nvidia-cutlass-dsl-libs-base 4.9.0.dev1__py3-none-any.whl
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- nvidia_cutlass_dsl/dsl_packages/cutlass/_mlir/_mlir_libs/__init__.py +324 -0
- nvidia_cutlass_dsl/dsl_packages/cutlass/_mlir/_mlir_libs/_cutlass_ir/_mlir/py.typed +0 -0
- nvidia_cutlass_dsl/dsl_packages/cutlass/_mlir/_mlir_libs/_cutlass_ir/_mlirExecutionEngine.pyi +19 -0
- nvidia_cutlass_dsl/dsl_packages/cutlass/_mlir/dialects/_arith_enum_gen.py +293 -0
- nvidia_cutlass_dsl/dsl_packages/cutlass/_mlir/dialects/_arith_ops_gen.py +4031 -0
- nvidia_cutlass_dsl/dsl_packages/cutlass/_mlir/dialects/_builtin_ops_gen.py +199 -0
- nvidia_cutlass_dsl/dsl_packages/cutlass/_mlir/dialects/_cf_ops_gen.py +400 -0
- nvidia_cutlass_dsl/dsl_packages/cutlass/_mlir/dialects/_core_enum_gen.py +567 -0
- nvidia_cutlass_dsl/dsl_packages/cutlass/_mlir/dialects/_core_ops_gen.py +6378 -0
- nvidia_cutlass_dsl/dsl_packages/cutlass/_mlir/dialects/_cuda_enum_gen.py +900 -0
- nvidia_cutlass_dsl/dsl_packages/cutlass/_mlir/dialects/_cuda_ops_gen.py +16207 -0
- nvidia_cutlass_dsl/dsl_packages/cutlass/_mlir/dialects/_cute_enum_gen.py +143 -0
- nvidia_cutlass_dsl/dsl_packages/cutlass/_mlir/dialects/_cute_nvgpu_enum_gen.py +1536 -0
- nvidia_cutlass_dsl/dsl_packages/cutlass/_mlir/dialects/_cute_nvgpu_ops_gen.py +6865 -0
- nvidia_cutlass_dsl/dsl_packages/cutlass/_mlir/dialects/_cute_ops_gen.py +8755 -0
- nvidia_cutlass_dsl/dsl_packages/cutlass/_mlir/dialects/_func_ops_gen.py +600 -0
- nvidia_cutlass_dsl/dsl_packages/cutlass/_mlir/dialects/_gpu_enum_gen.py +440 -0
- nvidia_cutlass_dsl/dsl_packages/cutlass/_mlir/dialects/_gpu_ops_gen.py +7183 -0
- nvidia_cutlass_dsl/dsl_packages/cutlass/_mlir/dialects/_iket_enum_gen.py +8 -0
- nvidia_cutlass_dsl/dsl_packages/cutlass/_mlir/dialects/_iket_ops_gen.py +454 -0
- nvidia_cutlass_dsl/dsl_packages/cutlass/_mlir/dialects/_lir_enum_gen.py +96 -0
- nvidia_cutlass_dsl/dsl_packages/cutlass/_mlir/dialects/_lir_ops_gen.py +6817 -0
- nvidia_cutlass_dsl/dsl_packages/cutlass/_mlir/dialects/_llvm_enum_gen.py +1098 -0
- nvidia_cutlass_dsl/dsl_packages/cutlass/_mlir/dialects/_llvm_ops_gen.py +21996 -0
- nvidia_cutlass_dsl/dsl_packages/cutlass/_mlir/dialects/_math_ops_gen.py +3115 -0
- nvidia_cutlass_dsl/dsl_packages/cutlass/_mlir/dialects/_nvgpu_enum_gen.py +257 -0
- nvidia_cutlass_dsl/dsl_packages/cutlass/_mlir/dialects/_nvgpu_ops_gen.py +2914 -0
- nvidia_cutlass_dsl/dsl_packages/cutlass/_mlir/dialects/_nvvm_enum_gen.py +2828 -0
- nvidia_cutlass_dsl/dsl_packages/cutlass/_mlir/dialects/_nvvm_ops_gen.py +25559 -0
- nvidia_cutlass_dsl/dsl_packages/cutlass/_mlir/dialects/_ods_common.py +274 -0
- nvidia_cutlass_dsl/dsl_packages/cutlass/_mlir/dialects/_pyir_ops_gen.py +174 -0
- nvidia_cutlass_dsl/dsl_packages/cutlass/_mlir/dialects/_scf_ops_gen.py +1404 -0
- nvidia_cutlass_dsl/dsl_packages/cutlass/_mlir/dialects/_ts_enum_gen.py +257 -0
- nvidia_cutlass_dsl/dsl_packages/cutlass/_mlir/dialects/_ts_ops_gen.py +1650 -0
- nvidia_cutlass_dsl/dsl_packages/cutlass/_mlir/dialects/_ub_ops_gen.py +123 -0
