/primitive/decompose_padding.cc | 11 +-
src/{tir => s_tir}/schedule/primitive/for_kind.cc | 13 +-
.../schedule/primitive/get_block_loop.cc | 23 +-
.../schedule/primitive/hide_buffer_access.cc | 9 +-
.../schedule/primitive/layout_transformation.cc | 11 +-
.../schedule/primitive/loop_transformation.cc | 21 +-
.../schedule/primitive/pad_einsum.cc | 7 +-
.../schedule/primitive/read_write_at.cc | 9 +-
src/{tir => s_tir}/schedule/primitive/reduction.cc | 11 +-
.../schedule/primitive/reorder_block_iter_var.cc | 7 +-
.../schedule/primitive/rolling_buffer.cc | 9 +-
src/{tir => s_tir}/schedule/primitive/sampling.cc | 15 +-
src/{tir => s_tir}/schedule/schedule.cc | 146 +--
src/{tir => s_tir}/schedule/state.cc | 15 +-
src/{tir => s_tir}/schedule/trace.cc | 27 +-
src/{tir => s_tir}/schedule/traced_schedule.cc | 17 +-
src/{tir => s_tir}/schedule/traced_schedule.h | 11 +-
src/{tir => s_tir}/schedule/transform.cc | 18 +-
src/{tir => s_tir}/schedule/transform.h | 25 +-
src/{tir => s_tir}/schedule/utils.h | 32 +-
.../transform}/annotate_irregular_loop.cc | 18 +-
.../transform}/canonicalize_loop.cc | 31 +-
.../transform}/compact_buffer_region.cc | 33 +-
.../transform}/convert_blocks_to_opaque.cc | 23 +-
.../transform}/inject_double_buffer.cc | 27 +-
.../transform}/inject_permuted_layout.cc | 13 +-
.../transform}/inject_software_pipeline.cc | 30 +-
.../transform}/inject_virtual_thread.cc | 18 +-
.../transform}/lift_thread_binding.cc | 23 +-
.../transform}/loop_partition.cc | 37 +-
.../transform}/lower_cross_thread_reduction.cc | 25 +-
.../transform}/lower_init_block.cc | 23 +-
.../transform}/lower_match_buffer.cc | 25 +-
.../transform}/lower_opaque_block.cc | 33 +-
.../manifest_shared_memory_local_stage.cc | 15 +-
.../transform}/memhammer_coalesce.cc | 5 +-
.../transform}/memhammer_intermediate_stage.cc | 5 +-
.../transform}/memhammer_lower_auto_copy.cc | 13 +-
.../transform}/memhammer_rewrite_rule.h | 13 +-
.../transform}/memhammer_tensorcore_rewrite.cc | 5 +-
.../plan_update_buffer_allocation_location.cc | 25 +-
.../transform}/transform_mma_buffer_layout.cc | 13 +-
.../transform}/unify_thread_binding.cc | 25 +-
src/tir/analysis/calculate_allocated_memory.cc | 15 +-
src/tir/analysis/oob_checker.cc | 4 +-
src/tir/transforms/default_gpu_schedule.cc | 22 +-
tests/python/codegen/test_gpu_codegen_allreduce.py | 100 +-
tests/python/codegen/test_target_codegen.py | 2 +-
.../python/codegen/test_target_codegen_aarch64.py | 604 ++++++-----
tests/python/codegen/test_target_codegen_arm.py | 87 +-
tests/python/codegen/test_target_codegen_bool.py | 92 +-
tests/python/codegen/test_target_codegen_c_host.py | 159 +--
.../codegen/test_target_codegen_cross_llvm.py | 39 +-
tests/python/codegen/test_target_codegen_cuda.py | 576 ++++++----
.../python/codegen/test_target_codegen_cuda_fp4.py | 178 +--
.../python/codegen/test_target_codegen_cuda_fp8.py | 263 +++--
tests/python/codegen/test_target_codegen_device.py | 101 +-
.../codegen/test_target_codegen_gpu_common.py | 37 +-
.../python/codegen/test_target_codegen_hexagon.py | 70 +-
tests/python/codegen/test_target_codegen_llvm.py | 1136 +++++++++++---------
