Merged
fix(foreach): 8个issue批量修复——错误码表对齐161002、out shape约束、列表长度错误码、资料修正、UT补齐 #13610
Tian_1122Add on 18 days ago
fix(foreach): 8个issue批量修复——错误码表对齐161002、out shape约束、列表长度错误码、资料修正、UT补齐 #13610
Merged
Total: 25 files changed+321-102
| @@ -118,7 +118,7 @@ aclnnStatus aclnnForeachAddList( | |||
| 118 | <td>out(aclTensorList*)</td> | 118 | <td>out(aclTensorList*)</td> |
| 119 | <td>输出</td> | 119 | <td>输出</td> |
| 120 | <td>表示加法运算的输出张量列表。对应公式中的`y`。</td> | 120 | <td>表示加法运算的输出张量列表。对应公式中的`y`。</td> |
| 121 | - <td><ul><li>支持空Tensor。</li><li>该参数中所有Tensor的数据类型保持一致。</li><li>数据类型和数据格式与入参`x1`的数据类型和数据格式一致,shape size大于等于入参`x1`的shape size。</li></ul></td> | 121 | + <td><ul><li>支持空Tensor。</li><li>该参数中所有Tensor的数据类型保持一致。</li><li>数据类型和数据格式与入参`x1`的数据类型和数据格式一致,shape与入参`x1`一致。</li></ul></td> |
| 122 | <td>FLOAT32、FLOAT16、BFLOAT16、INT32、INT16、INT8、UINT8</td> | 122 | <td>FLOAT32、FLOAT16、BFLOAT16、INT32、INT16、INT8、UINT8</td> |
| 123 | <td>ND</td> | 123 | <td>ND</td> |
| 124 | <td>0-8</td> | 124 | <td>0-8</td> |
| @@ -172,16 +172,14 @@ aclnnStatus aclnnForeachAddList( | |||
| 172 | <td>传入的x1、x2、alpha、out是空指针。</td> | 172 | <td>传入的x1、x2、alpha、out是空指针。</td> |
| 173 | </tr> | 173 | </tr> |
| 174 | <tr> | 174 | <tr> |
| 175 | - <td rowspan="2">ACLNN_ERR_PARAM_INVALID</td> | 175 | + <td rowspan="5">ACLNN_ERR_PARAM_INVALID</td> |
| 176 | - <td rowspan="2">161002</td> | 176 | + <td rowspan="5">161002</td> |
| 177 | <td>x1、x2、alpha、out的数据类型不在支持的范围之内。</td> | 177 | <td>x1、x2、alpha、out的数据类型不在支持的范围之内。</td> |
| 178 | </tr> | 178 | </tr> |
| 179 | <tr> | 179 | <tr> |
| 180 | <td>x1、x2、out的数据类型不一致。</td> | 180 | <td>x1、x2、out的数据类型不一致。</td> |
| 181 | </tr> | 181 | </tr> |
| 182 | <tr> | 182 | <tr> |
| 183 | - <td rowspan="4">ACLNN_ERR_INNER_TILING_ERROR</td> | ||
| 184 | - <td rowspan="4">561002</td> | ||
| 185 | <td>x1、x2、out的shape不满足约束。</td> | 183 | <td>x1、x2、out的shape不满足约束。</td> |
| 186 | </tr> | 184 | </tr> |
| 187 | <tr> | 185 | <tr> |
| @@ -191,6 +189,8 @@ aclnnStatus aclnnForeachAddList( | |||
| 191 | <td>x1、x2或out中的Tensor维度超过8维。</td> | 189 | <td>x1、x2或out中的Tensor维度超过8维。</td> |
| 192 | </tr> | 190 | </tr> |
| 193 | <tr> | 191 | <tr> |
| 192 | + <td>ACLNN_ERR_INNER_TILING_ERROR</td> | ||
| 193 | + <td>561002</td> | ||
| 194 | <td>alpha元素个数不为1。</td> | 194 | <td>alpha元素个数不为1。</td> |
| 195 | </tr> | 195 | </tr> |
| 196 | </tbody></table> | 196 | </tbody></table> |
| @@ -87,7 +87,7 @@ aclnnStatus aclnnForeachAddListV2( | |||
| 87 | <td>x1(aclTensorList*)</td> | 87 | <td>x1(aclTensorList*)</td> |
| 88 | <td>输入</td> | 88 | <td>输入</td> |
| 89 | <td>表示加法运算的第一个输入张量列表。公式中的`x1`。</td> | 89 | <td>表示加法运算的第一个输入张量列表。公式中的`x1`。</td> |
| 90 | - <td><ul><li>不支持空Tensor。</li><li>该参数中所有Tensor的数据类型保持一致。</li></ul></td> | 90 | + <td><ul><li>支持空Tensor。</li><li>该参数中所有Tensor的数据类型保持一致。</li></ul></td> |
| 91 | <td>FLOAT32、FLOAT16、BFLOAT16、INT32、INT16、INT8、UINT8</td> | 91 | <td>FLOAT32、FLOAT16、BFLOAT16、INT32、INT16、INT8、UINT8</td> |
| 92 | <td>ND</td> | 92 | <td>ND</td> |
| 93 | <td>0-8</td> | 93 | <td>0-8</td> |
| @@ -97,7 +97,7 @@ aclnnStatus aclnnForeachAddListV2( | |||
| 97 | <td>x2(aclTensorList*)</td> | 97 | <td>x2(aclTensorList*)</td> |
| 98 | <td>输入</td> | 98 | <td>输入</td> |
| 99 | <td>表示加法运算的第二个输入张量列表。对应公式中的`x2`。</td> | 99 | <td>表示加法运算的第二个输入张量列表。对应公式中的`x2`。</td> |
| 100 | - <td><ul><li>不支持空Tensor。</li><li>该参数中所有Tensor的数据类型保持一致。</li><li>数据类型、数据格式和shape与入参`x1`一致。</li></ul></td> | 100 | + <td><ul><li>支持空Tensor。</li><li>该参数中所有Tensor的数据类型保持一致。</li><li>数据类型、数据格式和shape与入参`x1`一致。</li></ul></td> |
| 101 | <td>FLOAT32、FLOAT16、BFLOAT16、INT32、INT16、INT8、UINT8</td> | 101 | <td>FLOAT32、FLOAT16、BFLOAT16、INT32、INT16、INT8、UINT8</td> |
