[v2.9.0][bugfix]cann and pta header mixing bulid bugfix-pytorch-AtomGit

已合并

[v2.9.0][bugfix]cann and pta header mixing bulid bugfix #44992

Dring创建于 17 天前从Dring/pytorch_3： v2.9.0 合入到 Ascend/pytorch： v2.9.0

[v2.9.0][bugfix]cann and pta header mixing bulid bugfix #44992

已合并

Dring创建于 17 天前

从Dring/pytorch_3 v2.9.0合入到Ascend/pytorch v2.9.0

收起

build_libtorch_npu.py
ci
- access_control_test.py
setup.py
test
- allocator
  - test_pluggable_allocator_extensions.py
- cpp_extensions
  - pluggable_allocator_extensions.cpp
- distributed / rpc
  - test_single_machine_multi_devices.py
- npu
  - test_allocator_trace_tracker.py
- test_npu_expandable_segments.py
- test_sanitizer_pluggable_allocator.py
- trans_contiguous
  - test_single_reshape_copy_to_contiguous.py
third_party
- acl / libs
  - build_stub.sh
- hccl / inc / hccl
  - hccl.h
torch_npu
- _inductor
  - ascend_npu_ir / ascend_npu_ir / npu
    - utils.py
  - cpp_builder.py
- csrc
  - aten / common
    - CopyMemoryKernel.cpp
    - FormatCastHelper.h
    - LocalScalarDenseNpu.cpp
    - TensorShape.cpp
  - core
    - NPUSerialization.cpp
    - NPUStorageImpl.h
    - NPUTensorImpl.cpp
    - npu
      - CachingHostAllocator.h
      - GetCANNInfo.cpp
      - NPUCachingAllocator.cpp
      - NPUEvent.h
      - NPUEventManager.h
      - NPUException.h
      - NPUFunctions.cpp
      - NPUFunctions.h
      - NPUGraph.cpp
      - NPUGraph.h
      - NPUPeerToPeerAccess.cpp
      - NPUQueue.cpp
      - NPUQueue.h
      - NPUStream.cpp
      - NPUStream.h
      - NPUSwappedMemoryAllocator.cpp
      - NPUWorkspaceAllocator.cpp
      - impl
        NPUGuardImpl.h
      - interface
        AclInterface.h
        AsyncTaskQueueInterface.cpp
        AsyncTaskQueueInterface.h
        MlInterface.h
        SkInterface.h
      - npu_log.h
      - sys_ctrl
        npu_sys_ctrl.cpp
        npu_sys_ctrl.h
  - custom_dtype
    - Init.h
  - distributed
    - ProcessGroupHCCL.cpp
    - rpc
      - tensorpipe_agent.cpp
    - symm_mem
      - NPUSHMEMInterface.h
  - framework
    - InferFormat.h
    - LazyInitAclops.h
    - OpCmdHelper.h
    - OpParamMaker.h
    - aoe
      - AoeDumpGraphManager.h
    - contiguous
      - ContiguousUtils.h
    - interface
      - AclInterface.h
      - AclOpCompileInterface.cpp
      - AclOpCompileInterface.h
      - EnvVariables.cpp
      - MsProfilerInterface.cpp
      - MsProfilerInterface.h
    - utils
      - CalcuOpUtil.cpp
      - CalcuOpUtil.h
      - NPUDefinition.h
      - NpuUtils.h
  - inductor
    - aoti_runtime
      - device_utils.h
      - utils_npu.h
    - mlir
      - cpp_common.cpp
      - cpp_common.h
  - ipc
    - NPUIPCTypes.cpp
    - StorageSharing.cpp
  - npu
    - Graph.h
    - Module.cpp
    - Stream.cpp
  - profiler
    - feature_mgr.h
    - npu_profiler.h
    - profiler_mgr.h
    - profiler_npu.cpp
- utils
  - cpp_extension.py

共 81 个文件变更+223-150

Mbuild_libtorch_npu.py+32-2

	@@ -281,8 +281,6 @@ def copy_hpp():
281	"torch_npu/csrc/inductor/*/.h",	281	"torch_npu/csrc/inductor/*/.h",
282	"torch_npu/csrc/distributed/*.h",	282	"torch_npu/csrc/distributed/*.h",
283	"torch_npu/csrc/distributed/*.hpp",	283	"torch_npu/csrc/distributed/*.hpp",
284	- "third_party/acl/inc//.h",
285	- "third_party/acl/inc///*.h",
286	"third_party/hccl/inc//.h",	284	"third_party/hccl/inc//.h",
287	]	285	]
288	glob_header_files = []	286	glob_header_files = []
	@@ -299,6 +297,38 @@ def copy_hpp():
299	os.makedirs(os.path.dirname(dst), exist_ok=True)	297	os.makedirs(os.path.dirname(dst), exist_ok=True)
300	ret.append((src, dst))	298	ret.append((src, dst))
301		299
		300	+ acl_include_root = os.path.join(BASE_DIR, "third_party", "acl", "inc")
		301	+ acl_header_files = glob.glob(
		302	+ os.path.join(acl_include_root, "*", ".h"),
		303	+ recursive=True,
		304	+ )
		305	+ for src in acl_header_files:
		306	+ relative_header = os.path.relpath(src, acl_include_root)
		307	+ dst = os.path.join(
		308	+ BASE_DIR,
		309	+ "libtorch_npu/include",
		310	+ relative_header,
		311	+ )
		312	+ os.makedirs(os.path.dirname(dst), exist_ok=True)
		313	+ ret.append((src, dst))
		314	+
		315	+ # Preserve legacy include paths with forwarding headers, not duplicate ACL headers.
		316	+ wrapper_src = os.path.join(
		317	+ BASE_DIR, "build", "acl_compat_headers", relative_header
		318	+ )
		319	+ os.makedirs(os.path.dirname(wrapper_src), exist_ok=True)
		320	+ with open(wrapper_src, "w", encoding="utf-8", newline="\n") as wrapper_file:
		321	+ wrapper_file.write("#pragma once\n#include <{}>\n".format(
		322	+ relative_header.replace(os.sep, "/")
		323	+ ))
		324	+ wrapper_dst = os.path.join(
		325	+ BASE_DIR,
		326	+ "libtorch_npu/include/third_party/acl/inc",
		327	+ relative_header,
		328	+ )
		329	+ os.makedirs(os.path.dirname(wrapper_dst), exist_ok=True)
		330	+ ret.append((wrapper_src, wrapper_dst))
		331	+
302	return ret	332	return ret
303		333
304	ret = get_src_py_and_dst()	334	ret = get_src_py_and_dst()

Mci/access_control_test.py+1-2

	@@ -55,8 +55,7 @@ def fetch_acl_headers():
55	# 2. Fallback: copy from installed torch_npu	55	# 2. Fallback: copy from installed torch_npu
56	try:	56	try:
57	import torch_npu	57	import torch_npu
58	- installed_acl = Path(	58	+ installed_acl = Path(torch_npu.__file__).resolve().parent / 'include' / 'acl'
59	- torch_npu.__file__).resolve().parent / 'include' / 'third_party' / 'acl' / 'inc' / 'acl'
60	if installed_acl.is_dir():	59	if installed_acl.is_dir():
61	acl_dest.mkdir(parents=True, exist_ok=True)	60	acl_dest.mkdir(parents=True, exist_ok=True)
62	shutil.copytree(str(installed_acl), str(acl_dest), dirs_exist_ok=True)	61	shutil.copytree(str(installed_acl), str(acl_dest), dirs_exist_ok=True)

Msetup.py+32-2

	@@ -473,9 +473,7 @@ def get_src_py_and_dst():
473	"torch_npu/csrc///*.h",	473	"torch_npu/csrc///*.h",
474	"torch_npu/csrc////.h",	474	"torch_npu/csrc////.h",
475	"torch_npu/csrc/////*.h",	475	"torch_npu/csrc/////*.h",
476	- "third_party/acl/inc//.h",
477	"third_party/hccl/inc//.h",	476	"third_party/hccl/inc//.h",
478	- "third_party/acl/inc///*.h",
479	"torch_npu/csrc/distributed/HCCLUtils.hpp",	477	"torch_npu/csrc/distributed/HCCLUtils.hpp",
480	"torch_npu/csrc/distributed/ProcessGroupHCCL.hpp"	478	"torch_npu/csrc/distributed/ProcessGroupHCCL.hpp"
481	]	479	]
	@@ -491,6 +489,38 @@ def get_src_py_and_dst():
491	os.makedirs(os.path.dirname(dst), exist_ok=True)	489	os.makedirs(os.path.dirname(dst), exist_ok=True)
492	ret.append((src, dst))	490	ret.append((src, dst))
493		491
		492	+ acl_include_root = os.path.join(BASE_DIR, "third_party", "acl", "inc")
		493	+ acl_header_files = glob.glob(
		494	+ os.path.join(acl_include_root, "*", ".h"),
		495	+ recursive=True,
		496	+ )
		497	+ for src in acl_header_files:
		498	+ relative_header = os.path.relpath(src, acl_include_root)
		499	+ dst = os.path.join(
		500	+ BASE_DIR,
		501	+ "build/packages/torch_npu/include",
		502	+ relative_header,
		503	+ )
		504	+ os.makedirs(os.path.dirname(dst), exist_ok=True)
		505	+ ret.append((src, dst))
		506	+
		507	+ # Preserve legacy include paths with forwarding headers, not duplicate ACL headers.
		508	+ wrapper_src = os.path.join(
		509	+ BASE_DIR, "build", "acl_compat_headers", relative_header
		510	+ )
		511	+ os.makedirs(os.path.dirname(wrapper_src), exist_ok=True)
		512	+ with open(wrapper_src, "w", encoding="utf-8", newline="\n") as wrapper_file:
		513	+ wrapper_file.write("#pragma once\n#include <{}>\n".format(
		514	+ relative_header.replace(os.sep, "/")
		515	+ ))
		516	+ wrapper_dst = os.path.join(
		517	+ BASE_DIR,
		518	+ "build/packages/torch_npu/include/third_party/acl/inc",
		519	+ relative_header,
		520	+ )
		521	+ os.makedirs(os.path.dirname(wrapper_dst), exist_ok=True)
		522	+ ret.append((wrapper_src, wrapper_dst))
		523	+
494	torch_header_files = [	524	torch_header_files = [
495	"/.h",	525	"/.h",
496	"//*.h",	526	"//*.h",

Mtest/allocator/test_pluggable_allocator_extensions.py+6-2

	@@ -21,7 +21,12 @@ def create_build_path(build_directory):
21		21
22		22
23	def build_stub(base_dir):	23	def build_stub(base_dir):
24	- build_stub_cmd = ["sh", os.path.join(base_dir, 'third_party/acl/libs/build_stub.sh')]	24	+ build_stub_cmd = [
		25	+ "sh",
		26	+ os.path.join(base_dir, 'third_party/acl/libs/build_stub.sh'),
		27	+ # Use the installed header root for both the ACL stub and C++ Extension.
		28	+ os.path.join(PYTORCH_NPU_INSTALL_PATH, "include"),
		29	+ ]
25	if subprocess.call(build_stub_cmd) != 0:	30	if subprocess.call(build_stub_cmd) != 0:
26	raise RuntimeError('Failed to build stub: {}'.format(build_stub_cmd))	31	raise RuntimeError('Failed to build stub: {}'.format(build_stub_cmd))
27		32
	@@ -47,7 +52,6 @@ class TestPluggableAllocator(TestCase):
47	extra_ldflags.append(f"-L{PYTORCH_INSTALL_PATH}")	52	extra_ldflags.append(f"-L{PYTORCH_INSTALL_PATH}")
48	extra_include_paths = [os.path.join(TEST_DIR, "cpp_extensions")]	53	extra_include_paths = [os.path.join(TEST_DIR, "cpp_extensions")]
49	extra_include_paths.append(os.path.join(PYTORCH_NPU_INSTALL_PATH, 'include'))	54	extra_include_paths.append(os.path.join(PYTORCH_NPU_INSTALL_PATH, 'include'))
50	- extra_include_paths.append(os.path.join(PYTORCH_NPU_INSTALL_PATH, 'include', 'third_party', 'acl', 'inc'))
51		55
52	cls.module = torch.utils.cpp_extension.load(	56	cls.module = torch.utils.cpp_extension.load(
53	name="pluggable_allocator_extensions",	57	name="pluggable_allocator_extensions",

