bug fix: experimental 构建报错 'multiple definition of `ops::op_impl_register_infershape_ZerosLike` + gen_compile

已合并

bug fix: experimental 构建报错 'multiple definition of `ops::op_impl_register_infershape_ZerosLike` + gen_compile_option.sh awk 重复告警 #3516

zhaohujie创建于 6月24日从zhaohujie/ops-math： bug/zeros-like 合入到 cann/ops-math： master

bug fix: experimental 构建报错 'multiple definition of `ops::op_impl_register_infershape_ZerosLike` + gen_compile_option.sh awk 重复告警 #3516

已合并

zhaohujie创建于 6月24日

从zhaohujie/ops-math bug/zeros-like合入到cann/ops-math master

共 3 个文件变更+52-107

Mexperimental/math/isin_part_v1/CMakeLists.txt+0-1

	@@ -12,7 +12,6 @@
12	add_all_modules_sources(	12	add_all_modules_sources(
13	OPTYPE isin_part_v1	13	OPTYPE isin_part_v1
14	ACLNNTYPE aclnn	14	ACLNNTYPE aclnn
15	- DEPENDENCIES concat_d sort zeros_like zero_op cast
16	COMPUTE_UNIT ascend950	15	COMPUTE_UNIT ascend950
17	TILING_DIR arch35	16	TILING_DIR arch35
18	DISABLE_IN_OPP TRUE	17	DISABLE_IN_OPP TRUE