| 102 | <td>ND</td> | 102 | <td>ND</td> |
| 103 | <td>0-8</td> | 103 | <td>0-8</td> |
| @@ -117,7 +117,7 @@ aclnnStatus aclnnForeachAddListV2( | |||
| 117 | <td>out(aclTensorList*)</td> | 117 | <td>out(aclTensorList*)</td> |
| 118 | <td>输出</td> | 118 | <td>输出</td> |
| 119 | <td>表示加法运算的输出张量列表。对应公式中的`y`。</td> | 119 | <td>表示加法运算的输出张量列表。对应公式中的`y`。</td> |
| 120 | - <td><ul><li>不支持空Tensor。</li><li>该参数中所有Tensor的数据类型保持一致。</li><li>数据类型和数据格式与入参`x1`的数据类型和数据格式一致,shape size大于等于入参`x1`的shape size。</li></ul></td> | 120 | + <td><ul><li>支持空Tensor。</li><li>该参数中所有Tensor的数据类型保持一致。</li><li>数据类型和数据格式与入参`x1`的数据类型和数据格式一致,shape与入参`x1`一致。</li></ul></td> |
| 121 | <td>FLOAT32、FLOAT16、BFLOAT16、INT32、INT16、INT8、UINT8</td> | 121 | <td>FLOAT32、FLOAT16、BFLOAT16、INT32、INT16、INT8、UINT8</td> |
| 122 | <td>ND</td> | 122 | <td>ND</td> |
| 123 | <td>0-8</td> | 123 | <td>0-8</td> |
| @@ -89,7 +89,7 @@ aclnnStatus aclnnForeachAddListV2( | |||
| 89 | <td>x1 (aclTensorList*)</td> | 89 | <td>x1 (aclTensorList*)</td> |
| 90 | <td>Input</td> | 90 | <td>Input</td> |
| 91 | <td>Indicates the first input tensor list of the addition operation. `x1` in the formula.</td> | 91 | <td>Indicates the first input tensor list of the addition operation. `x1` in the formula.</td> |
| 92 | - <td><ul><li>Does not support empty tensors.</li><li>All tensors in this parameter must have the same data type.</li></ul></td> | 92 | + <td><ul><li>Supports empty tensors.</li><li>All tensors in this parameter must have the same data type.</li></ul></td> |
| 93 | <td>FLOAT32, FLOAT16, BFLOAT16, INT32, INT16, INT8, UINT8</td> | 93 | <td>FLOAT32, FLOAT16, BFLOAT16, INT32, INT16, INT8, UINT8</td> |
| 94 | <td>ND</td> | 94 | <td>ND</td> |
| 95 | <td>0-8</td> | 95 | <td>0-8</td> |
| @@ -99,7 +99,7 @@ aclnnStatus aclnnForeachAddListV2( | |||
| 99 | <td>x2 (aclTensorList*)</td> | 99 | <td>x2 (aclTensorList*)</td> |
| 100 | <td>Input</td> | 100 | <td>Input</td> |
| 101 | <td>Indicates the second input tensor list of the addition operation. `x2` in the corresponding formula.</td> | 101 | <td>Indicates the second input tensor list of the addition operation. `x2` in the corresponding formula.</td> |
| 102 | - <td><ul><li>Does not support empty tensors.</li><li>All tensors in this parameter must have the same data type.</li><li>The data type, data format, and shape are the same as those of input `x1`.</li></ul></td> | 102 | + <td><ul><li>Supports empty tensors.</li><li>All tensors in this parameter must have the same data type.</li><li>The data type, data format, and shape are the same as those of input `x1`.</li></ul></td> |
| 103 | <td>FLOAT32, FLOAT16, BFLOAT16, INT32, INT16, INT8, UINT8</td> | 103 | <td>FLOAT32, FLOAT16, BFLOAT16, INT32, INT16, INT8, UINT8</td> |
| 104 | <td>ND</td> | 104 | <td>ND</td> |
| 105 | <td>0-8</td> | 105 | <td>0-8</td> |
| @@ -119,7 +119,7 @@ aclnnStatus aclnnForeachAddListV2( | |||
| 119 | <td>out (aclTensorList*)</td> | 119 | <td>out (aclTensorList*)</td> |
| 120 | <td>Output</td> | 120 | <td>Output</td> |
| 121 | <td>Output tensor list that represents addition. Corresponds to `y` in the corresponding formula.</td> | 121 | <td>Output tensor list that represents addition. Corresponds to `y` in the corresponding formula.</td> |
| 122 | - <td><ul><li>Does not support empty tensors.</li><li>All tensors in this parameter have the same data type.</li><li>The data type and data format must be the same as those of input `x1`, and the shape size must be greater than or equal to that of input `x1`.</li></ul></td> | 122 | + <td><ul><li>Supports empty tensors.</li><li>All tensors in this parameter have the same data type.</li><li>The data type and data format must be the same as those of input `x1`, and the shape is the same as that of input `x1`.</li></ul></td> |