tests/python/codegen/test_target_codegen_metal.py | 59 +-
tests/python/codegen/test_target_codegen_opencl.py | 225 ++--
tests/python/codegen/test_target_codegen_rocm.py | 65 +-
tests/python/codegen/test_target_codegen_vulkan.py | 382 +++----
tests/python/codegen/test_target_codegen_x86.py | 26 +-
tests/python/dlight/test_primitives.py | 2 +-
tests/python/relax/test_frontend_dynamo.py | 19 +
...est_s_tir_transform_annotate_irregular_loop.py} | 20 +-
.../test_s_tir_transform_canonicalize_loop.py} | 22 +-
.../test_s_tir_transform_compact_buffer_region.py} | 16 +-
...st_s_tir_transform_convert_blocks_to_opaque.py} | 6 +-
.../test_s_tir_transform_inject_double_buffer.py} | 8 +-
...test_s_tir_transform_inject_permuted_layout.py} | 3 +-
...st_s_tir_transform_inject_software_pipeline.py} | 32 +-
.../test_s_tir_transform_inject_virtual_thread.py} | 10 +-
.../test_s_tir_transform_lift_thread_binding.py} | 4 +-
.../test_s_tir_transform_loop_partition.py} | 44 +-
..._tir_transform_lower_cross_thread_reduction.py} | 6 +-
.../test_s_tir_transform_lower_init_block.py} | 6 +-
.../test_s_tir_transform_lower_match_buffer.py} | 5 +-
.../test_s_tir_transform_lower_opaque_block.py} | 9 +-
...ransform_manifest_shared_memory_local_stage.py} | 3 +-
...t_s_tir_transform_memhammer_lower_auto_copy.py} | 6 +-
...form_plan_update_buffer_allocation_location.py} | 4 +-
.../test_s_tir_transform_unify_thread_binding.py} | 6 +-
...test_tir_analysis_calculate_allocated_memory.py | 14 +-
.../test_tir_transform_inject_ptx_async_copy.py | 6 +-
tests/python/tvmscript/test_tvmscript_roundtrip.py | 2 +-
202 files changed, 4480 insertions(+), 3866 deletions(-)
rename include/tvm/{tir => s_tir}/schedule/instruction.h (94%)
rename include/tvm/{tir => s_tir}/schedule/schedule.h (98%)
rename include/tvm/{tir => s_tir}/schedule/state.h (96%)
rename include/tvm/{tir => s_tir}/schedule/trace.h (95%)
create mode 100644 include/tvm/s_tir/transform.h
copy python/tvm/{tir/analysis => s_tir/transform}/__init__.py (92%)
copy python/tvm/{tir => s_tir}/transform/_ffi_api.py (94%)
create mode 100644 python/tvm/s_tir/transform/transform.py
rename src/{tir => s_tir}/schedule/analysis.h (98%)
rename src/{tir => s_tir}/schedule/analysis/analysis.cc (97%)
rename src/{tir => s_tir}/schedule/analysis/layout.cc (98%)
rename src/{tir => s_tir}/schedule/analysis/reducer.cc (99%)
rename src/{tir => s_tir}/schedule/analysis/verify.cc (99%)
rename src/{tir => s_tir}/schedule/concrete_schedule.cc (87%)
rename src/{tir => s_tir}/schedule/concrete_schedule.h (98%)
rename src/{tir => s_tir}/schedule/error.cc (96%)
rename src/{tir => s_tir}/schedule/error.h (93%)
rename src/{tir => s_tir}/schedule/instruction.cc (96%)
rename src/{tir => s_tir}/schedule/instruction_traits.h (98%)
rename src/{tir => s_tir}/schedule/ir_comparator.cc (99%)
rename src/{tir => s_tir}/schedule/ir_comparator.h (97%)
rename src/{tir => s_tir}/schedule/primitive.h (99%)
rename src/{tir => s_tir}/schedule/primitive/annotate.cc (97%)
rename src/{tir => s_tir}/schedule/primitive/annotate_buffer_access.cc (96%)
rename src/{tir => s_tir}/schedule/primitive/block_annotate.cc (96%)