Mexperimental/math/isin_part_v1/examples/test_aclnn_isin_part_v1.cpp+36-98

	@@ -1,19 +1,15 @@
1	/**	1	/**
2	- * Copyright (c) 2026 Huawei Technologies Co., Ltd.	2	+ * Copyright (c) 2026 Huawei Technologies Co., Ltd.
3	- * This program is free software, you can redistribute it and/or modify it under the terms and conditions of	3	+ * This program is free software, you can redistribute it and/or modify it under the terms and conditions of
4	- * CANN Open Software License Agreement Version 2.0 (the "License").	4	+ * CANN Open Software License Agreement Version 2.0 (the "License").
5	- * Please refer to the License for details. You may not use this file except in compliance with the License.	5	+ * Please refer to the License for details. You may not use this file except in compliance with the License.
6	- * THIS SOFTWARE IS PROVIDED ON AN "AS IS" BASIS, WITHOUT WARRANTIES OF ANY KIND, EITHER EXPRESS OR IMPLIED,	6	+ * THIS SOFTWARE IS PROVIDED ON AN "AS IS" BASIS, WITHOUT WARRANTIES OF ANY KIND, EITHER EXPRESS OR IMPLIED,
7	- * INCLUDING BUT NOT LIMITED TO NON-INFRINGEMENT, MERCHANTABILITY, OR FITNESS FOR A PARTICULAR PURPOSE.	7	+ * INCLUDING BUT NOT LIMITED TO NON-INFRINGEMENT, MERCHANTABILITY, OR FITNESS FOR A PARTICULAR PURPOSE.
8	- * See LICENSE in the root of the software repository for the full text of the License.	8	+ * See LICENSE in the root of the software repository for the full text of the License.
9	- */	9	+ */
10	#include <iostream>	10	#include <iostream>
11	#include <vector>	11	#include <vector>
12	-#include <algorithm>
13	#include "acl/acl.h"	12	#include "acl/acl.h"
14	-#include "aclnnop/aclnn_cat.h"
15	-#include "aclnnop/aclnn_sort.h"
16	-#include "aclnnop/aclnn_cast.h"
17	#include "aclnnop/aclnn_isin_part_v1.h"	13	#include "aclnnop/aclnn_isin_part_v1.h"
18		14
19	#define CHECK_RET(cond, return_expr) \	15	#define CHECK_RET(cond, return_expr) \
	@@ -107,76 +103,35 @@ int main()
107	CHECK_RET(aclrtCreateContext(&context, deviceId) == ACL_SUCCESS, return -1);	103	CHECK_RET(aclrtCreateContext(&context, deviceId) == ACL_SUCCESS, return -1);
108	CHECK_RET(aclrtCreateStream(&stream) == ACL_SUCCESS, return -1);	104	CHECK_RET(aclrtCreateStream(&stream) == ACL_SUCCESS, return -1);
109		105
110	- // 2. 原始输入数据	106	+ // 2. 直接构造 IsinPartV1 的三个输入，仅校验本算子（不依赖 Cat/Sort/Cast 等其它算子）。
111	- std::vector<float> h_elements = {1.0, 1.0, 4.0, 1.0, 1.0};	107	+ // 语义等价于 isin(elements=[1,1,4,1,1], test_elements=[0,1,2,3])：
112	- std::vector<float> h_test_elements = {0.0, 1.0, 2.0, 3.0};	108	+ // - 将 elements 与 test_elements 拼接后按值排序，相等值相邻；
113	- int64_t elementsNum = h_elements.size();	109	+ // - value 为排序后的值，index 为其在拼接序列中的原始位置（0..elementsNum-1 为待判定元素）；
114	- int64_t totalSize = h_elements.size() + h_test_elements.size();	110	+ // - elementsNum 为待判定元素个数（单元素 INT32 张量）；
		111	+ // - 输出 z[i] 表示第 i 个待判定元素是否出现在 test_elements 中。
		112	+ std::vector<float> h_value = {0, 1, 1, 1, 1, 1, 2, 3, 4}; // 排序后的值
		113	+ std::vector<int32_t> h_index = {5, 0, 1, 3, 4, 6, 7, 8, 2}; // 对应的原始位置
		114	+ int32_t elementsNum = 5;
		115	+ std::vector<int32_t> h_elementsNum = {elementsNum};
		116	+ int64_t totalSize = static_cast<int64_t>(h_value.size());
115		117
116	std::vector<int64_t> totalShape = {totalSize};	118	std::vector<int64_t> totalShape = {totalSize};
117	std::vector<int64_t> outShape = {elementsNum};	119	std::vector<int64_t> outShape = {elementsNum};
118		120
119	// 3. 资源准备	121	// 3. 资源准备
120	- void elAddr = nullptr, teAddr = nullptr, *catAddr = nullptr;	122	+ void valueAddr = nullptr, indexAddr = nullptr, elementsNumAddr = nullptr, outAddr = nullptr;
121	- void sortVAddr = nullptr, sortI64Addr = nullptr, sortI32Addr = nullptr, outAddr = nullptr;	123	+ aclTensor valueTensor = nullptr, indexTensor = nullptr, elementsNumTensor = nullptr, outTensor = nullptr;
122	- aclTensor elTensor = nullptr, teTensor = nullptr, *catTensor = nullptr;	124	+ CHECK_RET(CreateAclTensor(h_value, totalShape, &valueAddr, ACL_FLOAT, &valueTensor) == 0, return -1);
123	- aclTensor sortVTensor = nullptr, sortI64Tensor = nullptr, sortI32Tensor = nullptr, outTensor = nullptr;	125	+ CHECK_RET(CreateAclTensor(h_index, totalShape, &indexAddr, ACL_INT32, &indexTensor) == 0, return -1);
124	-	126	+ CHECK_RET(CreateAclTensor(h_elementsNum, {1}, &elementsNumAddr, ACL_INT32, &elementsNumTensor) == 0, return -1);
125	- CHECK_RET(CreateAclTensor(h_elements, {elementsNum}, &elAddr, ACL_FLOAT, &elTensor) == 0, return -1);
126	- CHECK_RET(
127	- CreateAclTensor(h_test_elements, {(int64_t)h_test_elements.size()}, &teAddr, ACL_FLOAT, &teTensor) == 0,
128	- return -1);
129	-
130	- // 创建输入 Tensor
131	- CHECK_RET(CreateEmptyTensor(totalShape, ACL_FLOAT, &catAddr, &catTensor) == 0, return -1);
132	- CHECK_RET(CreateEmptyTensor(totalShape, ACL_FLOAT, &sortVAddr, &sortVTensor) == 0, return -1);
133	- CHECK_RET(CreateEmptyTensor(totalShape, ACL_INT64, &sortI64Addr, &sortI64Tensor) == 0, return -1);