- nvidia_cutlass_dsl/dsl_packages/cutlass/_mlir/dialects/_vector_enum_gen.py +119 -0
- nvidia_cutlass_dsl/dsl_packages/cutlass/_mlir/dialects/_vector_ops_gen.py +4221 -0
- nvidia_cutlass_dsl/dsl_packages/cutlass/_mlir/dialects/arith.py +88 -0
- nvidia_cutlass_dsl/dsl_packages/cutlass/_mlir/dialects/builtin.py +46 -0
- nvidia_cutlass_dsl/dsl_packages/cutlass/_mlir/dialects/cf.py +6 -0
- nvidia_cutlass_dsl/dsl_packages/cutlass/_mlir/dialects/core.py +35 -0
- nvidia_cutlass_dsl/dsl_packages/cutlass/_mlir/dialects/cuda.py +276 -0
- nvidia_cutlass_dsl/dsl_packages/cutlass/_mlir/dialects/cute.py +16 -0
- nvidia_cutlass_dsl/dsl_packages/cutlass/_mlir/dialects/cute_nvgpu.py +15 -0
- nvidia_cutlass_dsl/dsl_packages/cutlass/_mlir/dialects/ext.py +961 -0
- nvidia_cutlass_dsl/dsl_packages/cutlass/_mlir/dialects/func.py +321 -0
- nvidia_cutlass_dsl/dsl_packages/cutlass/_mlir/dialects/gpu/__init__.py +337 -0
- nvidia_cutlass_dsl/dsl_packages/cutlass/_mlir/dialects/gpu/passes/__init__.py +5 -0
- nvidia_cutlass_dsl/dsl_packages/cutlass/_mlir/dialects/iket.py +13 -0
- nvidia_cutlass_dsl/dsl_packages/cutlass/_mlir/dialects/lir.py +14 -0
- nvidia_cutlass_dsl/dsl_packages/cutlass/_mlir/dialects/llvm.py +26 -0
- nvidia_cutlass_dsl/dsl_packages/cutlass/_mlir/dialects/math.py +6 -0
- nvidia_cutlass_dsl/dsl_packages/cutlass/_mlir/dialects/nvgpu.py +8 -0
- nvidia_cutlass_dsl/dsl_packages/cutlass/_mlir/dialects/nvvm.py +7 -0
- nvidia_cutlass_dsl/dsl_packages/cutlass/_mlir/dialects/pyir.py +17 -0
- nvidia_cutlass_dsl/dsl_packages/cutlass/_mlir/dialects/scf.py +324 -0
- nvidia_cutlass_dsl/dsl_packages/cutlass/_mlir/dialects/ts.py +64 -0
- nvidia_cutlass_dsl/dsl_packages/cutlass/_mlir/dialects/ub.py +6 -0
- nvidia_cutlass_dsl/dsl_packages/cutlass/_mlir/dialects/vector.py +7 -0
- nvidia_cutlass_dsl/dsl_packages/cutlass/_mlir/execution_engine.py +41 -0
- nvidia_cutlass_dsl/dsl_packages/cutlass/_mlir/extras/meta.py +81 -0
- nvidia_cutlass_dsl/dsl_packages/cutlass/_mlir/extras/types.py +170 -0
- nvidia_cutlass_dsl/dsl_packages/cutlass/_mlir/ir.py +386 -0
- nvidia_cutlass_dsl/dsl_packages/cutlass/_mlir/passmanager.py +5 -0
- nvidia_cutlass_dsl/dsl_packages/cutlass/_mlir/rewrite.py +5 -0
- nvidia_cutlass_dsl/dsl_packages/cutlass/_mlir/runtime/__init__.py +1 -0
- nvidia_cutlass_dsl/dsl_packages/cutlass/_mlir/runtime/np_to_memref.py +181 -0
- nvidia_cutlass_dsl/dsl_packages/cutlass/_mlir_helpers/__init__.py +29 -0
- nvidia_cutlass_dsl/dsl_packages/cutlass/_mlir_helpers/arith.py +1221 -0
- nvidia_cutlass_dsl/dsl_packages/cutlass/_mlir_helpers/cuda.py +99 -0
- nvidia_cutlass_dsl/dsl_packages/cutlass/_mlir_helpers/dialect_proxy.py +121 -0
- nvidia_cutlass_dsl/dsl_packages/cutlass/_mlir_helpers/dominance.py +75 -0
- nvidia_cutlass_dsl/dsl_packages/cutlass/_mlir_helpers/gpu.py +63 -0
- nvidia_cutlass_dsl/dsl_packages/cutlass/_mlir_helpers/lru_cache_ir.py +79 -0
- nvidia_cutlass_dsl/dsl_packages/cutlass/_mlir_helpers/math.py +2514 -0
- nvidia_cutlass_dsl/dsl_packages/cutlass/_mlir_helpers/nanobind_compat.py +71 -0