Mtest/cpp_extensions/pluggable_allocator_extensions.cpp+2-2

	@@ -2,8 +2,8 @@
2	#include <iostream>	2	#include <iostream>
3	#include <torch/extension.h>	3	#include <torch/extension.h>
4		4
5	-#include "third_party/acl/inc/acl/acl_base.h"	5	+#include <acl/acl_base.h>
6	-#include "third_party/acl/inc/acl/acl_rt.h"	6	+#include <acl/acl_rt.h>
7	#include "torch_npu/csrc/core/npu/NPUCachingAllocator.h"	7	#include "torch_npu/csrc/core/npu/NPUCachingAllocator.h"
8		8
9	extern "C" {	9	extern "C" {

Mtest/distributed/rpc/test_single_machine_multi_devices.py+1-1

	@@ -335,7 +335,7 @@ class TestRpc(TestCase):
335		335
336	@skipIfUnsupportMultiNPU(2)	336	@skipIfUnsupportMultiNPU(2)
337	def test_async_call_for_cpu(self):	337	def test_async_call_for_cpu(self):
338	- inputs = [torch.rand(1024, 1024).cpu(), torch.rand(1024, 1024, 1024).cpu()]	338	+ inputs = [torch.rand(1024, 1024).cpu(), torch.rand(1024, 1024, 4).cpu()]
339	self._test_multiprocess(TestRpc._test_async_call_for_cpu, inputs, self.world_size_2p)	339	self._test_multiprocess(TestRpc._test_async_call_for_cpu, inputs, self.world_size_2p)
340		340
341	@skipIfUnsupportMultiNPU(2)	341	@skipIfUnsupportMultiNPU(2)

Mtest/npu/test_allocator_trace_tracker.py+5-2

	@@ -23,7 +23,11 @@ def create_build_path(build_directory):
23		23
24		24
25	def build_stub(base_dir):	25	def build_stub(base_dir):
26	- build_stub_cmd = ["sh", os.path.join(base_dir, "third_party/acl/libs/build_stub.sh")]	26	+ build_stub_cmd = [
		27	+ "sh",
		28	+ os.path.join(base_dir, "third_party/acl/libs/build_stub.sh"),
		29	+ os.path.join(PYTORCH_NPU_INSTALL_PATH, "include"),
		30	+ ]
27	if subprocess.call(build_stub_cmd) != 0:	31	if subprocess.call(build_stub_cmd) != 0:
28	raise RuntimeError(f"Failed to build stub: {build_stub_cmd}")	32	raise RuntimeError(f"Failed to build stub: {build_stub_cmd}")
29		33
	@@ -43,7 +47,6 @@ class TestAllocatorTraceTracker(TestCase):
43	torch_npu_lib_path = os.path.join(PYTORCH_NPU_INSTALL_PATH, "lib")	47	torch_npu_lib_path = os.path.join(PYTORCH_NPU_INSTALL_PATH, "lib")
44	extra_include_paths = [	48	extra_include_paths = [
45	os.path.join(PYTORCH_NPU_INSTALL_PATH, "include"),	49	os.path.join(PYTORCH_NPU_INSTALL_PATH, "include"),
46	- os.path.join(PYTORCH_NPU_INSTALL_PATH, "include", "third_party", "acl", "inc"),
47	]	50	]
48	extra_ldflags = [	51	extra_ldflags = [
49	f"-L{cann_lib_path}",	52	f"-L{cann_lib_path}",

Mtest/test_npu_expandable_segments.py+5-2

	@@ -23,7 +23,11 @@ def create_build_path(build_directory):
23		23
24		24
25	def build_stub(base_dir):	25	def build_stub(base_dir):
26	- build_stub_cmd = ["sh", os.path.join(base_dir, 'third_party/acl/libs/build_stub.sh')]	26	+ build_stub_cmd = [
		27	+ "sh",
		28	+ os.path.join(base_dir, 'third_party/acl/libs/build_stub.sh'),
		29	+ os.path.join(PYTORCH_NPU_INSTALL_PATH, "include"),
		30	+ ]
27	if subprocess.call(build_stub_cmd) != 0:	31	if subprocess.call(build_stub_cmd) != 0:
28	raise RuntimeError('Failed to build stub: {}'.format(build_stub_cmd))	32	raise RuntimeError('Failed to build stub: {}'.format(build_stub_cmd))
29		33
	@@ -56,7 +60,6 @@ class TestPluggableAllocator(TestCase):
56	extra_ldflags.append(f"-L{PYTORCH_INSTALL_PATH}")	60	extra_ldflags.append(f"-L{PYTORCH_INSTALL_PATH}")
57	extra_include_paths = [os.path.join(TEST_DIR, "cpp_extensions")]	61	extra_include_paths = [os.path.join(TEST_DIR, "cpp_extensions")]
58	extra_include_paths.append(os.path.join(PYTORCH_NPU_INSTALL_PATH, "include"))	62	extra_include_paths.append(os.path.join(PYTORCH_NPU_INSTALL_PATH, "include"))
59	- extra_include_paths.append(os.path.join(PYTORCH_NPU_INSTALL_PATH, 'include', 'third_party', 'acl', 'inc'))
60		63
61	cls.module = torch.utils.cpp_extension.load(	64	cls.module = torch.utils.cpp_extension.load(
62	name="pluggable_allocator_extensions",	65	name="pluggable_allocator_extensions",

Mtest/test_sanitizer_pluggable_allocator.py+1-1

	@@ -41,6 +41,7 @@ def build_stub(base_dir):
41	build_stub_cmd = [	41	build_stub_cmd = [
42	"sh",	42	"sh",
43	os.path.join(base_dir, "third_party/acl/libs/build_stub.sh"),	43	os.path.join(base_dir, "third_party/acl/libs/build_stub.sh"),
		44	+ os.path.join(PYTORCH_NPU_INSTALL_PATH, "include"),
44	]	45	]
45	if subprocess.call(build_stub_cmd) != 0:	46	if subprocess.call(build_stub_cmd) != 0:
46	raise RuntimeError(f"Failed to build stub: {build_stub_cmd}")	47	raise RuntimeError(f"Failed to build stub: {build_stub_cmd}")
	@@ -79,7 +80,6 @@ class TestSanitizerPluggableAllocator(TestCase):
79	extra_ldflags.append(f"-L{PYTORCH_INSTALL_PATH}")	80	extra_ldflags.append(f"-L{PYTORCH_INSTALL_PATH}")
80	extra_include_paths = [os.path.join(TEST_DIR, "cpp_extensions")]	81	extra_include_paths = [os.path.join(TEST_DIR, "cpp_extensions")]
81	extra_include_paths.append(os.path.join(PYTORCH_NPU_INSTALL_PATH, 'include'))	82	extra_include_paths.append(os.path.join(PYTORCH_NPU_INSTALL_PATH, 'include'))
82	- extra_include_paths.append(os.path.join(PYTORCH_NPU_INSTALL_PATH, 'include', 'third_party', 'acl', 'inc'))
83		83
84	cls.module = torch.utils.cpp_extension.load(	84	cls.module = torch.utils.cpp_extension.load(
85	name="sanitizer_pluggable_allocator_extensions",	85	name="sanitizer_pluggable_allocator_extensions",

Mtest/trans_contiguous/test_single_reshape_copy_to_contiguous.py+6-0

	@@ -9,6 +9,8 @@ from torch_npu.testing.common_utils import create_common_tensor, check_operators
9		9
10	os.environ["COMBINED_ENABLE"] = "1" # Open combined-view cases optimization	10	os.environ["COMBINED_ENABLE"] = "1" # Open combined-view cases optimization
11		11
		12	+SKIP_REASON = "Temporarily skipped; see https://gitcode.com/Ascend/pytorch/issues/4356"
		13	+
12	# Optimized view Ops contains Transpose, permute, narrow, strideslice, select, unfold	14	# Optimized view Ops contains Transpose, permute, narrow, strideslice, select, unfold
13		15
14	# The test case is a continuous optimization test case for aclop	16	# The test case is a continuous optimization test case for aclop
	@@ -19,6 +21,7 @@ os.environ["COMBINED_ENABLE"] = "1" # Open combined-view cases optimization
19		21
20		22
21	class SingleViewCopyToContiguous(TestCase):	23	class SingleViewCopyToContiguous(TestCase):
		24	+ @unittest.skip(SKIP_REASON)
22	def test_view_copy(self, device="npu"):	25	def test_view_copy(self, device="npu"):
23	dtype_list1 = [np.float16, np.float32]	26	dtype_list1 = [np.float16, np.float32]
24	format_list1 = [0, 3, 29]	27	format_list1 = [0, 3, 29]
	@@ -66,6 +69,7 @@ class SingleViewCopyToContiguous(TestCase):
66	cpu_out2 = cpu_input.view(1, 6, cpu_input.size(2) * cpu_input.size(3), 1).clone()	69	cpu_out2 = cpu_input.view(1, 6, cpu_input.size(2) * cpu_input.size(3), 1).clone()
67	self.assertRtolEqual(npu_out2.to("cpu").numpy(), cpu_out2.numpy())	70	self.assertRtolEqual(npu_out2.to("cpu").numpy(), cpu_out2.numpy())
68		71
		72	+ @unittest.skip(SKIP_REASON)
69	def test_unsqueeze_copy(self, device="npu"):	73	def test_unsqueeze_copy(self, device="npu"):
70	dtype_list2 = [np.float16, np.float32]	74	dtype_list2 = [np.float16, np.float32]
71	format_list2 = [2, 3, 29]	75	format_list2 = [2, 3, 29]
	@@ -98,6 +102,7 @@ class SingleViewCopyToContiguous(TestCase):
98	cpu_out = cpu_input.unsqueeze(i).clone()	102	cpu_out = cpu_input.unsqueeze(i).clone()
99	self.assertRtolEqual(npu_out.to("cpu").numpy(), cpu_out.numpy())	103	self.assertRtolEqual(npu_out.to("cpu").numpy(), cpu_out.numpy())
100		104
		105	+ @unittest.skip(SKIP_REASON)
101	def test_flatten_copy(self, device="npu"):	106	def test_flatten_copy(self, device="npu"):
102	dtype_list3 = [np.float16, np.float32]	107	dtype_list3 = [np.float16, np.float32]
103	format_list3 = [0, 3, 29]	108	format_list3 = [0, 3, 29]
	@@ -127,6 +132,7 @@ class SingleViewCopyToContiguous(TestCase):
127	cpu_out = torch.flatten(cpu_input, 0, 1).clone()	132	cpu_out = torch.flatten(cpu_input, 0, 1).clone()
128	self.assertRtolEqual(npu_out.to("cpu").numpy(), cpu_out.numpy())	133	self.assertRtolEqual(npu_out.to("cpu").numpy(), cpu_out.numpy())
129		134
		135	+ @unittest.skip(SKIP_REASON)
130	def test_narrow_at_first_axis_copy(self, device="npu"):	136	def test_narrow_at_first_axis_copy(self, device="npu"):
131	# this case: slice at the first dim, tensor with offset remains contiguous	137	# this case: slice at the first dim, tensor with offset remains contiguous
132	dtype_list4 = [np.float16, np.float32]	138	dtype_list4 = [np.float16, np.float32]