134	- CHECK_RET(CreateEmptyTensor(totalShape, ACL_INT32, &sortI32Addr, &sortI32Tensor) == 0, return -1);
135	CHECK_RET(CreateEmptyTensor(outShape, ACL_BOOL, &outAddr, &outTensor) == 0, return -1);	127	CHECK_RET(CreateEmptyTensor(outShape, ACL_BOOL, &outAddr, &outTensor) == 0, return -1);
136		128
137	- // 4. 调用 aclnnCat	129	+ // 4. 调用自定义算子 aclnnIsinPartV1
138	- LOG_PRINT("[Process] Step A: aclnnCat\n");	130	+ LOG_PRINT("[Process] aclnnIsinPartV1\n");
139	- std::vector<aclTensor*> catListVec = {elTensor, teTensor};
140	- aclTensorList* catList = aclCreateTensorList(catListVec.data(), catListVec.size());
141	- uint64_t wsCat = 0;
142	- aclOpExecutor* execCat = nullptr;
143	- CHECK_RET(aclnnCatGetWorkspaceSize(catList, 0, catTensor, &wsCat, &execCat) == ACL_SUCCESS, return -1);
144	- void* wsCatAddr = nullptr;
145	- if (wsCat > 0)
146	- aclrtMalloc(&wsCatAddr, wsCat, ACL_MEM_MALLOC_HUGE_FIRST);
147	- CHECK_RET(aclnnCat(wsCatAddr, wsCat, execCat, stream) == ACL_SUCCESS, return -1);
148	-
149	- // 5. 调用 aclnnSort (输出为 INT64)
150	- LOG_PRINT("[Process] Step B: aclnnSort (INT64 Indices)\n");
151	- uint64_t wsSort = 0;
152	- aclOpExecutor* execSort = nullptr;
153	- CHECK_RET(
154	- aclnnSortGetWorkspaceSize(catTensor, false, 0, false, sortVTensor, sortI64Tensor, &wsSort, &execSort) ==
155	- ACL_SUCCESS,
156	- return -1);
157	- void* wsSortAddr = nullptr;
158	- if (wsSort > 0)
159	- aclrtMalloc(&wsSortAddr, wsSort, ACL_MEM_MALLOC_HUGE_FIRST);
160	- CHECK_RET(aclnnSort(wsSortAddr, wsSort, execSort, stream) == ACL_SUCCESS, return -1);
161	-
162	- // 6. 调用 aclnnCast (INT64 -> INT32)
163	- LOG_PRINT("[Process] Step C: aclnnCast (INT64 -> INT32)\n");
164	- uint64_t wsCast = 0;
165	- aclOpExecutor* execCast = nullptr;
166	- CHECK_RET(
167	- aclnnCastGetWorkspaceSize(sortI64Tensor, ACL_INT32, sortI32Tensor, &wsCast, &execCast) == ACL_SUCCESS,
168	- return -1);
169	- void* wsCastAddr = nullptr;
170	- if (wsCast > 0)
171	- aclrtMalloc(&wsCastAddr, wsCast, ACL_MEM_MALLOC_HUGE_FIRST);
172	- CHECK_RET(aclnnCast(wsCastAddr, wsCast, execCast, stream) == ACL_SUCCESS, return -1);
173	-
174	- // 7. 调用自定义算子 aclnnIsinPartV1
175	- LOG_PRINT("[Process] Step D: aclnnIsinPartV1\n");
176	uint64_t wsIsin = 0;	131	uint64_t wsIsin = 0;
177	aclOpExecutor* execIsin = nullptr;	132	aclOpExecutor* execIsin = nullptr;
178	CHECK_RET(	133	CHECK_RET(
179	- aclnnIsinPartV1GetWorkspaceSize(sortVTensor, sortI32Tensor, elementsNum, outTensor, &wsIsin, &execIsin) ==	134	+ aclnnIsinPartV1GetWorkspaceSize(valueTensor, indexTensor, elementsNumTensor, outTensor, &wsIsin, &execIsin) ==
180	ACL_SUCCESS,	135	ACL_SUCCESS,
181	return -1);	136	return -1);
182	void* wsIsinAddr = nullptr;	137	void* wsIsinAddr = nullptr;
	@@ -184,38 +139,21 @@ int main()
184	aclrtMalloc(&wsIsinAddr, wsIsin, ACL_MEM_MALLOC_HUGE_FIRST);	139	aclrtMalloc(&wsIsinAddr, wsIsin, ACL_MEM_MALLOC_HUGE_FIRST);
185	CHECK_RET(aclnnIsinPartV1(wsIsinAddr, wsIsin, execIsin, stream) == ACL_SUCCESS, return -1);	140	CHECK_RET(aclnnIsinPartV1(wsIsinAddr, wsIsin, execIsin, stream) == ACL_SUCCESS, return -1);
186		141
187	- // 8. 同步并获取结果	142	+ // 5. 同步并获取结果（期望 [1, 1, 0, 1, 1]，即 1 在测试集中、4 不在）
188	CHECK_RET(aclrtSynchronizeStream(stream) == ACL_SUCCESS, return -1);	143	CHECK_RET(aclrtSynchronizeStream(stream) == ACL_SUCCESS, return -1);
189	- std::vector<uint8_t> h_out(elementsNum, 0); // Bool 在 Host 用 uint8_t 接收
190	- aclrtMemcpy(h_out.data(), elementsNum, outAddr, elementsNum, ACL_MEMCPY_DEVICE_TO_HOST);
191	-
192	LOG_PRINT("--- Final Result ---\n");	144	LOG_PRINT("--- Final Result ---\n");
193	PrintOutResult(outShape, elementsNum, &outAddr);	145	PrintOutResult(outShape, elementsNum, &outAddr);
194		146
195	- // 9. 资源清理	147	+ // 6. 资源清理
196	- aclDestroyTensorList(catList);	148	+ aclDestroyTensor(valueTensor);
197	- aclDestroyTensor(elTensor);	149	+ aclDestroyTensor(indexTensor);
198	- aclDestroyTensor(teTensor);	150	+ aclDestroyTensor(elementsNumTensor);
199	- aclDestroyTensor(catTensor);
200	- aclDestroyTensor(sortVTensor);
201	- aclDestroyTensor(sortI64Tensor);
202	- aclDestroyTensor(sortI32Tensor);
203	aclDestroyTensor(outTensor);	151	aclDestroyTensor(outTensor);
204		152
205	- aclrtFree(elAddr);	153	+ aclrtFree(valueAddr);
206	- aclrtFree(teAddr);	154	+ aclrtFree(indexAddr);
207	- aclrtFree(catAddr);	155	+ aclrtFree(elementsNumAddr);
208	- aclrtFree(sortVAddr);
209	- aclrtFree(sortI64Addr);
210	- aclrtFree(sortI32Addr);
211	aclrtFree(outAddr);	156	aclrtFree(outAddr);
212	-
213	- if (wsCatAddr)
214	- aclrtFree(wsCatAddr);
215	- if (wsSortAddr)
216	- aclrtFree(wsSortAddr);
217	- if (wsCastAddr)
218	- aclrtFree(wsCastAddr);
219	if (wsIsinAddr)	157	if (wsIsinAddr)
220	aclrtFree(wsIsinAddr);	158	aclrtFree(wsIsinAddr);
221		159
	@@ -224,4 +162,4 @@ int main()
224	aclFinalize();	162	aclFinalize();
225		163
226	return 0;	164	return 0;
227	-}	165	+}