| 123 | <td>FLOAT32, FLOAT16, BFLOAT16, INT32, INT16, INT8, UINT8</td> | 123 | <td>FLOAT32, FLOAT16, BFLOAT16, INT32, INT16, INT8, UINT8</td> |
| 124 | <td>ND</td> | 124 | <td>ND</td> |
| 125 | <td>0-8</td> | 125 | <td>0-8</td> |
| @@ -120,7 +120,7 @@ aclnnStatus aclnnForeachAddList( | |||
| 120 | <td>out (aclTensorList*)</td> | 120 | <td>out (aclTensorList*)</td> |
| 121 | <td>Output</td> | 121 | <td>Output</td> |
| 122 | <td>Output tensor list that represents the addition operation. It corresponds to `y` in the formula.</td> | 122 | <td>Output tensor list that represents the addition operation. It corresponds to `y` in the formula.</td> |
| 123 | - <td><ul><li>Supports empty tensors.</li><li>All tensors in this parameter have the same data type.</li><li>The data type and data format are the same as those of input parameter `x1`, and the shape size is greater than or equal to that of input parameter `x1`.</li></ul></td> | 123 | + <td><ul><li>Supports empty tensors.</li><li>All tensors in this parameter have the same data type.</li><li>The data type and data format are the same as those of input parameter `x1`, and the shape is the same as that of input parameter `x1`.</li></ul></td> |
| 124 | <td>FLOAT32, FLOAT16, BFLOAT16, INT32, INT16, INT8, UINT8</td> | 124 | <td>FLOAT32, FLOAT16, BFLOAT16, INT32, INT16, INT8, UINT8</td> |
| 125 | <td>ND</td> | 125 | <td>ND</td> |
| 126 | <td>0-8</td> | 126 | <td>0-8</td> |
| @@ -174,16 +174,14 @@ aclnnStatus aclnnForeachAddList( | |||
| 174 | <td>x1, x2, alpha, or out is a null pointer.</td> | 174 | <td>x1, x2, alpha, or out is a null pointer.</td> |
| 175 | </tr> | 175 | </tr> |
| 176 | <tr> | 176 | <tr> |
| 177 | - <td rowspan="2">ACLNN_ERR_PARAM_INVALID</td> | 177 | + <td rowspan="5">ACLNN_ERR_PARAM_INVALID</td> |
| 178 | - <td rowspan="2">161002</td> | 178 | + <td rowspan="5">161002</td> |
| 179 | <td>The data type of x1, x2, alpha, or out is not supported.</td> | 179 | <td>The data type of x1, x2, alpha, or out is not supported.</td> |
| 180 | </tr> | 180 | </tr> |
| 181 | <tr> | 181 | <tr> |
| 182 | <td>The data types of x1, x2, and out are inconsistent.</td> | 182 | <td>The data types of x1, x2, and out are inconsistent.</td> |
| 183 | </tr> | 183 | </tr> |
| 184 | <tr> | 184 | <tr> |
| 185 | - <td rowspan="4">ACLNN_ERR_INNER_TILING_ERROR</td> | ||
| 186 | - <td rowspan="4">561002</td> | ||
| 187 | <td>The shapes of x1, x2, and out do not meet the constraints.</td> | 185 | <td>The shapes of x1, x2, and out do not meet the constraints.</td> |
| 188 | </tr> | 186 | </tr> |
| 189 | <tr> | 187 | <tr> |
| @@ -193,6 +191,8 @@ aclnnStatus aclnnForeachAddList( | |||
| 193 | <td>The dimensions of Tensors in x1, x2, or out exceed 8.</td> | 191 | <td>The dimensions of Tensors in x1, x2, or out exceed 8.</td> |
| 194 | </tr> | 192 | </tr> |
| 195 | <tr> | 193 | <tr> |
| 194 | + <td>ACLNN_ERR_INNER_TILING_ERROR</td> | ||
| 195 | + <td>561002</td> | ||
| 196 | <td>The number of alpha elements is not 1.</td> | 196 | <td>The number of alpha elements is not 1.</td> |
| 197 | </tr> | 197 | </tr> |
| 198 | </tbody></table> | 198 | </tbody></table> |
| @@ -31,13 +31,13 @@ extern "C" { | |||
| 31 | * @domain aclnnop_math | 31 | * @domain aclnnop_math |
| 32 | * 参数描述: | 32 | * 参数描述: |
| 33 | * @param [in] x1 | 33 | * @param [in] x1 |
| 34 | - * 输入Tensor,数据类型支持FLOAT,FLOAT16,BFLOAT16,INT32。数据格式支持ND。 | 34 | + * 输入Tensor,数据类型支持FLOAT,FLOAT16,BFLOAT16,INT32,INT16,INT8,UINT8。数据格式支持ND。 |
| 35 | * @param [in] x2 | 35 | * @param [in] x2 |