Mthird_party/acl/libs/build_stub.sh+11-8

	@@ -1,20 +1,23 @@
1	#!/bin/bash	1	#!/bin/bash
2		2
		3	+set -e
		4	+
3	CDIR="$(cd "$(dirname "$0")" ; pwd -P)"	5	CDIR="$(cd "$(dirname "$0")" ; pwd -P)"
4		6
5	-cd ${CDIR}	7	+cd "${CDIR}"
		8	+
		9	+ACL_INCLUDE_DIR="${1:-../inc}"
6		10
7	gcc -fPIC -shared -o libhccl.so -I./ hccl.cpp	11	gcc -fPIC -shared -o libhccl.so -I./ hccl.cpp
8		12
9	-gcc -fPIC -shared -o libascendcl.so -I../inc acl.cpp	13	+gcc -fPIC -shared -o libascendcl.so -I"${ACL_INCLUDE_DIR}" acl.cpp
10		14
11	-gcc -fPIC -shared -o libacl_op_compiler.so -I../inc acl_op_compiler.cpp	15	+gcc -fPIC -shared -o libacl_op_compiler.so -I"${ACL_INCLUDE_DIR}" acl_op_compiler.cpp
12		16
13	-gcc -fPIC -shared -o libge_runner.so -I../inc ge_runner.cpp ge_api.cpp	17	+gcc -fPIC -shared -o libge_runner.so -I"${ACL_INCLUDE_DIR}" ge_runner.cpp ge_api.cpp
14		18
15	-gcc -fPIC -shared -o libgraph.so -I../inc graph.cpp operator_factory.cpp operator.cpp tensor.cpp	19	+gcc -fPIC -shared -o libgraph.so -I"${ACL_INCLUDE_DIR}" graph.cpp operator_factory.cpp operator.cpp tensor.cpp
16		20
17	-gcc -fPIC -shared -o libacl_tdt_channel.so -I../inc acl_tdt.cpp	21	+gcc -fPIC -shared -o libacl_tdt_channel.so -I"${ACL_INCLUDE_DIR}" acl_tdt.cpp
18	-
19	-gcc -fPIC -shared -o libascend_ml.so -I../inc aml_fwk_detect.cpp
20		22
		23	+gcc -fPIC -shared -o libascend_ml.so -I"${ACL_INCLUDE_DIR}" aml_fwk_detect.cpp

Mthird_party/hccl/inc/hccl/hccl.h+1-1

	@@ -12,7 +12,7 @@
12	#define HCCL_H_	12	#define HCCL_H_
13		13
14	#include "third_party/hccl/inc/hccl/hccl_types.h"	14	#include "third_party/hccl/inc/hccl/hccl_types.h"
15	-#include "third_party/acl/inc/acl/acl.h"	15	+#include <acl/acl.h>
16		16
17	#ifdef __cplusplus	17	#ifdef __cplusplus
18	extern "C" {	18	extern "C" {

Mtorch_npu/_inductor/ascend_npu_ir/ascend_npu_ir/npu/utils.py+0-2

	@@ -95,13 +95,11 @@ def _build_npu_ext(obj_name: str, src_path, src_dir) -> str:
95		95
96	torch_npu_dir = torch_npu_root / "include"	96	torch_npu_dir = torch_npu_root / "include"
97	torch_npu_lib_dir = torch_npu_root / "lib"	97	torch_npu_lib_dir = torch_npu_root / "lib"
98	- acl_inc_dir = torch_npu_dir / "third_party" / "acl" / "inc"
99		98
100	cc_cmd += [	99	cc_cmd += [
101	f"-I{torch_npu_dir}",	100	f"-I{torch_npu_dir}",
102	f"-I{cpp_common_dir}",	101	f"-I{cpp_common_dir}",
103	f"-L{torch_npu_lib_dir}",	102	f"-L{torch_npu_lib_dir}",
104	- f"-I{acl_inc_dir}",
105	"-ltorch_npu",	103	"-ltorch_npu",
106	f"-Wl,-rpath",	104	f"-Wl,-rpath",
107	"-std=c++17",	105	"-std=c++17",

Mtorch_npu/_inductor/cpp_builder.py+2-3

	@@ -43,7 +43,7 @@ def include_paths(npu: bool = False) -> List[str]:
43	os.path.join(lib_include, 'TH'),	43	os.path.join(lib_include, 'TH'),
44	os.path.join(lib_include, 'THC')	44	os.path.join(lib_include, 'THC')
45	]	45	]
46	- include_path = os.path.join(PYTORCH_NPU_INSTALL_PATH, "include", "third_party", "acl", "inc")	46	+ include_path = os.path.join(PYTORCH_NPU_INSTALL_PATH, "include")
47	paths.extend([include_path])	47	paths.extend([include_path])
48	if npu:	48	if npu:
49	ASCEND_HOME = get_ascend_home()	49	ASCEND_HOME = get_ascend_home()
	@@ -55,7 +55,6 @@ def include_paths(npu: bool = False) -> List[str]:
55	os.path.join(ASCEND_HOME, "include/experiment/msprof"),	55	os.path.join(ASCEND_HOME, "include/experiment/msprof"),
56	])	56	])
57		57
58	- paths.append(os.path.join(PYTORCH_NPU_INSTALL_PATH, "include"))
59	return paths	58	return paths
60		59
61		60
	@@ -163,4 +162,4 @@ def patch_get_cpp_torch_device_options():
163		162
164		163
165	def patch_get_optimization_cflags():	164	def patch_get_optimization_cflags():
166	- cpp_builder._get_optimization_cflags = _get_optimization_cflags	165	+ cpp_builder._get_optimization_cflags = _get_optimization_cflags

Mtorch_npu/csrc/aten/common/CopyMemoryKernel.cpp+1-1

	@@ -8,7 +8,7 @@
8	#include "torch_npu/csrc/aten/NPUNativeFunctions.h"	8	#include "torch_npu/csrc/aten/NPUNativeFunctions.h"
9	#include "torch_npu/csrc/core/NPUBridge.h"	9	#include "torch_npu/csrc/core/NPUBridge.h"
10	#include "torch_npu/csrc/core/npu/interface/AsyncTaskQueueInterface.h"	10	#include "torch_npu/csrc/core/npu/interface/AsyncTaskQueueInterface.h"
11	-#include "third_party/acl/inc/acl/acl.h"	11	+#include <acl/acl.h>
12		12
13	namespace at_npu {	13	namespace at_npu {
14	namespace native {	14	namespace native {

Mtorch_npu/csrc/aten/common/FormatCastHelper.h+1-1

	@@ -3,7 +3,7 @@
3		3
4	#include <ATen/ATen.h>	4	#include <ATen/ATen.h>
5		5
6	-#include "third_party/acl/inc/acl/acl_base.h"	6	+#include <acl/acl_base.h>
7		7
8	namespace at_npu {	8	namespace at_npu {
9	namespace native {	9	namespace native {

Mtorch_npu/csrc/aten/common/LocalScalarDenseNpu.cpp+2-2

	@@ -2,8 +2,8 @@
2	#include <ATen/NativeFunctions.h>	2	#include <ATen/NativeFunctions.h>
3	#include <ATen/Dispatch_v2.h>	3	#include <ATen/Dispatch_v2.h>
4		4
5	-#include "third_party/acl/inc/acl/acl_base.h"	5	+#include <acl/acl_base.h>
6	-#include "third_party/acl/inc/acl/acl_rt.h"	6	+#include <acl/acl_rt.h>
7	#include "torch_npu/csrc/core/npu/NPUStream.h"	7	#include "torch_npu/csrc/core/npu/NPUStream.h"
8	#include "torch_npu/csrc/framework/utils/CalcuOpUtil.h"	8	#include "torch_npu/csrc/framework/utils/CalcuOpUtil.h"
9	#include "torch_npu/csrc/aten/NPUNativeFunctions.h"	9	#include "torch_npu/csrc/aten/NPUNativeFunctions.h"

Mtorch_npu/csrc/aten/common/TensorShape.cpp+1-1

	@@ -23,7 +23,7 @@
23	#include "torch_npu/csrc/aten/common/ResizeNpu.h"	23	#include "torch_npu/csrc/aten/common/ResizeNpu.h"
24	#include "torch_npu/csrc/framework/StorageDescHelper.h"	24	#include "torch_npu/csrc/framework/StorageDescHelper.h"
25	#include "torch_npu/csrc/core/npu/NPUGuard.h"	25	#include "torch_npu/csrc/core/npu/NPUGuard.h"
26	-#include "third_party/acl/inc/acl/acl_base.h"	26	+#include <acl/acl_base.h>
27	#include "op_plugin/OpInterface.h"	27	#include "op_plugin/OpInterface.h"
28		28
29	namespace {	29	namespace {

Mtorch_npu/csrc/core/NPUSerialization.cpp+1-1

	@@ -3,7 +3,7 @@
3	#include "torch_npu/csrc/core/npu/NPUGuard.h"	3	#include "torch_npu/csrc/core/npu/NPUGuard.h"
4	#include "torch_npu/csrc/core/NPUSerialization.h"	4	#include "torch_npu/csrc/core/NPUSerialization.h"
5	#include "torch_npu/csrc/framework/FormatHelper.h"	5	#include "torch_npu/csrc/framework/FormatHelper.h"
6	-#include "third_party/acl/inc/acl/acl_base.h"	6	+#include <acl/acl_base.h>
7	#include "torch_npu/csrc/framework/StorageDescHelper.h"	7	#include "torch_npu/csrc/framework/StorageDescHelper.h"
8		8
9	namespace torch_npu {	9	namespace torch_npu {

Mtorch_npu/csrc/core/NPUStorageImpl.h+2-2

	@@ -8,8 +8,8 @@
8	#include <c10/util/typeid.h>	8	#include <c10/util/typeid.h>
9	#include <c10/util/order_preserving_flat_hash_map.h>	9	#include <c10/util/order_preserving_flat_hash_map.h>
10		10
11	-#include "third_party/acl/inc/acl/acl_rt.h"	11	+#include <acl/acl_rt.h>
12	-#include "third_party/acl/inc/acl/acl_base.h"	12	+#include <acl/acl_base.h>
13		13
14	namespace torch_npu {	14	namespace torch_npu {
15		15

Mtorch_npu/csrc/core/NPUTensorImpl.cpp+1-1

	@@ -5,7 +5,7 @@
5		5
6	#include "torch_npu/csrc/framework/StorageDescHelper.h"	6	#include "torch_npu/csrc/framework/StorageDescHelper.h"
7	#include "torch_npu/csrc/core/NPUTensorImpl.h"	7	#include "torch_npu/csrc/core/NPUTensorImpl.h"
8	-#include "third_party/acl/inc/acl/acl_rt.h"	8	+#include <acl/acl_rt.h>
9	#include "torch_npu/csrc/core/NPUStorageImpl.h"	9	#include "torch_npu/csrc/core/NPUStorageImpl.h"
10		10
11	namespace torch_npu {	11	namespace torch_npu {