- nvidia_cutlass_dsl/dsl_packages/cutlass/_mlir_helpers/op.py +810 -0
- nvidia_cutlass_dsl/dsl_packages/cutlass/_mlir_helpers/vector.py +1470 -0
- nvidia_cutlass_dsl_libs_base-4.9.0.dev1.dist-info/METADATA +46 -0
- nvidia_cutlass_dsl_libs_base-4.9.0.dev1.dist-info/RECORD +83 -0
- nvidia_cutlass_dsl_libs_base-4.9.0.dev1.dist-info/WHEEL +5 -0
- nvidia_cutlass_dsl_libs_base-4.9.0.dev1.dist-info/licenses/LICENSE +188 -0
- nvidia_cutlass_dsl_libs_base-4.9.0.dev1.dist-info/top_level.txt +1 -0
|
@@ -0,0 +1,71 @@
|
|
|
1
|
+
# SPDX-FileCopyrightText: Copyright (c) 2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
|
|
2
|
+
# SPDX-License-Identifier: LicenseRef-NvidiaProprietary
|
|
3
|
+
#
|
|
4
|
+
# Use of this software is governed by the terms and conditions of the
|
|
5
|
+
# NVIDIA End User License Agreement (EULA), available at:
|
|
6
|
+
# https://docs.nvidia.com/cutlass/latest/media/docs/pythonDSL/license.html
|
|
7
|
+
#
|
|
8
|
+
# Any use, reproduction, disclosure, or distribution of this software
|
|
9
|
+
# and related documentation outside the scope permitted by the EULA
|
|
10
|
+
# is strictly prohibited.
|
|
11
|
+
|
|
12
|
+
"""
|
|
13
|
+
Compatibility patches for nanobind-based MLIR Python bindings.
|
|
14
|
+
|
|
15
|
+
The LLVM prebuilt is compiled with -fno-exceptions. Several nanobind
|
|
16
|
+
binding methods signal errors or end-of-iteration by throwing C++
|
|
17
|
+
exceptions (nb::stop_iteration, nb::value_error, ...). With
|
|
18
|
+
-fno-exceptions the throw cannot unwind and calls std::terminate/abort.
|
|
19
|
+
The patches below replace those code paths with alternatives
|
|
20
|
+
that never reach the throwing C++ code.
|
|
21
|
+
"""
|
|
22
|
+
|
|
23
|
+
from threading import local
|
|
24
|
+
from typing import Any, Iterator
|
|
25
|
+
|
|
26
|
+
from .._mlir import ir
|
|
27
|
+
|
|
28
|
+
_thread_local = local()
|
|
29
|
+
|
|
30
|
+
|
|
31
|
+
# -- InsertionPoint.current --------------------------------------------------
|
|
32
|
+
# ir.InsertionPoint.current throws nb::value_error when no insertion
|
|
33
|
+
# point is active. Track enter/exit at the Python level so callers can
|
|
34
|
+
# check before touching the C++ property.
|
|
35
|
+
|
|
36
|
+
|
|
37
|
+
def has_insertion_point() -> bool:
|
|
38
|
+
return getattr(_thread_local, "ip_depth", 0) > 0
|
|
39
|
+
|
|
40
|
+
|
|
41
|
+
_orig_ip_enter = ir.InsertionPoint.__enter__
|
|
42
|
+
_orig_ip_exit = ir.InsertionPoint.__exit__
|
|
43
|
+
|
|
44
|
+
|
|
45
|
+
def _tracked_ip_enter(self: ir.InsertionPoint) -> ir.InsertionPoint:
|
|
46
|
+
_thread_local.ip_depth = getattr(_thread_local, "ip_depth", 0) + 1
|
|
47
|
+
return _orig_ip_enter(self)
|
|
48
|
+
|
|
49
|
+
|
|
50
|
+
def _tracked_ip_exit(self: ir.InsertionPoint, *args: Any) -> None:
|
|
51
|
+
_thread_local.ip_depth = getattr(_thread_local, "ip_depth", 1) - 1
|
|
52
|
+
_orig_ip_exit(self, *args)
|
|
53
|
+
|
|
54
|
+
|
|
55
|
+
ir.InsertionPoint.__enter__ = _tracked_ip_enter
|
|
56
|
+
ir.InsertionPoint.__exit__ = _tracked_ip_exit
|
|
57
|
+
|
|
58
|
+
|
|
59
|
+
# -- Block.__iter__ -----------------------------------------------------------
|
|
60
|
+
# PyOperationIterator.__next__ throws nb::stop_iteration() at
|
|
61
|
+
# end-of-block. Replace with indexed access via Block.operations
|
|
62
|
+
# whose __len__ and __getitem__ use the C API without throwing.
|
|
63
|
+
|
|
64
|
+
|
|
65
|
+
def _safe_block_iter(self: ir.Block) -> Iterator[Any]:
|
|
66
|
+
ops = self.operations
|
|
67
|
+
for i in range(len(ops)):
|
|
68
|
+
yield ops[i]
|
|
69
|
+
|
|
70
|
+
|
|
71
|
+
ir.Block.__iter__ = _safe_block_iter
|