| 36 | - * 输入Tensor,数据类型支持FLOAT,FLOAT16,BFLOAT16,INT32。数据格式支持ND。 | 36 | + * 输入Tensor,数据类型支持FLOAT,FLOAT16,BFLOAT16,INT32,INT16,INT8,UINT8。数据格式支持ND。 |
| 37 | * @param [in] alpha | 37 | * @param [in] alpha |
| 38 | - * 输入Scalar,数据类型支持FLOAT、FLOAT16和INT32。数据格式支持ND。 | 38 | + * 输入Scalar,数据类型支持FLOAT、FLOAT16、INT32、DOUBLE、INT64。数据格式支持ND。 |
| 39 | * @param [in] out | 39 | * @param [in] out |
| 40 | - * 输出Tensor,数据类型支持FLOAT,FLOAT16,BFLOAT16,INT32。数据格式支持ND。 | 40 | + * 输出Tensor,数据类型支持FLOAT,FLOAT16,BFLOAT16,INT32,INT16,INT8,UINT8。数据格式支持ND。 |
| 41 | * @param [out] workspaceSize 返回用户需要在npu device侧申请的workspace大小。 | 41 | * @param [out] workspaceSize 返回用户需要在npu device侧申请的workspace大小。 |
| 42 | * @param [out] executor 返回op执行器,包含了算子计算流程。 | 42 | * @param [out] executor 返回op执行器,包含了算子计算流程。 |
| 43 | * @return aclnnStatus 返回状态码 | 43 | * @return aclnnStatus 返回状态码 |
| @@ -213,3 +213,37 @@ TEST_F(l2_foreach_add_list_v2_test, ascend910B2_foreach_add_list_v2_test_non_con | |||
| 213 | aclnnStatus getWorkspaceResult = ut.TestGetWorkspaceSize(&workspaceSize); | 213 | aclnnStatus getWorkspaceResult = ut.TestGetWorkspaceSize(&workspaceSize); |
| 214 | EXPECT_EQ(getWorkspaceResult, ACL_SUCCESS); | 214 | EXPECT_EQ(getWorkspaceResult, ACL_SUCCESS); |
| 215 | } | 215 | } |
| 216 | + | ||
| 217 | +// out 元素数大于 x1:实现要求严格同 shape,应拒绝 | ||
| 218 | +TEST_F(l2_foreach_add_list_v2_test, ascend910B2_foreach_add_list_v2_test_out_shape_larger_numel) | ||
| 219 | +{ | ||
| 220 | + auto alpha = ScalarDesc(1.0); | ||
| 221 | + auto x1 = TensorDesc({2, 3}, ACL_FLOAT, ACL_FORMAT_ND).ValueRange(-1, 1); | ||
| 222 | + auto x2 = TensorDesc({2, 3}, ACL_FLOAT, ACL_FORMAT_ND).ValueRange(-1, 1); | ||
| 223 | + auto out = TensorDesc({2, 4}, ACL_FLOAT, ACL_FORMAT_ND).Precision(0.001, 0.001); | ||
| 224 | + auto x1List = TensorListDesc({x1}); | ||
| 225 | + auto x2List = TensorListDesc({x2}); | ||
| 226 | + auto outList = TensorListDesc({out}); | ||
| 227 | + | ||
| 228 | + auto ut = OP_API_UT(aclnnForeachAddListV2, INPUT(x1List, x2List, alpha), OUTPUT(outList)); | ||
| 229 | + uint64_t workspaceSize = 0; | ||
| 230 | + aclnnStatus getWorkspaceResult = ut.TestGetWorkspaceSize(&workspaceSize); | ||
| 231 | + EXPECT_EQ(getWorkspaceResult, ACLNN_ERR_PARAM_INVALID); | ||
| 232 | +} | ||
| 233 | + | ||
| 234 | +// 空 Tensor:实现不拦截空 Tensor,应放行 | ||
| 235 | +TEST_F(l2_foreach_add_list_v2_test, ascend910B2_foreach_add_list_v2_test_empty_tensor) | ||
| 236 | +{ | ||
| 237 | + auto alpha = ScalarDesc(1.0); | ||
| 238 | + auto x1 = TensorDesc({0}, ACL_FLOAT, ACL_FORMAT_ND); | ||
| 239 | + auto x2 = TensorDesc({0}, ACL_FLOAT, ACL_FORMAT_ND); | ||
| 240 | + auto out = TensorDesc({0}, ACL_FLOAT, ACL_FORMAT_ND); | ||
| 241 | + auto x1List = TensorListDesc({x1}); | ||
| 242 | + auto x2List = TensorListDesc({x2}); | ||
| 243 | + auto outList = TensorListDesc({out}); | ||
| 244 | + | ||
| 245 | + auto ut = OP_API_UT(aclnnForeachAddListV2, INPUT(x1List, x2List, alpha), OUTPUT(outList)); | ||
| 246 | + uint64_t workspaceSize = 0; | ||
| 247 | + aclnnStatus getWorkspaceResult = ut.TestGetWorkspaceSize(&workspaceSize); | ||
| 248 | + EXPECT_EQ(getWorkspaceResult, ACL_SUCCESS); | ||
| 249 | +} | ||
| @@ -179,13 +179,16 @@ aclnnStatus aclnnForeachAddScalar( | |||
| 179 | <td>传入的x、scalar、out是空指针。</td> | 179 | <td>传入的x、scalar、out是空指针。</td> |
| 180 | </tr> | 180 | </tr> |
| 181 | <tr> | 181 | <tr> |
| 182 | - <td rowspan="2">ACLNN_ERR_PARAM_INVALID</td> | 182 | + <td rowspan="3">ACLNN_ERR_PARAM_INVALID</td> |
| 183 | - <td rowspan="2">161002</td> | 183 | + <td rowspan="3">161002</td> |
| 184 | <td>x、scalar、out的数据类型不在支持的范围之内。</td> | 184 | <td>x、scalar、out的数据类型不在支持的范围之内。</td> |