Mtorch_npu/csrc/core/npu/CachingHostAllocator.h+2-2

	@@ -6,8 +6,8 @@
6	#include "torch_npu/csrc/core/npu/NPUMacros.h"	6	#include "torch_npu/csrc/core/npu/NPUMacros.h"
7	#include "torch_npu/csrc/core/npu/NPUStream.h"	7	#include "torch_npu/csrc/core/npu/NPUStream.h"
8	#include "torch_npu/csrc/core/npu/NPUException.h"	8	#include "torch_npu/csrc/core/npu/NPUException.h"
9	-#include <third_party/acl/inc/acl/acl.h>	9	+#include <acl/acl.h>
10	-#include <third_party/acl/inc/acl/acl_rt.h>	10	+#include <acl/acl_rt.h>
11		11
12	#include <c10/core/DeviceGuard.h>	12	#include <c10/core/DeviceGuard.h>
13	#include <ATen/DeviceGuard.h>	13	#include <ATen/DeviceGuard.h>

Mtorch_npu/csrc/core/npu/GetCANNInfo.cpp+1-1

	@@ -9,7 +9,7 @@
9	#include "torch_npu/csrc/core/npu/register/FunctionLoader.h"	9	#include "torch_npu/csrc/core/npu/register/FunctionLoader.h"
10	#include "torch_npu/csrc/core/npu/NPUException.h"	10	#include "torch_npu/csrc/core/npu/NPUException.h"
11	#include "torch_npu/csrc/core/npu/interface/AclInterface.h"	11	#include "torch_npu/csrc/core/npu/interface/AclInterface.h"
12	-#include "third_party/acl/inc/acl/acl.h"	12	+#include <acl/acl.h>
13		13
14		14
15	constexpr size_t kVersionIndex1 = 1;	15	constexpr size_t kVersionIndex1 = 1;

Mtorch_npu/csrc/core/npu/NPUCachingAllocator.cpp+2-2

	@@ -17,8 +17,8 @@
17	#include <c10/util/llvmMathExtras.h>	17	#include <c10/util/llvmMathExtras.h>
18	#include <c10/util/ScopeExit.h>	18	#include <c10/util/ScopeExit.h>
19		19
20	-#include "third_party/acl/inc/acl/acl_base.h"	20	+#include <acl/acl_base.h>
21	-#include "third_party/acl/inc/acl/acl_rt.h"	21	+#include <acl/acl_rt.h>
22	#include "torch_npu/csrc/core/npu/interface/AsyncTaskQueueInterface.h"	22	#include "torch_npu/csrc/core/npu/interface/AsyncTaskQueueInterface.h"
23	#include "torch_npu/csrc/core/npu/NPUCachingAllocator.h"	23	#include "torch_npu/csrc/core/npu/NPUCachingAllocator.h"
24	#include "torch_npu/csrc/core/npu/NPUAllocatorConfig.h"	24	#include "torch_npu/csrc/core/npu/NPUAllocatorConfig.h"

Mtorch_npu/csrc/core/npu/NPUEvent.h+1-1

	@@ -4,7 +4,7 @@
4	#include "torch_npu/csrc/core/npu/NPUMacros.h"	4	#include "torch_npu/csrc/core/npu/NPUMacros.h"
5	#include "torch_npu/csrc/core/npu/NPUGuard.h"	5	#include "torch_npu/csrc/core/npu/NPUGuard.h"
6	#include "torch_npu/csrc/core/npu/NPUFunctions.h"	6	#include "torch_npu/csrc/core/npu/NPUFunctions.h"
7	-#include "third_party/acl/inc/acl/acl.h"	7	+#include <acl/acl.h>
8	#include <cstdint>	8	#include <cstdint>
9	#include <utility>	9	#include <utility>
10		10

Mtorch_npu/csrc/core/npu/NPUEventManager.h+2-2

	@@ -5,7 +5,7 @@
5	#include <unordered_set>	5	#include <unordered_set>
6	#include <c10/core/thread_pool.h>	6	#include <c10/core/thread_pool.h>
7	#include <c10/util/flat_hash_map.h>	7	#include <c10/util/flat_hash_map.h>
8	-#include <third_party/acl/inc/acl/acl.h>	8	+#include <acl/acl.h>
9		9
10	#include "torch_npu/csrc/core/npu/NPUException.h"	10	#include "torch_npu/csrc/core/npu/NPUException.h"
11		11
	@@ -49,4 +49,4 @@ private:
49	std::unordered_set<aclrtEvent> ipc_events_;	49	std::unordered_set<aclrtEvent> ipc_events_;
50	};	50	};
51		51
52	-} // namespace c10_npu	52	+} // namespace c10_npu

Mtorch_npu/csrc/core/npu/NPUException.h+1-1

	@@ -15,7 +15,7 @@
15	#include <regex>	15	#include <regex>
16	#include <c10/macros/Macros.h>	16	#include <c10/macros/Macros.h>
17	#include <c10/util/Exception.h>	17	#include <c10/util/Exception.h>
18	-#include <third_party/acl/inc/acl/acl_base.h>	18	+#include <acl/acl_base.h>
19	#include "torch_npu/csrc/core/npu/NPUMacros.h"	19	#include "torch_npu/csrc/core/npu/NPUMacros.h"
20	#include "torch_npu/csrc/core/npu/interface/AclInterface.h"	20	#include "torch_npu/csrc/core/npu/interface/AclInterface.h"
21	#include "torch_npu/csrc/core/npu/NPUErrorCodes.h"	21	#include "torch_npu/csrc/core/npu/NPUErrorCodes.h"

Mtorch_npu/csrc/core/npu/NPUFunctions.cpp+1-1

	@@ -11,7 +11,7 @@
11	#include "torch_npu/csrc/core/npu/register/OptionsManager.h"	11	#include "torch_npu/csrc/core/npu/register/OptionsManager.h"
12	#include "torch_npu/csrc/core/npu/GetCANNInfo.h"	12	#include "torch_npu/csrc/core/npu/GetCANNInfo.h"
13	#include "torch_npu/csrc/framework/interface/EnvVariables.h"	13	#include "torch_npu/csrc/framework/interface/EnvVariables.h"
14	-#include "third_party/acl/inc/acl/acl_rt.h"	14	+#include <acl/acl_rt.h>
15	#ifndef BUILD_LIBTORCH	15	#ifndef BUILD_LIBTORCH
16	#include "torch_npu/csrc/sanitizer/NPUTrace.h"	16	#include "torch_npu/csrc/sanitizer/NPUTrace.h"
17	#endif	17	#endif

Mtorch_npu/csrc/core/npu/NPUFunctions.h+1-1

	@@ -16,7 +16,7 @@
16	#include "torch_npu/csrc/core/npu/npu_log.h"	16	#include "torch_npu/csrc/core/npu/npu_log.h"
17	#include "torch_npu/csrc/core/npu/NPUMacros.h"	17	#include "torch_npu/csrc/core/npu/NPUMacros.h"
18	#include "torch_npu/csrc/core/npu/NPUStream.h"	18	#include "torch_npu/csrc/core/npu/NPUStream.h"
19	-#include <third_party/acl/inc/acl/acl.h>	19	+#include <acl/acl.h>
20		20
21	namespace c10_npu {	21	namespace c10_npu {
22		22

Mtorch_npu/csrc/core/npu/NPUGraph.cpp+1-1

	@@ -4,7 +4,7 @@
4	#include "torch_npu/csrc/core/npu/NPUStreamUtils.h"	4	#include "torch_npu/csrc/core/npu/NPUStreamUtils.h"
5	#include "torch_npu/csrc/aten/NPUGeneratorImpl.h"	5	#include "torch_npu/csrc/aten/NPUGeneratorImpl.h"
6	#include "torch_npu/csrc/core/npu/register/OptionRegister.h"	6	#include "torch_npu/csrc/core/npu/register/OptionRegister.h"
7	-#include "third_party/acl/inc/acl/error_codes/rt_error_codes.h"	7	+#include <acl/error_codes/rt_error_codes.h>
8	#include "torch_npu/csrc/core/npu/sys_ctrl/npu_sys_ctrl.h"	8	#include "torch_npu/csrc/core/npu/sys_ctrl/npu_sys_ctrl.h"
9	#include "torch_npu/csrc/core/npu/GetCANNInfo.h"	9	#include "torch_npu/csrc/core/npu/GetCANNInfo.h"
10		10

Mtorch_npu/csrc/core/npu/NPUGraph.h+3-3

	@@ -4,9 +4,9 @@
4	#include <c10/core/Device.h>	4	#include <c10/core/Device.h>
5	#include <c10/util/flat_hash_map.h>	5	#include <c10/util/flat_hash_map.h>
6		6
7	-#include "third_party/acl/inc/acl/acl_base.h"	7	+#include <acl/acl_base.h>
8	-#include "third_party/acl/inc/acl/acl_rt.h"	8	+#include <acl/acl_rt.h>
9	-#include "third_party/acl/inc/acl/super_kernel.h"	9	+#include <acl/super_kernel.h>
10	#include "torch_npu/csrc/core/npu/interface/SkInterface.h"	10	#include "torch_npu/csrc/core/npu/interface/SkInterface.h"
11	#include "torch_npu/csrc/core/npu/NPUGraphsUtils.h"	11	#include "torch_npu/csrc/core/npu/NPUGraphsUtils.h"
12	#include "torch_npu/csrc/core/npu/NPUMacros.h"	12	#include "torch_npu/csrc/core/npu/NPUMacros.h"

Mtorch_npu/csrc/core/npu/NPUPeerToPeerAccess.cpp+1-1

	@@ -1,7 +1,7 @@
1	#include <vector>	1	#include <vector>
2	#include <c10/util/Exception.h>	2	#include <c10/util/Exception.h>
3	#include <c10/util/irange.h>	3	#include <c10/util/irange.h>
4	-#include <third_party/acl/inc/acl/acl_rt.h>	4	+#include <acl/acl_rt.h>
5	#include "torch_npu/csrc/core/npu/NpuVariables.h"	5	#include "torch_npu/csrc/core/npu/NpuVariables.h"
6	#include "torch_npu/csrc/core/npu/NPUPeerToPeerAccess.h"	6	#include "torch_npu/csrc/core/npu/NPUPeerToPeerAccess.h"
7	#include "torch_npu/csrc/core/npu/NPUGuard.h"	7	#include "torch_npu/csrc/core/npu/NPUGuard.h"

Mtorch_npu/csrc/core/npu/NPUQueue.cpp+1-1

	@@ -20,7 +20,7 @@
20	#include <sstream>	20	#include <sstream>
21	#include <sys/time.h>	21	#include <sys/time.h>
22	#include <sys/eventfd.h>	22	#include <sys/eventfd.h>
23	-#include <third_party/acl/inc/acl/acl_rt.h>	23	+#include <acl/acl_rt.h>
24		24
25	namespace c10_npu {	25	namespace c10_npu {
26	struct timeval delay = { 0, 1 };	26	struct timeval delay = { 0, 1 };