rename src/{tir => s_tir}/schedule/primitive/blockize_tensorize.cc (98%)
rename src/{tir => s_tir}/schedule/primitive/cache_index.cc (98%)
rename src/{tir => s_tir}/schedule/primitive/cache_read_write.cc (99%)
rename src/{tir => s_tir}/schedule/primitive/compute_at.cc (99%)
rename src/{tir => s_tir}/schedule/primitive/compute_inline.cc (99%)
rename src/{tir => s_tir}/schedule/primitive/decompose_padding.cc (98%)
rename src/{tir => s_tir}/schedule/primitive/for_kind.cc (97%)
rename src/{tir => s_tir}/schedule/primitive/get_block_loop.cc (93%)
rename src/{tir => s_tir}/schedule/primitive/hide_buffer_access.cc (97%)
rename src/{tir => s_tir}/schedule/primitive/layout_transformation.cc (99%)
rename src/{tir => s_tir}/schedule/primitive/loop_transformation.cc (99%)
rename src/{tir => s_tir}/schedule/primitive/pad_einsum.cc (99%)
rename src/{tir => s_tir}/schedule/primitive/read_write_at.cc (99%)
rename src/{tir => s_tir}/schedule/primitive/reduction.cc (99%)
rename src/{tir => s_tir}/schedule/primitive/reorder_block_iter_var.cc (97%)
rename src/{tir => s_tir}/schedule/primitive/rolling_buffer.cc (98%)
rename src/{tir => s_tir}/schedule/primitive/sampling.cc (98%)
rename src/{tir => s_tir}/schedule/schedule.cc (67%)
rename src/{tir => s_tir}/schedule/state.cc (99%)
rename src/{tir => s_tir}/schedule/trace.cc (96%)
rename src/{tir => s_tir}/schedule/traced_schedule.cc (98%)
rename src/{tir => s_tir}/schedule/traced_schedule.h (97%)
rename src/{tir => s_tir}/schedule/transform.cc (97%)
rename src/{tir => s_tir}/schedule/transform.h (93%)
rename src/{tir => s_tir}/schedule/utils.h (95%)
rename src/{tir/transforms => s_tir/transform}/annotate_irregular_loop.cc (82%)
rename src/{tir/transforms => s_tir/transform}/canonicalize_loop.cc (80%)
rename src/{tir/transforms => s_tir/transform}/compact_buffer_region.cc (96%)
rename src/{tir/transforms => s_tir/transform}/convert_blocks_to_opaque.cc
(90%)
rename src/{tir/transforms => s_tir/transform}/inject_double_buffer.cc (93%)
rename src/{tir/transforms => s_tir/transform}/inject_permuted_layout.cc (97%)
rename src/{tir/transforms => s_tir/transform}/inject_software_pipeline.cc
(98%)
rename src/{tir/transforms => s_tir/transform}/inject_virtual_thread.cc (97%)
rename src/{tir/transforms => s_tir/transform}/lift_thread_binding.cc (93%)
rename src/{tir/transforms => s_tir/transform}/loop_partition.cc (96%)
rename src/{tir/transforms => s_tir/transform}/lower_cross_thread_reduction.cc
(98%)
rename src/{tir/transforms => s_tir/transform}/lower_init_block.cc (84%)
rename src/{tir/transforms => s_tir/transform}/lower_match_buffer.cc (95%)
rename src/{tir/transforms => s_tir/transform}/lower_opaque_block.cc (91%)
rename src/{tir/transforms =>
s_tir/transform}/manifest_shared_memory_local_stage.cc (96%)
rename src/{tir/transforms => s_tir/transform}/memhammer_coalesce.cc (99%)
rename src/{tir/transforms => s_tir/transform}/memhammer_intermediate_stage.cc
(99%)
rename src/{tir/transforms => s_tir/transform}/memhammer_lower_auto_copy.cc
(99%)
rename src/{tir/transforms => s_tir/transform}/memhammer_rewrite_rule.h (97%)