| 185 | </tr> | 185 | </tr> |
| 186 | <tr> | 186 | <tr> |
| 187 | <td>x和out的数据类型不一致。</td> | 187 | <td>x和out的数据类型不一致。</td> |
| 188 | </tr> | 188 | </tr> |
| 189 | + <tr> | ||
| 190 | + <td>x与out的tensorList长度(包含的Tensor个数)不一致。</td> | ||
| 191 | + </tr> | ||
| 189 | <tr> | 192 | <tr> |
| 190 | <td rowspan="4">ACLNN_ERR_INNER_TILING_ERROR</td> | 193 | <td rowspan="4">ACLNN_ERR_INNER_TILING_ERROR</td> |
| 191 | <td rowspan="4">561002</td> | 194 | <td rowspan="4">561002</td> |
| @@ -181,8 +181,6 @@ aclnnStatus aclnnForeachAddScalarV2( | |||
| 181 | <tr> | 181 | <tr> |
| 182 | <td rowspan="6">ACLNN_ERR_PARAM_INVALID</td> | 182 | <td rowspan="6">ACLNN_ERR_PARAM_INVALID</td> |
| 183 | <td rowspan="6">161002</td> | 183 | <td rowspan="6">161002</td> |
| 184 | - </tr> | ||
| 185 | - <tr> | ||
| 186 | <td>x、scalar和out的数据类型不在支持的范围之内。</td> | 184 | <td>x、scalar和out的数据类型不在支持的范围之内。</td> |
| 187 | </tr> | 185 | </tr> |
| 188 | <tr> | 186 | <tr> |
| @@ -190,12 +188,15 @@ aclnnStatus aclnnForeachAddScalarV2( | |||
| 190 | </tr> | 188 | </tr> |
| 191 | <tr> | 189 | <tr> |
| 192 | <td>x与out的shape不满足约束。</td> | 190 | <td>x与out的shape不满足约束。</td> |
| 191 | + </tr> | ||
| 193 | <tr> | 192 | <tr> |
| 194 | <td>x或out中的Tensor的数据类型不一致。</td> | 193 | <td>x或out中的Tensor的数据类型不一致。</td> |
| 195 | </tr> | 194 | </tr> |
| 196 | <tr> | 195 | <tr> |
| 197 | <td>x或out中的Tensor维度超过8维。</td> | 196 | <td>x或out中的Tensor维度超过8维。</td> |
| 198 | </tr> | 197 | </tr> |
| 198 | + <tr> | ||
| 199 | + <td>x与out的tensorList长度(包含的Tensor个数)不一致。</td> | ||
| 199 | </tr> | 200 | </tr> |
| 200 | </tbody></table> | 201 | </tbody></table> |
| 201 | 202 | ||
| @@ -52,7 +52,6 @@ static inline bool ForeachAddScalarCheckNotNull(const aclTensorList* self, const | |||
| 52 | OP_CHECK_NULL(self, return false); | 52 | OP_CHECK_NULL(self, return false); |
| 53 | OP_CHECK_NULL(scalar, return false); | 53 | OP_CHECK_NULL(scalar, return false); |
| 54 | OP_CHECK_NULL(out, return false); | 54 | OP_CHECK_NULL(out, return false); |
| 55 | - OP_CHECK_TENSORLIST_SIZE_EQUAL(self, out, return false); | ||
| 56 | return true; | 55 | return true; |
| 57 | } | 56 | } |
| 58 | 57 | ||
| @@ -62,7 +62,6 @@ static inline bool ForeachAddScalarV2CheckNotNull(const aclTensorList* self, con | |||
| 62 | OP_CHECK_NULL(self, return false); | 62 | OP_CHECK_NULL(self, return false); |
| 63 | OP_CHECK_NULL(scalar, return false); | 63 | OP_CHECK_NULL(scalar, return false); |
| 64 | OP_CHECK_NULL(out, return false); | 64 | OP_CHECK_NULL(out, return false); |
| 65 | - OP_CHECK_TENSORLIST_SIZE_EQUAL(self, out, return false); | ||
| 66 | return true; | 65 | return true; |
| 67 | } | 66 | } |
| 68 | 67 | ||
| @@ -43,3 +43,18 @@ TEST_F(l2_foreach_add_scalar_test, ascend910B2_foreach_add_scalar_test_non_conti | |||
| 43 | aclnnStatus getWorkspaceResult = ut.TestGetWorkspaceSize(&workspaceSize); | 43 | aclnnStatus getWorkspaceResult = ut.TestGetWorkspaceSize(&workspaceSize); |
| 44 | EXPECT_EQ(getWorkspaceResult, ACL_SUCCESS); | 44 | EXPECT_EQ(getWorkspaceResult, ACL_SUCCESS); |
| 45 | } | 45 | } |
| 46 | + | ||
| 47 | +// x 与 out 的 tensorList 长度不一致:应返回 161002 | ||
| 48 | +TEST_F(l2_foreach_add_scalar_test, ascend910B2_foreach_add_scalar_test_list_length_mismatch) | ||
| 49 | +{ | ||
| 50 | + auto scalar = TensorDesc({1}, ACL_FLOAT, ACL_FORMAT_ND).ValueRange(0, 2); | ||
| 51 | + auto x = TensorDesc({2, 3}, ACL_FLOAT, ACL_FORMAT_ND).ValueRange(-1, 1); | ||
| 52 | + auto out = TensorDesc({2, 3}, ACL_FLOAT, ACL_FORMAT_ND).Precision(0.001, 0.001); | ||
| 53 | + auto xList = TensorListDesc({x, x}); | ||
| 54 | + auto outList = TensorListDesc({out}); | ||
| 55 | + | ||