Mtorch_npu/csrc/core/npu/NPUQueue.h+2-2

	@@ -7,7 +7,7 @@
7		7
8	#include <c10/core/Device.h>	8	#include <c10/core/Device.h>
9	#include "torch_npu/csrc/logging/LogContext.h"	9	#include "torch_npu/csrc/logging/LogContext.h"
10	-#include <third_party/acl/inc/acl/acl_op.h>	10	+#include <acl/acl_op.h>
11		11
12	namespace c10_npu {	12	namespace c10_npu {
13		13
	@@ -184,4 +184,4 @@ public:
184	#define REGISTER_QUEUE_FUNC(execF, copyF, releaseF, newF, deleteF, copyReleaseParamF, releaseParamF) \	184	#define REGISTER_QUEUE_FUNC(execF, copyF, releaseF, newF, deleteF, copyReleaseParamF, releaseParamF) \
185	static ::c10_npu::register_queue_cb::NPUCallBackRegisterBuilder \	185	static ::c10_npu::register_queue_cb::NPUCallBackRegisterBuilder \
186	register_queue_func_builder(execF, copyF, releaseF, newF, deleteF, copyReleaseParamF, releaseParamF);	186	register_queue_func_builder(execF, copyF, releaseF, newF, deleteF, copyReleaseParamF, releaseParamF);
187	-} // namespace c10_npu	187	+} // namespace c10_npu

Mtorch_npu/csrc/core/npu/NPUStream.cpp+1-1

	@@ -20,7 +20,7 @@
20	#include "torch_npu/csrc/core/npu/NPUStreamUtils.h"	20	#include "torch_npu/csrc/core/npu/NPUStreamUtils.h"
21	#include "torch_npu/csrc/core/npu/sys_ctrl/npu_sys_ctrl.h"	21	#include "torch_npu/csrc/core/npu/sys_ctrl/npu_sys_ctrl.h"
22	#include "torch_npu/csrc/core/npu/interface/AsyncTaskQueueInterface.h"	22	#include "torch_npu/csrc/core/npu/interface/AsyncTaskQueueInterface.h"
23	-#include "third_party/acl/inc/acl/acl_rt.h"	23	+#include <acl/acl_rt.h>
24	#ifndef BUILD_LIBTORCH	24	#ifndef BUILD_LIBTORCH
25	#include "torch_npu/csrc/sanitizer/NPUTrace.h"	25	#include "torch_npu/csrc/sanitizer/NPUTrace.h"
26	#endif	26	#endif

Mtorch_npu/csrc/core/npu/NPUStream.h+2-2

	@@ -10,8 +10,8 @@
10	#include "torch_npu/csrc/core/npu/NPUException.h"	10	#include "torch_npu/csrc/core/npu/NPUException.h"
11	#include "torch_npu/csrc/core/npu/NPUQueue.h"	11	#include "torch_npu/csrc/core/npu/NPUQueue.h"
12	#include "torch_npu/csrc/core/npu/npu_log.h"	12	#include "torch_npu/csrc/core/npu/npu_log.h"
13	-#include "third_party/acl/inc/acl/acl.h"	13	+#include <acl/acl.h>
14	-#include "third_party/acl/inc/acl/acl_op.h"	14	+#include <acl/acl_op.h>
15	#include "torch_npu/csrc/aten/NPUNativeFunctions.h"	15	#include "torch_npu/csrc/aten/NPUNativeFunctions.h"
16		16
17	namespace c10_npu {	17	namespace c10_npu {

Mtorch_npu/csrc/core/npu/NPUSwappedMemoryAllocator.cpp+2-2

	@@ -1,8 +1,8 @@
1	#include <unistd.h>	1	#include <unistd.h>
2	#include <c10/util/flat_hash_map.h>	2	#include <c10/util/flat_hash_map.h>
3		3
4	-#include "third_party/acl/inc/acl/acl_base.h"	4	+#include <acl/acl_base.h>
5	-#include "third_party/acl/inc/acl/acl_rt.h"	5	+#include <acl/acl_rt.h>
6	#include "torch_npu/csrc/core/npu/NPUFunctions.h"	6	#include "torch_npu/csrc/core/npu/NPUFunctions.h"
7	#include "torch_npu/csrc/core/npu/NpuVariables.h"	7	#include "torch_npu/csrc/core/npu/NpuVariables.h"
8	#include "torch_npu/csrc/core/npu/NPUSwappedMemoryAllocator.h"	8	#include "torch_npu/csrc/core/npu/NPUSwappedMemoryAllocator.h"

Mtorch_npu/csrc/core/npu/NPUWorkspaceAllocator.cpp+2-2

	@@ -4,8 +4,8 @@
4	#include <c10/util/flat_hash_map.h>	4	#include <c10/util/flat_hash_map.h>
5	#include <c10/util/irange.h>	5	#include <c10/util/irange.h>
6		6
7	-#include "third_party/acl/inc/acl/acl_base.h"	7	+#include <acl/acl_base.h>
8	-#include "third_party/acl/inc/acl/acl_rt.h"	8	+#include <acl/acl_rt.h>
9	#include "torch_npu/csrc/core/npu/interface/AsyncTaskQueueInterface.h"	9	#include "torch_npu/csrc/core/npu/interface/AsyncTaskQueueInterface.h"
10	#include "torch_npu/csrc/core/npu/register/OptionsManager.h"	10	#include "torch_npu/csrc/core/npu/register/OptionsManager.h"
11	#include "torch_npu/csrc/core/npu/NPUFunctions.h"	11	#include "torch_npu/csrc/core/npu/NPUFunctions.h"

Mtorch_npu/csrc/core/npu/impl/NPUGuardImpl.h+3-3

	@@ -7,9 +7,9 @@
7	#include "torch_npu/csrc/core/npu/NPUException.h"	7	#include "torch_npu/csrc/core/npu/NPUException.h"
8	#include "torch_npu/csrc/core/npu/NPUFunctions.h"	8	#include "torch_npu/csrc/core/npu/NPUFunctions.h"
9	#include "torch_npu/csrc/core/npu/NPUStream.h"	9	#include "torch_npu/csrc/core/npu/NPUStream.h"
10	-#include "third_party/acl/inc/acl/acl.h"	10	+#include <acl/acl.h>
11	-#include "third_party/acl/inc/acl/acl_base.h"	11	+#include <acl/acl_base.h>
12	-#include "third_party/acl/inc/acl/acl_rt.h"	12	+#include <acl/acl_rt.h>
13		13
14	namespace c10_npu {	14	namespace c10_npu {
15	namespace impl {	15	namespace impl {

Mtorch_npu/csrc/core/npu/interface/AclInterface.h+5-5

	@@ -2,12 +2,12 @@
2		2
3	#include <c10/core/Device.h>	3	#include <c10/core/Device.h>
4		4
5	-#include "third_party/acl/inc/acl/acl_rt.h"	5	+#include <acl/acl_rt.h>
6	-#include "third_party/acl/inc/acl/acl_base.h"	6	+#include <acl/acl_base.h>
7	-#include "third_party/acl/inc/acl/acl_mdl.h"	7	+#include <acl/acl_mdl.h>
8	-#include "third_party/acl/inc/acl/acl_prof.h"	8	+#include <acl/acl_prof.h>
9	#include "torch_npu/csrc/core/npu/interface/HcclInterface.h"	9	#include "torch_npu/csrc/core/npu/interface/HcclInterface.h"
10	-#include "third_party/acl/inc/acl/acl.h"	10	+#include <acl/acl.h>
11		11
12	using aclrtHostFunc = void ()(void args);	12	using aclrtHostFunc = void ()(void args);
13	struct aclrtMemUsageInfo;	13	struct aclrtMemUsageInfo;

Mtorch_npu/csrc/core/npu/interface/AsyncTaskQueueInterface.cpp+1-1

	@@ -6,7 +6,7 @@
6	#include <ATen/record_function.h>	6	#include <ATen/record_function.h>
7	#include "torch_npu/csrc/framework/utils/NpuUtils.h"	7	#include "torch_npu/csrc/framework/utils/NpuUtils.h"
8	#include "torch_npu/csrc/framework/NPUDefine.h"	8	#include "torch_npu/csrc/framework/NPUDefine.h"
9	-#include "third_party/acl/inc/acl/acl_rt.h"	9	+#include <acl/acl_rt.h>
10	#ifndef BUILD_LIBTORCH	10	#ifndef BUILD_LIBTORCH
11	#include "torch_npu/csrc/sanitizer/NPUTrace.h"	11	#include "torch_npu/csrc/sanitizer/NPUTrace.h"
12	#endif	12	#endif

Mtorch_npu/csrc/core/npu/interface/AsyncTaskQueueInterface.h+1-1

	@@ -2,7 +2,7 @@
2		2
3	#include "c10/core/Storage.h"	3	#include "c10/core/Storage.h"
4	#include "torch_npu/csrc/core/npu/NPUStream.h"	4	#include "torch_npu/csrc/core/npu/NPUStream.h"
5	-#include "third_party/acl/inc/acl/acl_rt.h"	5	+#include <acl/acl_rt.h>
6		6
7	namespace c10_npu {	7	namespace c10_npu {
8	namespace queue {	8	namespace queue {

Mtorch_npu/csrc/core/npu/interface/MlInterface.h+1-1

	@@ -1,5 +1,5 @@
1	#pragma once	1	#pragma once
2	-#include "third_party/acl/inc/aml/aml_fwk_detect.h"	2	+#include <aml/aml_fwk_detect.h>
3		3
4	namespace c10_npu {	4	namespace c10_npu {
5	namespace amlapi {	5	namespace amlapi {

Mtorch_npu/csrc/core/npu/interface/SkInterface.h+3-3

	@@ -1,8 +1,8 @@
1	#pragma once	1	#pragma once
2		2
3	-#include "third_party/acl/inc/acl/super_kernel.h"	3	+#include <acl/super_kernel.h>
4	-#include "third_party/acl/inc/acl/acl_base.h"	4	+#include <acl/acl_base.h>
5	-#include "third_party/acl/inc/acl/acl_mdl.h"	5	+#include <acl/acl_mdl.h>
6		6
7	namespace c10_npu {	7	namespace c10_npu {
8	namespace skapi {	8	namespace skapi {

Mtorch_npu/csrc/core/npu/npu_log.h+1-1

	@@ -3,7 +3,7 @@
3	#include <iostream>	3	#include <iostream>
4	#include <string>	4	#include <string>
5	#include <stdio.h>	5	#include <stdio.h>
6	-#include "third_party/acl/inc/acl/acl_base.h"	6	+#include <acl/acl_base.h>
7	#include "torch_npu/csrc/core/npu/register/OptionsManager.h"	7	#include "torch_npu/csrc/core/npu/register/OptionsManager.h"
8		8
9	#if defined(__GNUC__) && __GNUC__ >= 12	9	#if defined(__GNUC__) && __GNUC__ >= 12

Mtorch_npu/csrc/core/npu/sys_ctrl/npu_sys_ctrl.cpp+2-2

	@@ -22,8 +22,8 @@
22	#include "torch_npu/csrc/core/npu/register/OptionsManager.h"	22	#include "torch_npu/csrc/core/npu/register/OptionsManager.h"
23	#include "torch_npu/csrc/core/npu/NpuVariables.h"	23	#include "torch_npu/csrc/core/npu/NpuVariables.h"
24	#include "torch_npu/csrc/distributed/symm_mem/NPUSHMEMInterface.h"	24	#include "torch_npu/csrc/distributed/symm_mem/NPUSHMEMInterface.h"
25	-#include "third_party/acl/inc/acl/acl_op_compiler.h"	25	+#include <acl/acl_op_compiler.h>
26	-#include "third_party/acl/inc/acl/acl_rt.h"	26	+#include <acl/acl_rt.h>
27	#include "torch_npu/csrc/framework/interface/AclOpCompileInterface.h"	27	#include "torch_npu/csrc/framework/interface/AclOpCompileInterface.h"
28	#include "torch_npu/csrc/framework/LazyInitAclops.h"	28	#include "torch_npu/csrc/framework/LazyInitAclops.h"
29	#include "torch_npu/csrc/core/npu/NPUFunctions.h"	29	#include "torch_npu/csrc/core/npu/NPUFunctions.h"