Mscripts/util/gen_compile_option.sh+16-8

	@@ -185,18 +185,26 @@ main() {
185	fi	185	fi
186		186
187	compute_units=$(echo "$json_line" \| awk '	187	compute_units=$(echo "$json_line" \| awk '
188	- match($0, /"compute_units"[[:space:]]:[[:space:]]\[([^]]*)\]/, arr) {	188	+ {
189	- str = arr[1]	189	+ if (match($0, /"compute_units"[[:space:]]:[[:space:]]\[[^]]*\]/)) {
190	- gsub(/"/, "", str)	190	+ str = substr($0, RSTART, RLENGTH)
191	- gsub(/^[[:space:]]+\|[[:space:]]+$/, "", str)	191	+ sub(/^"compute_units"[[:space:]]:[[:space:]]\[/, "", str)
192	- gsub(/,[[:space:]]*/, " ", str)	192	+ sub(/\][[:space:]]*$/, "", str)
193	- print str	193	+ gsub(/"/, "", str)
		194	+ gsub(/^[[:space:]]+\|[[:space:]]+$/, "", str)
		195	+ gsub(/,[[:space:]]*/, " ", str)
		196	+ print str
		197	+ }
194	}')	198	}')
195		199
196	if [ -z "$compile_options" ]; then	200	if [ -z "$compile_options" ]; then
197	compile_options=$(echo "$json_line" \| awk '	201	compile_options=$(echo "$json_line" \| awk '
198	- match($0, /"compile_options"[[:space:]]:[[:space:]](\{[^}]*\})/, arr) {	202	+ {
199	- print arr[1]	203	+ if (match($0, /"compile_options"[[:space:]]:[[:space:]]\{[^}]*\}/)) {
		204	+ s = substr($0, RSTART, RLENGTH)
		205	+ sub(/^"compile_options"[[:space:]]:[[:space:]]/, "", s)
		206	+ print s
		207	+ }
200	}')	208	}')
201	fi	209	fi
202		210