| 56 | + auto ut = OP_API_UT(aclnnForeachAddScalar, INPUT(xList, scalar), OUTPUT(outList)); | ||
| 57 | + uint64_t workspaceSize = 0; | ||
| 58 | + aclnnStatus getWorkspaceResult = ut.TestGetWorkspaceSize(&workspaceSize); | ||
| 59 | + EXPECT_EQ(getWorkspaceResult, ACLNN_ERR_PARAM_INVALID); | ||
| 60 | +} | ||
| @@ -208,3 +208,18 @@ TEST_F(l2_foreach_add_scalar_v2_test, ascend910B2_foreach_add_scalar_v2_test_non | |||
| 208 | aclnnStatus getWorkspaceResult = ut.TestGetWorkspaceSize(&workspaceSize); | 208 | aclnnStatus getWorkspaceResult = ut.TestGetWorkspaceSize(&workspaceSize); |
| 209 | EXPECT_EQ(getWorkspaceResult, ACL_SUCCESS); | 209 | EXPECT_EQ(getWorkspaceResult, ACL_SUCCESS); |
| 210 | } | 210 | } |
| 211 | + | ||
| 212 | +// x 与 out 的 tensorList 长度不一致:应返回 161002 | ||
| 213 | +TEST_F(l2_foreach_add_scalar_v2_test, ascend910B2_foreach_add_scalar_v2_test_list_length_mismatch) | ||
| 214 | +{ | ||
| 215 | + auto scalar_desc = ScalarDesc(1.0); | ||
| 216 | + auto x = TensorDesc({2, 3}, ACL_FLOAT, ACL_FORMAT_ND).ValueRange(-1, 1); | ||
| 217 | + auto out = TensorDesc({2, 3}, ACL_FLOAT, ACL_FORMAT_ND).Precision(0.001, 0.001); | ||
| 218 | + auto xList = TensorListDesc({x, x}); | ||
| 219 | + auto outList = TensorListDesc({out}); | ||
| 220 | + | ||
| 221 | + auto ut = OP_API_UT(aclnnForeachAddScalarV2, INPUT(xList, scalar_desc), OUTPUT(outList)); | ||
| 222 | + uint64_t workspaceSize = 0; | ||
| 223 | + aclnnStatus getWorkspaceResult = ut.TestGetWorkspaceSize(&workspaceSize); | ||
| 224 | + EXPECT_EQ(getWorkspaceResult, ACLNN_ERR_PARAM_INVALID); | ||
| 225 | +} | ||
| @@ -208,7 +208,7 @@ aclnnStatus aclnnForeachAddcmulScalarV2( | |||
| 208 | <td>x1、x2、x3、out的数据类型不一致。</td> | 208 | <td>x1、x2、x3、out的数据类型不一致。</td> |
| 209 | </tr> | 209 | </tr> |
| 210 | <tr> | 210 | <tr> |
| 211 | - <td>x1、x2、x3、out的shape不满足约束。</td> | 211 | + <td>x1、x2、x3、out的shape不满足约束(含out的shape size小于x1的shape size)。</td> |
| 212 | </tr> | 212 | </tr> |
| 213 | <tr> | 213 | <tr> |
| 214 | <td>x1、x2、x3或out中的Tensor的元素数据类型不一致。</td> | 214 | <td>x1、x2、x3或out中的Tensor的元素数据类型不一致。</td> |
| @@ -182,11 +182,24 @@ static inline bool ForeachAddcMulScalarV2CheckShape(const aclTensorList* self, c | |||
| 182 | // tensor 维度检查 | 182 | // tensor 维度检查 |
| 183 | for (uint64_t i = 0; i < self->Size(); i++) { | 183 | for (uint64_t i = 0; i < self->Size(); i++) { |
| 184 | OP_CHECK_MAX_DIM((*self)[i], MAX_SUPPORT_DIMS_NUMS, return false); | 184 | OP_CHECK_MAX_DIM((*self)[i], MAX_SUPPORT_DIMS_NUMS, return false); |
| 185 | + OP_CHECK_MAX_DIM((*out)[i], MAX_SUPPORT_DIMS_NUMS, return false); | ||
| 185 | } | 186 | } |
| 186 | 187 | ||
| 187 | - // self和out的shape必须一致 | 188 | + // out的元素数必须大于等于x1(文档:out的shape size大于等于入参x1的shape size) |
| 188 | for (uint64_t i = 0; i < self->Size(); i++) { | 189 | for (uint64_t i = 0; i < self->Size(); i++) { |
| 189 | - OP_CHECK_SHAPE_NOT_EQUAL((*self)[i], (*out)[i], return false); | 190 | + const int64_t x1Size = (*self)[i]->GetViewShape().GetShapeSize(); |
| 191 | + const int64_t outSize = (*out)[i]->GetViewShape().GetShapeSize(); | ||
| 192 | + if (outSize < x1Size) { | ||
| 193 | + const std::string reason = ("The shape size of the " + std::to_string(i) + | ||
| 194 | + "th tensor in out must be greater than or equal to that of x1, but got " + | ||
| 195 | + std::to_string(outSize) + " and " + std::to_string(x1Size) + "."); | ||
| 196 | + OP_LOGE_WITHOUT_REPORT(ACLNN_ERR_PARAM_INVALID, "%s", reason.c_str()); | ||
| 197 | + REPORT_PREDEFINED_ERR_MSG("EZ0015", | ||
| 198 | + (std::vector<const char*>{"param_name", "op_name", "incorrect_size", "reason"}), | ||