Mtorch_npu/csrc/core/npu/sys_ctrl/npu_sys_ctrl.h+1-1

	@@ -1,6 +1,6 @@
1	#pragma once	1	#pragma once
2		2
3	-#include <third_party/acl/inc/acl/acl.h>	3	+#include <acl/acl.h>
4	#include <map>	4	#include <map>
5	#include <string>	5	#include <string>
6	#include <vector>	6	#include <vector>

Mtorch_npu/csrc/custom_dtype/Init.h+1-1

	@@ -7,7 +7,7 @@
7	#include "torch_npu/csrc/core/npu/NPUMacros.h"	7	#include "torch_npu/csrc/core/npu/NPUMacros.h"
8	#include "torch_npu/csrc/core/npu/NPUException.h"	8	#include "torch_npu/csrc/core/npu/NPUException.h"
9	#include "torch_npu/csrc/framework/utils/OpPreparation.h"	9	#include "torch_npu/csrc/framework/utils/OpPreparation.h"
10	-#include "third_party/acl/inc/acl/acl_base.h"	10	+#include <acl/acl_base.h>
11		11
12	namespace c10_npu {	12	namespace c10_npu {
13	const int g_toAclOffset = 256;	13	const int g_toAclOffset = 256;

Mtorch_npu/csrc/distributed/ProcessGroupHCCL.cpp+2-2

	@@ -33,8 +33,8 @@
33	#include <arpa/inet.h>	33	#include <arpa/inet.h>
34		34
35	#include "op_plugin/OpInterface.h"	35	#include "op_plugin/OpInterface.h"
36	-#include "third_party/acl/inc/acl/acl.h"	36	+#include <acl/acl.h>
37	-#include "third_party/acl/inc/acl/acl_base.h"	37	+#include <acl/acl_base.h>
38	#include "torch_npu/csrc/aten/CustomFunctions.h"	38	#include "torch_npu/csrc/aten/CustomFunctions.h"
39	#include "torch_npu/csrc/aten/NPUNativeFunctions.h"	39	#include "torch_npu/csrc/aten/NPUNativeFunctions.h"
40	#include "torch_npu/csrc/core/npu/GetCANNInfo.h"	40	#include "torch_npu/csrc/core/npu/GetCANNInfo.h"

Mtorch_npu/csrc/distributed/rpc/tensorpipe_agent.cpp+2-2

	@@ -16,7 +16,7 @@
16		16
17	#include "third_party/Tensorpipe/tensorpipe/common/device_id.h"	17	#include "third_party/Tensorpipe/tensorpipe/common/device_id.h"
18	#include "third_party/Tensorpipe/tensorpipe/tensorpipe.h"	18	#include "third_party/Tensorpipe/tensorpipe/tensorpipe.h"
19	-#include "third_party/acl/inc/acl/acl_rt.h"	19	+#include <acl/acl_rt.h>
20	#include "torch_npu/csrc/core/npu/sys_ctrl/npu_sys_ctrl.h"	20	#include "torch_npu/csrc/core/npu/sys_ctrl/npu_sys_ctrl.h"
21	#include "torch_npu/csrc/distributed/rpc/tensorpipe_utils.h"	21	#include "torch_npu/csrc/distributed/rpc/tensorpipe_utils.h"
22	#include "torch_npu/csrc/core/npu/NPUFunctions.h"	22	#include "torch_npu/csrc/core/npu/NPUFunctions.h"
	@@ -362,7 +362,7 @@ void TensorPipeAgent::checkAndSetStaticGroup(const c10::intrusive_ptr<::c10d::St
362	returnedVec = store->compareSet(isStaticGroupKey, std::vector<uint8_t>(), isStaticGroupVec);	362	returnedVec = store->compareSet(isStaticGroupKey, std::vector<uint8_t>(), isStaticGroupVec);
363	std::string returnedVal = std::string(returnedVec.begin(), returnedVec.end());	363	std::string returnedVal = std::string(returnedVec.begin(), returnedVec.end());
364	// In both cases, the returned value should be the value of isStaticGroupStr,	364	// In both cases, the returned value should be the value of isStaticGroupStr,
365	- // otherwise there is a discrepency with initialization among one of the	365	+ // otherwise there is a discrepancy with initialization among one of the
366	// members	366	// members
367	}	367	}
368		368

Mtorch_npu/csrc/distributed/symm_mem/NPUSHMEMInterface.h+1-1

	@@ -3,7 +3,7 @@
3	#include <cstddef>	3	#include <cstddef>
4	#include <cstdint>	4	#include <cstdint>
5	#include "third_party/shmem/include/shmem_host_def.h"	5	#include "third_party/shmem/include/shmem_host_def.h"
6	-#include "third_party/acl/inc/acl/acl_base.h"	6	+#include <acl/acl_base.h>
7		7
8	namespace c10d {	8	namespace c10d {
9	namespace symmetric_memory {	9	namespace symmetric_memory {

Mtorch_npu/csrc/framework/InferFormat.h+2-2

	@@ -4,7 +4,7 @@
4	#include <ATen/ATen.h>	4	#include <ATen/ATen.h>
5		5
6	#include "torch_npu/csrc/framework/utils/NPUDefinition.h"	6	#include "torch_npu/csrc/framework/utils/NPUDefinition.h"
7	-#include "third_party/acl/inc/acl/acl_base.h"	7	+#include <acl/acl_base.h>
8		8
9	namespace at_npu {	9	namespace at_npu {
10	namespace native {	10	namespace native {
	@@ -47,4 +47,4 @@ public:
47	} // namespace native	47	} // namespace native
48	} // namespace at_npu	48	} // namespace at_npu
49		49
50	-#endif // __NATIVE_NPU_UTILS_FORMAT_INFER__	50	+#endif // __NATIVE_NPU_UTILS_FORMAT_INFER__

Mtorch_npu/csrc/framework/LazyInitAclops.h+2-2

	@@ -5,7 +5,7 @@
5	#include <vector>	5	#include <vector>
6	#include <string>	6	#include <string>
7	#include <utility>	7	#include <utility>
8	-#include "third_party/acl/inc/acl/acl_op_compiler.h"	8	+#include <acl/acl_op_compiler.h>
9		9
10	namespace at_npu {	10	namespace at_npu {
11	namespace aclops {	11	namespace aclops {
	@@ -27,4 +27,4 @@ void InitializeJitCompilationMode();
27	} // namespace aclops	27	} // namespace aclops
28	} // namespace at_npu	28	} // namespace at_npu
29		29
30	-#endif // AT_NPU_ACLOPS_LAZYINITACLOPS_H_	30	+#endif // AT_NPU_ACLOPS_LAZYINITACLOPS_H_

Mtorch_npu/csrc/framework/OpCmdHelper.h+3-3

	@@ -4,8 +4,8 @@
4	#include <ATen/ATen.h>	4	#include <ATen/ATen.h>
5		5
6	#include "torch_npu/csrc/framework/utils/NpuUtils.h"	6	#include "torch_npu/csrc/framework/utils/NpuUtils.h"
7	-#include "third_party/acl/inc/acl/acl.h"	7	+#include <acl/acl.h>
8	-#include "third_party/acl/inc/acl/acl_base.h"	8	+#include <acl/acl_base.h>
9		9
10	namespace at_npu {	10	namespace at_npu {
11	namespace native {	11	namespace native {
	@@ -39,4 +39,4 @@ public:
39	} // namespace native	39	} // namespace native
40	} // namespace at_npu	40	} // namespace at_npu
41		41
42	-#endif	42	+#endif

Mtorch_npu/csrc/framework/OpParamMaker.h+2-2

	@@ -4,7 +4,7 @@
4	#include "torch_npu/csrc/core/npu/NPUStream.h"	4	#include "torch_npu/csrc/core/npu/NPUStream.h"
5	#include "torch_npu/csrc/logging/LogContext.h"	5	#include "torch_npu/csrc/logging/LogContext.h"
6		6
7	-#include "third_party/acl/inc/acl/acl_base.h"	7	+#include <acl/acl_base.h>
8	#include "torch_npu/csrc/framework/interface/AclOpCompileInterface.h"	8	#include "torch_npu/csrc/framework/interface/AclOpCompileInterface.h"
9	#include "torch_npu/csrc/framework/NPUDefine.h"	9	#include "torch_npu/csrc/framework/NPUDefine.h"
10	#include "torch_npu/csrc/framework/utils/ForceJitCompileList.h"	10	#include "torch_npu/csrc/framework/utils/ForceJitCompileList.h"
	@@ -451,4 +451,4 @@ extern std::recursive_mutex deterministic_launch_mutex;
451	} // namespace native	451	} // namespace native
452	} // namespace at_npu	452	} // namespace at_npu
453		453
454	-#endif	454	+#endif

Mtorch_npu/csrc/framework/aoe/AoeDumpGraphManager.h+2-2

	@@ -2,7 +2,7 @@
2	#define __NATIVE_NPU_TOOLS_AOEUTILS__	2	#define __NATIVE_NPU_TOOLS_AOEUTILS__
3		3
4	#include <unordered_set>	4	#include <unordered_set>
5	-#include <third_party/acl/inc/acl/acl_op_compiler.h>	5	+#include <acl/acl_op_compiler.h>
6	#include "torch_npu/csrc/core/npu/NPUException.h"	6	#include "torch_npu/csrc/core/npu/NPUException.h"
7		7
8	namespace at_npu {	8	namespace at_npu {
	@@ -147,4 +147,4 @@ AoeDumpGraphManager& aoe_manager();
147	} // namespace native	147	} // namespace native
148	} // namespace at_npu	148	} // namespace at_npu
149		149
150	-#endif // __NATIVE_NPU_TOOLS_AOEUTILS__	150	+#endif // __NATIVE_NPU_TOOLS_AOEUTILS__

Mtorch_npu/csrc/framework/contiguous/ContiguousUtils.h+2-2

	@@ -3,7 +3,7 @@
3		3
4	#include <c10/util/SmallVector.h>	4	#include <c10/util/SmallVector.h>
5		5
6	-#include "third_party/acl/inc/acl/acl_base.h"	6	+#include <acl/acl_base.h>
7	#include "torch_npu/csrc/framework/utils/NpuUtils.h"	7	#include "torch_npu/csrc/framework/utils/NpuUtils.h"
8	#include "torch_npu/csrc/framework/utils/NPUDefinition.h"	8	#include "torch_npu/csrc/framework/utils/NPUDefinition.h"
9		9
	@@ -40,4 +40,4 @@ struct ContiguousTensorDesc {
40	} // namespace native	40	} // namespace native
41	} // namespace at_npu	41	} // namespace at_npu
42		42
43	-#endif	43	+#endif

Mtorch_npu/csrc/framework/interface/AclInterface.h+5-5

	@@ -1,10 +1,10 @@
1	#ifndef __TORCH_NPU_INTERFACE_ACLINTERFACE__	1	#ifndef __TORCH_NPU_INTERFACE_ACLINTERFACE__
2	#define __TORCH_NPU_INTERFACE_ACLINTERFACE__	2	#define __TORCH_NPU_INTERFACE_ACLINTERFACE__
3		3
4	-#include "third_party/acl/inc/acl/acl_rt.h"	4	+#include <acl/acl_rt.h>
5	-#include <third_party/acl/inc/acl/acl_base.h>	5	+#include <acl/acl_base.h>
6	-#include <third_party/acl/inc/acl/acl_prof.h>	6	+#include <acl/acl_prof.h>
7	-#include <third_party/acl/inc/acl/acl_op.h>	7	+#include <acl/acl_op.h>
8		8
9	namespace at_npu {	9	namespace at_npu {
10	namespace native {	10	namespace native {
	@@ -76,4 +76,4 @@ aclError AclopStopDumpArgs(uint32_t dumpType);
76	} // namespace native	76	} // namespace native
77	} // namespace at_npu	77	} // namespace at_npu
78		78
79	-#endif // __TORCH_NPU_INTERFACE_ACLINTERFACE__	79	+#endif // __TORCH_NPU_INTERFACE_ACLINTERFACE__