rename src/{tir/transforms => s_tir/transform}/memhammer_tensorcore_rewrite.cc
(99%)
rename src/{tir/transforms =>
s_tir/transform}/plan_update_buffer_allocation_location.cc (94%)
rename src/{tir/transforms => s_tir/transform}/transform_mma_buffer_layout.cc
(96%)
rename src/{tir/transforms => s_tir/transform}/unify_thread_binding.cc (93%)
rename
tests/python/{tir-transform/test_tir_transform_annotate_irregular_loop.py =>
s_tir/transform/test_s_tir_transform_annotate_irregular_loop.py} (93%)
rename tests/python/{tir-transform/test_tir_transform_canonicalize_loop.py =>
s_tir/transform/test_s_tir_transform_canonicalize_loop.py} (76%)
rename tests/python/{tir-transform/test_tir_transform_compact_buffer_region.py
=> s_tir/transform/test_s_tir_transform_compact_buffer_region.py} (99%)
rename
tests/python/{tir-transform/test_tir_transform_convert_blocks_to_opaque.py =>
s_tir/transform/test_s_tir_transform_convert_blocks_to_opaque.py} (95%)
rename tests/python/{tir-transform/test_tir_transform_inject_double_buffer.py
=> s_tir/transform/test_s_tir_transform_inject_double_buffer.py} (95%)
rename
tests/python/{tir-transform/test_tir_transform_inject_permuted_layout.py =>
s_tir/transform/test_s_tir_transform_inject_permuted_layout.py} (99%)
rename
tests/python/{tir-transform/test_tir_transform_inject_software_pipeline.py =>
s_tir/transform/test_s_tir_transform_inject_software_pipeline.py} (98%)
rename tests/python/{tir-transform/test_tir_transform_inject_virtual_thread.py
=> s_tir/transform/test_s_tir_transform_inject_virtual_thread.py} (95%)
rename tests/python/{tir-transform/test_tir_transform_lift_thread_binding.py
=> s_tir/transform/test_s_tir_transform_lift_thread_binding.py} (99%)
rename tests/python/{tir-transform/test_tir_transform_loop_partition.py =>
s_tir/transform/test_s_tir_transform_loop_partition.py} (95%)
rename
tests/python/{tir-transform/test_tir_transform_lower_cross_thread_reduction.py
=> s_tir/transform/test_s_tir_transform_lower_cross_thread_reduction.py} (99%)
rename tests/python/{tir-transform/test_tir_transform_lower_init_block.py =>
s_tir/transform/test_s_tir_transform_lower_init_block.py} (96%)
rename tests/python/{tir-transform/test_tir_transform_lower_match_buffer.py =>
s_tir/transform/test_s_tir_transform_lower_match_buffer.py} (99%)
rename tests/python/{tir-transform/test_tir_transform_lower_opaque_block.py =>
s_tir/transform/test_s_tir_transform_lower_opaque_block.py} (98%)
rename
tests/python/{tir-transform/test_tir_transform_manifest_shared_memory_local_stage.py
=> s_tir/transform/test_s_tir_transform_manifest_shared_memory_local_stage.py}
(99%)
rename
tests/python/{tir-transform/test_tir_transform_memhammer_lower_auto_copy.py =>
s_tir/transform/test_s_tir_transform_memhammer_lower_auto_copy.py} (99%)
rename
tests/python/{tir-transform/test_tir_transform_plan_update_buffer_allocation_location.py
=>
s_tir/transform/test_s_tir_transform_plan_update_buffer_allocation_location.py}
(99%)
rename tests/python/{tir-transform/test_tir_transform_unify_thread_binding.py
=> s_tir/transform/test_s_tir_transform_unify_thread_binding.py} (98%)