| 199 | + (std::vector<const char*>{"out", "aclnnForeachAddcmulScalarV2", | ||
| 200 | + (std::to_string(outSize)).c_str(), reason.c_str()})); | ||
| 201 | + return false; | ||
| 202 | + } | ||
| 190 | } | 203 | } |
| 191 | return true; | 204 | return true; |
| 192 | } | 205 | } |
Mforeach/foreach_addcmul_scalar/tests/ut/op_host/op_api/test_aclnn_foreach_addcmul_scalar_v2.cpp+65-2
| @@ -51,7 +51,7 @@ TEST_F(l2_foreach_addcmul_scalar_v2_test, ascend910B2_foreach_addcmul_scalar_v2_ | |||
| 51 | EXPECT_EQ(getWorkspaceResult, ACLNN_ERR_PARAM_INVALID); | 51 | EXPECT_EQ(getWorkspaceResult, ACLNN_ERR_PARAM_INVALID); |
| 52 | } | 52 | } |
| 53 | 53 | ||
| 54 | -// self and out diffrent shape | 54 | +// self and out diffrent shape(out 元素数 >= x1:文档契约,应放行) |
| 55 | TEST_F(l2_foreach_addcmul_scalar_v2_test, ascend910B2_foreach_addcmul_scalar_v2_test_shape_different) | 55 | TEST_F(l2_foreach_addcmul_scalar_v2_test, ascend910B2_foreach_addcmul_scalar_v2_test_shape_different) |
| 56 | { | 56 | { |
| 57 | vector<vector<int64_t>> selfDims = {{2, 2}}; | 57 | vector<vector<int64_t>> selfDims = {{2, 2}}; |
| @@ -69,7 +69,7 @@ TEST_F(l2_foreach_addcmul_scalar_v2_test, ascend910B2_foreach_addcmul_scalar_v2_ | |||
| 69 | auto ut = OP_API_UT(aclnnForeachAddcmulScalarV2, INPUT(xList, x2List, x3List, scalar_desc), OUTPUT(outList)); | 69 | auto ut = OP_API_UT(aclnnForeachAddcmulScalarV2, INPUT(xList, x2List, x3List, scalar_desc), OUTPUT(outList)); |
| 70 | uint64_t workspaceSize = 0; | 70 | uint64_t workspaceSize = 0; |
| 71 | aclnnStatus getWorkspaceResult = ut.TestGetWorkspaceSize(&workspaceSize); | 71 | aclnnStatus getWorkspaceResult = ut.TestGetWorkspaceSize(&workspaceSize); |
| 72 | - EXPECT_EQ(getWorkspaceResult, ACLNN_ERR_PARAM_INVALID); | 72 | + EXPECT_EQ(getWorkspaceResult, ACL_SUCCESS); |
| 73 | } | 73 | } |
| 74 | 74 | ||
| 75 | // self and x2 diffrent shape | 75 | // self and x2 diffrent shape |
| @@ -215,3 +215,66 @@ TEST_F(l2_foreach_addcmul_scalar_v2_test, ascend910B2_foreach_addcmul_scalar_v2_ | |||
| 215 | aclnnStatus getWorkspaceResult = ut.TestGetWorkspaceSize(&workspaceSize); | 215 | aclnnStatus getWorkspaceResult = ut.TestGetWorkspaceSize(&workspaceSize); |
| 216 | EXPECT_EQ(getWorkspaceResult, ACL_SUCCESS); | 216 | EXPECT_EQ(getWorkspaceResult, ACL_SUCCESS); |
| 217 | } | 217 | } |
| 218 | + | ||
| 219 | +// out 元素数大于 x1:文档要求 shape size >=,应放行 | ||
| 220 | +TEST_F(l2_foreach_addcmul_scalar_v2_test, ascend910B2_foreach_addcmul_scalar_v2_test_out_shape_larger_numel) | ||
| 221 | +{ | ||
| 222 | + vector<vector<int64_t>> selfDims = {{2, 2}}; | ||
| 223 | + auto scalar_desc = ScalarDesc(1.0); | ||
| 224 | + vector<vector<int64_t>> outDims = {{2, 4}}; | ||
| 225 | + auto x = TensorDesc(selfDims[0], ACL_FLOAT, ACL_FORMAT_ND).ValueRange(-1, 1); | ||
| 226 | + auto x2 = TensorDesc(selfDims[0], ACL_FLOAT, ACL_FORMAT_ND).ValueRange(-1, 1); | ||
| 227 | + auto x3 = TensorDesc(selfDims[0], ACL_FLOAT, ACL_FORMAT_ND).ValueRange(-1, 1); | ||
| 228 | + auto out = TensorDesc(outDims[0], ACL_FLOAT, ACL_FORMAT_ND).Precision(0.001, 0.001); | ||
| 229 | + auto xList = TensorListDesc({x}); | ||
| 230 | + auto x2List = TensorListDesc({x2}); | ||
| 231 | + auto x3List = TensorListDesc({x3}); | ||
| 232 | + auto outList = TensorListDesc({out}); | ||
| 233 | + | ||
| 234 | + auto ut = OP_API_UT(aclnnForeachAddcmulScalarV2, INPUT(xList, x2List, x3List, scalar_desc), OUTPUT(outList)); | ||
| 235 | + uint64_t workspaceSize = 0; | ||
| 236 | + aclnnStatus getWorkspaceResult = ut.TestGetWorkspaceSize(&workspaceSize); | ||
| 237 | + EXPECT_EQ(getWorkspaceResult, ACL_SUCCESS); | ||
| 238 | +} | ||
| 239 | + | ||