Mtorch_npu/csrc/framework/interface/AclOpCompileInterface.cpp+2-2

	@@ -5,7 +5,7 @@
5	#include "torch_npu/csrc/core/npu/register/FunctionLoader.h"	5	#include "torch_npu/csrc/core/npu/register/FunctionLoader.h"
6	#include "torch_npu/csrc/framework/interface/AclOpCompileInterface.h"	6	#include "torch_npu/csrc/framework/interface/AclOpCompileInterface.h"
7	#include "torch_npu/csrc/core/npu/register/OptionsManager.h"	7	#include "torch_npu/csrc/core/npu/register/OptionsManager.h"
8	-#include "third_party/acl/inc/acl/acl_base.h"	8	+#include <acl/acl_base.h>
9		9
10	namespace at_npu	10	namespace at_npu
11	{	11	{
	@@ -207,4 +207,4 @@ aclError AclDestroyAclOpExecutor(aclOpExecutor *executor)
207	}	207	}
208		208
209	} // namespace native	209	} // namespace native
210	-} // namespace at_npu	210	+} // namespace at_npu

Mtorch_npu/csrc/framework/interface/AclOpCompileInterface.h+2-2

	@@ -1,7 +1,7 @@
1	#ifndef __PLUGIN_NATIVE_NPU_INTERFACE_ACLOPCOMPILE__	1	#ifndef __PLUGIN_NATIVE_NPU_INTERFACE_ACLOPCOMPILE__
2	#define __PLUGIN_NATIVE_NPU_INTERFACE_ACLOPCOMPILE__	2	#define __PLUGIN_NATIVE_NPU_INTERFACE_ACLOPCOMPILE__
3	#include <c10/util/Optional.h>	3	#include <c10/util/Optional.h>
4	-#include "third_party/acl/inc/acl/acl_op_compiler.h"	4	+#include <acl/acl_op_compiler.h>
5		5
6	typedef struct aclOpExecutor aclOpExecutor;	6	typedef struct aclOpExecutor aclOpExecutor;
7		7
	@@ -150,4 +150,4 @@ ACL_FUNC_VISIBILITY aclError AclDestroyAclOpExecutor(aclOpExecutor *executor);
150	} // namespace native	150	} // namespace native
151	} // namespace at_npu	151	} // namespace at_npu
152		152
153	-#endif // __NATIVE_NPU_INTERFACE_ACLOPCOMPILE__	153	+#endif // __NATIVE_NPU_INTERFACE_ACLOPCOMPILE__

Mtorch_npu/csrc/framework/interface/EnvVariables.cpp+1-1

	@@ -7,7 +7,7 @@
7	#include <string>	7	#include <string>
8	#include "torch_npu/csrc/core/npu/NPUException.h"	8	#include "torch_npu/csrc/core/npu/NPUException.h"
9		9
10	-#include "third_party/acl/inc/acl/acl_mdl.h"	10	+#include <acl/acl_mdl.h>
11	#include "torch_npu/csrc/framework/utils/ForceJitCompileList.h"	11	#include "torch_npu/csrc/framework/utils/ForceJitCompileList.h"
12	#include "torch_npu/csrc/framework/utils/ForceAclnnList.h"	12	#include "torch_npu/csrc/framework/utils/ForceAclnnList.h"
13	#include "torch_npu/csrc/framework/interface/AclOpCompileInterface.h"	13	#include "torch_npu/csrc/framework/interface/AclOpCompileInterface.h"

Mtorch_npu/csrc/framework/interface/MsProfilerInterface.cpp+1-1

	@@ -1,7 +1,7 @@
1	#include "torch_npu/csrc/framework/interface/MsProfilerInterface.h"	1	#include "torch_npu/csrc/framework/interface/MsProfilerInterface.h"
2	#include "torch_npu/csrc/core/npu/NPUException.h"	2	#include "torch_npu/csrc/core/npu/NPUException.h"
3	#include "torch_npu/csrc/core/npu/register/FunctionLoader.h"	3	#include "torch_npu/csrc/core/npu/register/FunctionLoader.h"
4	-#include "third_party/acl/inc/acl/acl_prof.h"	4	+#include <acl/acl_prof.h>
5		5
6	namespace at_npu {	6	namespace at_npu {
7	namespace native {	7	namespace native {

Mtorch_npu/csrc/framework/interface/MsProfilerInterface.h+1-1

	@@ -1,7 +1,7 @@
1	#ifndef __TORCH_NPU_MSPROFILERINTERFACE__	1	#ifndef __TORCH_NPU_MSPROFILERINTERFACE__
2	#define __TORCH_NPU_MSPROFILERINTERFACE__	2	#define __TORCH_NPU_MSPROFILERINTERFACE__
3		3
4	-#include <third_party/acl/inc/acl/acl_prof.h>	4	+#include <acl/acl_prof.h>
5	#include "torch_npu/csrc/core/npu/NPUException.h"	5	#include "torch_npu/csrc/core/npu/NPUException.h"
6		6
7	namespace at_npu {	7	namespace at_npu {

Mtorch_npu/csrc/framework/utils/CalcuOpUtil.cpp+2-2

	@@ -1,7 +1,7 @@
1	#include <ATen/record_function.h>	1	#include <ATen/record_function.h>
2		2
3	-#include "third_party/acl/inc/acl/acl_base.h"	3	+#include <acl/acl_base.h>
4	-#include "third_party/acl/inc/acl/acl_rt.h"	4	+#include <acl/acl_rt.h>
5	#include "torch_npu/csrc/aten/mirror/NPUMemoryOverlap.h"	5	#include "torch_npu/csrc/aten/mirror/NPUMemoryOverlap.h"
6	#include "torch_npu/csrc/core/NPUBridge.h"	6	#include "torch_npu/csrc/core/NPUBridge.h"
7	#include "torch_npu/csrc/core/NPUStorageImpl.h"	7	#include "torch_npu/csrc/core/NPUStorageImpl.h"

Mtorch_npu/csrc/framework/utils/CalcuOpUtil.h+2-2

	@@ -14,8 +14,8 @@
14	#include "torch_npu/csrc/framework/utils/NpuUtils.h"	14	#include "torch_npu/csrc/framework/utils/NpuUtils.h"
15	#include "torch_npu/csrc/core/npu/interface/AclInterface.h"	15	#include "torch_npu/csrc/core/npu/interface/AclInterface.h"
16	#include "torch_npu/csrc/core/npu/npu_log.h"	16	#include "torch_npu/csrc/core/npu/npu_log.h"
17	-#include "third_party/acl/inc/acl/acl_base.h"	17	+#include <acl/acl_base.h>
18	-#include "third_party/acl/inc/acl/acl.h"	18	+#include <acl/acl.h>
19		19
20	using std::string;	20	using std::string;
21	using std::vector;	21	using std::vector;

Mtorch_npu/csrc/framework/utils/NPUDefinition.h+2-2

	@@ -3,7 +3,7 @@
3		3
4		4
5	#include <c10/util/SmallVector.h>	5	#include <c10/util/SmallVector.h>
6	-#include <third_party/acl/inc/graph/operator.h>	6	+#include <graph/operator.h>
7		7
8	#include <functional>	8	#include <functional>
9	#include <vector>	9	#include <vector>
	@@ -21,4 +21,4 @@ using DynamicInputRegFunc = std::function<ge::OperatorPtr(DyNumAndIndex, std::st
21	} // namespace native	21	} // namespace native
22	} // namespace at_npu	22	} // namespace at_npu
23		23
24	-#endif // __NATIVE_NPU_UTILS_NPU_CONFIG__	24	+#endif // __NATIVE_NPU_UTILS_NPU_CONFIG__

Mtorch_npu/csrc/framework/utils/NpuUtils.h+3-3

	@@ -7,9 +7,9 @@
7	#include <ATen/ATen.h>	7	#include <ATen/ATen.h>
8	#include "torch_npu/csrc/core/npu/npu_log.h"	8	#include "torch_npu/csrc/core/npu/npu_log.h"
9		9
10	-#include "third_party/acl/inc/acl/acl.h"	10	+#include <acl/acl.h>
11	-#include "third_party/acl/inc/acl/acl_base.h"	11	+#include <acl/acl_base.h>
12	-#include "third_party/acl/inc/acl/acl_op.h"	12	+#include <acl/acl_op.h>
13		13
14	#include "torch_npu/csrc/core/npu/interface/AsyncTaskQueueInterface.h"	14	#include "torch_npu/csrc/core/npu/interface/AsyncTaskQueueInterface.h"
15	#include "torch_npu/csrc/framework/interface/AclOpCompileInterface.h"	15	#include "torch_npu/csrc/framework/interface/AclOpCompileInterface.h"

Mtorch_npu/csrc/inductor/aoti_runtime/device_utils.h+2-2

	@@ -2,8 +2,8 @@
2		2
3	#ifdef USE_NPU	3	#ifdef USE_NPU
4		4
5	-#include "third_party/acl/inc/acl/acl_base.h"	5	+#include <acl/acl_base.h>
6	-#include "third_party/acl/inc/acl/acl_rt.h"	6	+#include <acl/acl_rt.h>
7		7
8	typedef void* NPUdeviceptr;	8	typedef void* NPUdeviceptr;
9		9

Mtorch_npu/csrc/inductor/aoti_runtime/utils_npu.h+1-1

	@@ -5,7 +5,7 @@
5	// in model.so, and should not refer to any aten/c10 headers except the stable	5	// in model.so, and should not refer to any aten/c10 headers except the stable
6	// C ABI defined in torch_npu/csrc/inductor/aoti_torch/c/shim.h. The same rule	6	// C ABI defined in torch_npu/csrc/inductor/aoti_torch/c/shim.h. The same rule
7	// applies to other files under torch_npu/csrc/inductor/aoti_runtime/.	7	// applies to other files under torch_npu/csrc/inductor/aoti_runtime/.
8	-#include <third_party/acl/inc/acl/acl_base.h>	8	+#include <acl/acl_base.h>
9	#include <torch/csrc/inductor/aoti_runtime/utils.h>	9	#include <torch/csrc/inductor/aoti_runtime/utils.h>
10	#include <torch_npu/csrc/inductor/aoti_torch/c/shim_npu.h>	10	#include <torch_npu/csrc/inductor/aoti_torch/c/shim_npu.h>
11		11

Mtorch_npu/csrc/inductor/mlir/cpp_common.cpp+2-2

	@@ -8,8 +8,8 @@
8	#include <torch_npu/csrc/framework/OpCommand.h>	8	#include <torch_npu/csrc/framework/OpCommand.h>
9	#include <torch_npu/csrc/profiler/profiler_mgr.h>	9	#include <torch_npu/csrc/profiler/profiler_mgr.h>
10		10
11	-#include "third_party/acl/inc/profiling/prof_api.h"	11	+#include <profiling/prof_api.h>
12	-#include "third_party/acl/inc/profiling/prof_common.h"	12	+#include <profiling/prof_common.h>
13		13
14	struct TilingMem {	14	struct TilingMem {
15	std::unique_ptr<void, decltype(&aclrtFreeHost)> arg_tiling_host;	15	std::unique_ptr<void, decltype(&aclrtFreeHost)> arg_tiling_host;

Mtorch_npu/csrc/inductor/mlir/cpp_common.h+3-3

	@@ -1,8 +1,8 @@
1	#ifndef BUILD_LIBTORCH	1	#ifndef BUILD_LIBTORCH
2	#include <Python.h>	2	#include <Python.h>
3	#include <functional>	3	#include <functional>
4	-#include "third_party/acl/inc/acl/acl_base.h"	4	+#include <acl/acl_base.h>
5	-#include "third_party/acl/inc/acl/acl_rt.h"	5	+#include <acl/acl_rt.h>
6	#include "torch_npu/csrc/inductor/mlir/hacl_rt.h"	6	#include "torch_npu/csrc/inductor/mlir/hacl_rt.h"
7		7
8	rtError_t common_launch(char* kernelName, const void* func, uint32_t gridX, void* args, uint32_t argsSize,	8	rtError_t common_launch(char* kernelName, const void* func, uint32_t gridX, void* args, uint32_t argsSize,
	@@ -10,4 +10,4 @@ rtError_t common_launch(char* kernelName, const void* func, uint32_t gridX, void
10	rtError_t common_launch_dyn(char* kernelName, void* func, void* tiling_func, int64_t tilingSize, void* arg_tiling_host,	10	rtError_t common_launch_dyn(char* kernelName, void* func, void* tiling_func, int64_t tilingSize, void* arg_tiling_host,
11	void* arg_tiling_device, uint32_t gridX, void* args, uint32_t argsSize, rtStream_t stream);	11	void* arg_tiling_device, uint32_t gridX, void* args, uint32_t argsSize, rtStream_t stream);
12	void opcommand_call(const char* name, std::function<int()> launch_call);	12	void opcommand_call(const char* name, std::function<int()> launch_call);
13	-#endif // BUILD_LIBTORCH	13	+#endif // BUILD_LIBTORCH

Mtorch_npu/csrc/ipc/NPUIPCTypes.cpp+3-3

	@@ -8,8 +8,8 @@
8	#include "torch_npu/csrc/core/npu/interface/AsyncTaskQueueInterface.h"	8	#include "torch_npu/csrc/core/npu/interface/AsyncTaskQueueInterface.h"
9	#include "torch_npu/csrc/ipc/NPUIPCTypes.h"	9	#include "torch_npu/csrc/ipc/NPUIPCTypes.h"
10		10
11	-#include "third_party/acl/inc/acl/acl_base.h"	11	+#include <acl/acl_base.h>
12	-#include "third_party/acl/inc/acl/acl_rt.h"	12	+#include <acl/acl_rt.h>
13		13
14	namespace torch_npu {	14	namespace torch_npu {
15	namespace ipc {	15	namespace ipc {
	@@ -264,4 +264,4 @@ namespace NPUCachingAllocator {
264	REGISTER_FREE_MEMORY_CALLBACK("npu_ipc_collect", NpuIPCCollectCallback);	264	REGISTER_FREE_MEMORY_CALLBACK("npu_ipc_collect", NpuIPCCollectCallback);
265		265
266	} // namespace NPUCachingAllocator	266	} // namespace NPUCachingAllocator
267	-} // namespace c10_npu	267	+} // namespace c10_npu

Mtorch_npu/csrc/ipc/StorageSharing.cpp+2-2

	@@ -22,8 +22,8 @@
22	#include "torch_npu/csrc/ipc/NPUIPCTypes.h"	22	#include "torch_npu/csrc/ipc/NPUIPCTypes.h"
23	#include "torch_npu/csrc/ipc/StorageSharing.h"	23	#include "torch_npu/csrc/ipc/StorageSharing.h"
24		24
25	-#include "third_party/acl/inc/acl/acl_base.h"	25	+#include <acl/acl_base.h>
26	-#include "third_party/acl/inc/acl/acl_rt.h"	26	+#include <acl/acl_rt.h>
27		27
28	namespace torch_npu {	28	namespace torch_npu {
29	namespace reductions {	29	namespace reductions {

Mtorch_npu/csrc/npu/Graph.h+3-3

	@@ -5,9 +5,9 @@
5	#include <torch/csrc/jit/python/pybind_utils.h>	5	#include <torch/csrc/jit/python/pybind_utils.h>
6	#include <torch/csrc/utils/pybind.h>	6	#include <torch/csrc/utils/pybind.h>
7		7
8	-#include "third_party/acl/inc/acl/acl_base.h"	8	+#include <acl/acl_base.h>
9	-#include "third_party/acl/inc/acl/acl_rt.h"	9	+#include <acl/acl_rt.h>
10	-#include "third_party/acl/inc/acl/super_kernel.h"	10	+#include <acl/super_kernel.h>
11		11
12	struct PendingTensorData {	12	struct PendingTensorData {
13	PendingTensorData(uintptr_t dataPtr, Py_ssize_t nbytes, PyObject* shape, PyObject* dtype)	13	PendingTensorData(uintptr_t dataPtr, Py_ssize_t nbytes, PyObject* shape, PyObject* dtype)

Mtorch_npu/csrc/npu/Module.cpp+1-1

	@@ -27,7 +27,7 @@
27	#include <torch_npu/csrc/inductor/aoti_runner/pybind.h>	27	#include <torch_npu/csrc/inductor/aoti_runner/pybind.h>
28		28
29	#include <op_plugin/utils/custom_functions/opapi/FFTCommonOpApi.h>	29	#include <op_plugin/utils/custom_functions/opapi/FFTCommonOpApi.h>
30	-#include <third_party/acl/inc/acl/acl.h>	30	+#include <acl/acl.h>
31	#include <third_party/fmt/include/fmt/format.h>	31	#include <third_party/fmt/include/fmt/format.h>
32	#include <torch_npu/csrc/aten/NPUGeneratorImpl.h>	32	#include <torch_npu/csrc/aten/NPUGeneratorImpl.h>
33	#include <torch_npu/csrc/aten/NPUNativeFunctions.h>	33	#include <torch_npu/csrc/aten/NPUNativeFunctions.h>

Mtorch_npu/csrc/npu/Stream.cpp+3-3

	@@ -3,9 +3,9 @@
3	#include <torch/csrc/Device.h>	3	#include <torch/csrc/Device.h>
4	#include <torch/csrc/THP.h>	4	#include <torch/csrc/THP.h>
5		5
6	-#include "third_party/acl/inc/acl/acl.h"	6	+#include <acl/acl.h>
7	-#include "third_party/acl/inc/acl/acl_base.h"	7	+#include <acl/acl_base.h>
8	-#include "third_party/acl/inc/acl/acl_rt.h"	8	+#include <acl/acl_rt.h>
9	#include "torch_npu/csrc/core/npu/NPUGuard.h"	9	#include "torch_npu/csrc/core/npu/NPUGuard.h"
10	#include "torch_npu/csrc/core/npu/sys_ctrl/npu_sys_ctrl.h"	10	#include "torch_npu/csrc/core/npu/sys_ctrl/npu_sys_ctrl.h"
11	#include "torch_npu/csrc/npu/Module.h"	11	#include "torch_npu/csrc/npu/Module.h"

Mtorch_npu/csrc/profiler/feature_mgr.h+1-1

	@@ -7,7 +7,7 @@
7	#include <vector>	7	#include <vector>
8	#include <tuple>	8	#include <tuple>
9		9
10	-#include "third_party/acl/inc/acl/acl_prof.h"	10	+#include <acl/acl_prof.h>
11		11
12	#include "torch_npu/csrc/toolkit/profiler/common/singleton.h"	12	#include "torch_npu/csrc/toolkit/profiler/common/singleton.h"
13	#include "torch_npu/csrc/toolkit/profiler/common/utils.h"	13	#include "torch_npu/csrc/toolkit/profiler/common/utils.h"

Mtorch_npu/csrc/profiler/npu_profiler.h+2-2

	@@ -7,8 +7,8 @@
7		7
8	#include <ATen/record_function.h>	8	#include <ATen/record_function.h>
9		9
10	-#include "third_party/acl/inc/acl/acl_base.h"	10	+#include <acl/acl_base.h>
11	-#include "third_party/acl/inc/acl/acl_rt.h"	11	+#include <acl/acl_rt.h>
12		12
13	#include "torch_npu/csrc/toolkit/profiler/inc/data_reporter.h"	13	#include "torch_npu/csrc/toolkit/profiler/inc/data_reporter.h"
14	#include "torch_npu/csrc/profiler/profiler_mgr.h"	14	#include "torch_npu/csrc/profiler/profiler_mgr.h"

Mtorch_npu/csrc/profiler/profiler_mgr.h+1-1

	@@ -4,7 +4,7 @@
4	#include <mutex>	4	#include <mutex>
5	#include <map>	5	#include <map>
6		6
7	-#include "third_party/acl/inc/acl/acl_prof.h"	7	+#include <acl/acl_prof.h>
8		8
9	#include "torch_npu/csrc/toolkit/profiler/common/singleton.h"	9	#include "torch_npu/csrc/toolkit/profiler/common/singleton.h"
10	#include "torch_npu/csrc/toolkit/profiler/inc/data_dumper.h"	10	#include "torch_npu/csrc/toolkit/profiler/inc/data_dumper.h"

Mtorch_npu/csrc/profiler/profiler_npu.cpp+1-1

	@@ -7,7 +7,7 @@
7	#include "torch_npu/csrc/core/npu/NPUStream.h"	7	#include "torch_npu/csrc/core/npu/NPUStream.h"
8	#include "torch_npu/csrc/core/npu/NPUGuard.h"	8	#include "torch_npu/csrc/core/npu/NPUGuard.h"
9	#include "torch_npu/csrc/core/npu/interface/AclInterface.h"	9	#include "torch_npu/csrc/core/npu/interface/AclInterface.h"
10	-#include "third_party/acl/inc/acl/acl_rt.h"	10	+#include <acl/acl_rt.h>
11		11
12	namespace torch_npu {	12	namespace torch_npu {
13	namespace profiler {	13	namespace profiler {

Mtorch_npu/utils/cpp_extension.py+0-2

	@@ -1,7 +1,6 @@
1	import os	1	import os
2	import setuptools	2	import setuptools
3		3
4	-import torch
5	import torch.utils.cpp_extension as TorchExtension	4	import torch.utils.cpp_extension as TorchExtension
6		5
7	import torch_npu	6	import torch_npu
	@@ -38,7 +37,6 @@ def NpuExtension(name, sources, args, *kwargs):
38	torch_npu_dir = PYTORCH_NPU_INSTALL_PATH	37	torch_npu_dir = PYTORCH_NPU_INSTALL_PATH
39	include_dirs = kwargs.get('include_dirs', [])	38	include_dirs = kwargs.get('include_dirs', [])
40	include_dirs.append(os.path.join(torch_npu_dir, 'include'))	39	include_dirs.append(os.path.join(torch_npu_dir, 'include'))
41	- include_dirs.append(os.path.join(torch_npu_dir, 'include', 'third_party', 'acl', 'inc'))
42	include_dirs.append(os.path.join(torch_npu_dir, 'include', 'third_party', 'hccl', 'inc'))	40	include_dirs.append(os.path.join(torch_npu_dir, 'include', 'third_party', 'hccl', 'inc'))
43	include_dirs += TorchExtension.include_paths()	41	include_dirs += TorchExtension.include_paths()
44	kwargs['include_dirs'] = include_dirs	42	kwargs['include_dirs'] = include_dirs