| 240 | +// out 与 x1 元素数相同但 shape 不同:文档要求 shape size >=,应放行 | ||
| 241 | +TEST_F(l2_foreach_addcmul_scalar_v2_test, ascend910B2_foreach_addcmul_scalar_v2_test_out_shape_same_numel) | ||
| 242 | +{ | ||
| 243 | + vector<vector<int64_t>> selfDims = {{2, 2}}; | ||
| 244 | + auto scalar_desc = ScalarDesc(1.0); | ||
| 245 | + vector<vector<int64_t>> outDims = {{4}}; | ||
| 246 | + auto x = TensorDesc(selfDims[0], ACL_FLOAT, ACL_FORMAT_ND).ValueRange(-1, 1); | ||
| 247 | + auto x2 = TensorDesc(selfDims[0], ACL_FLOAT, ACL_FORMAT_ND).ValueRange(-1, 1); | ||
| 248 | + auto x3 = TensorDesc(selfDims[0], ACL_FLOAT, ACL_FORMAT_ND).ValueRange(-1, 1); | ||
| 249 | + auto out = TensorDesc(outDims[0], ACL_FLOAT, ACL_FORMAT_ND).Precision(0.001, 0.001); | ||
| 250 | + auto xList = TensorListDesc({x}); | ||
| 251 | + auto x2List = TensorListDesc({x2}); | ||
| 252 | + auto x3List = TensorListDesc({x3}); | ||
| 253 | + auto outList = TensorListDesc({out}); | ||
| 254 | + | ||
| 255 | + auto ut = OP_API_UT(aclnnForeachAddcmulScalarV2, INPUT(xList, x2List, x3List, scalar_desc), OUTPUT(outList)); | ||
| 256 | + uint64_t workspaceSize = 0; | ||
| 257 | + aclnnStatus getWorkspaceResult = ut.TestGetWorkspaceSize(&workspaceSize); | ||
| 258 | + EXPECT_EQ(getWorkspaceResult, ACL_SUCCESS); | ||
| 259 | +} | ||
| 260 | + | ||
| 261 | +// out 元素数小于 x1:应拒绝 | ||
| 262 | +TEST_F(l2_foreach_addcmul_scalar_v2_test, ascend910B2_foreach_addcmul_scalar_v2_test_out_shape_smaller_numel) | ||
| 263 | +{ | ||
| 264 | + vector<vector<int64_t>> selfDims = {{2, 4}}; | ||
| 265 | + auto scalar_desc = ScalarDesc(1.0); | ||
| 266 | + vector<vector<int64_t>> outDims = {{2, 2}}; | ||
| 267 | + auto x = TensorDesc(selfDims[0], ACL_FLOAT, ACL_FORMAT_ND).ValueRange(-1, 1); | ||
| 268 | + auto x2 = TensorDesc(selfDims[0], ACL_FLOAT, ACL_FORMAT_ND).ValueRange(-1, 1); | ||
| 269 | + auto x3 = TensorDesc(selfDims[0], ACL_FLOAT, ACL_FORMAT_ND).ValueRange(-1, 1); | ||
| 270 | + auto out = TensorDesc(outDims[0], ACL_FLOAT, ACL_FORMAT_ND).Precision(0.001, 0.001); | ||
| 271 | + auto xList = TensorListDesc({x}); | ||
| 272 | + auto x2List = TensorListDesc({x2}); | ||
| 273 | + auto x3List = TensorListDesc({x3}); | ||
| 274 | + auto outList = TensorListDesc({out}); | ||
| 275 | + | ||
| 276 | + auto ut = OP_API_UT(aclnnForeachAddcmulScalarV2, INPUT(xList, x2List, x3List, scalar_desc), OUTPUT(outList)); | ||
| 277 | + uint64_t workspaceSize = 0; | ||
| 278 | + aclnnStatus getWorkspaceResult = ut.TestGetWorkspaceSize(&workspaceSize); | ||
| 279 | + EXPECT_EQ(getWorkspaceResult, ACLNN_ERR_PARAM_INVALID); | ||
| 280 | +} | ||
| @@ -1,12 +1,12 @@ | |||
| 1 | #!/usr/bin/env python3 | 1 | #!/usr/bin/env python3 |
| 2 | # -*- coding: utf-8 -*- | 2 | # -*- coding: utf-8 -*- |
| 3 | # ---------------------------------------------------------------------------- | 3 | # ---------------------------------------------------------------------------- |
| 4 | -# This program is free software, you can redistribute it and/or modify. | 4 | +# Copyright (c) 2025-2026 Huawei Technologies Co., Ltd. |
| 5 | -# Copyright (c) 2025 Huawei Technologies Co., Ltd. | 5 | +# This program is free software, you can redistribute it and/or modify it under the terms and conditions of |
| 6 | -# This file is a part of the CANN Open Software. | 6 | +# CANN Open Software License Agreement Version 2.0 (the "License"). |
| 7 | -# Licensed under CANN Open Software License Agreement Version 2.0 (the "License"). | ||
| 8 | # Please refer to the License for details. You may not use this file except in compliance with the License. | 7 | # Please refer to the License for details. You may not use this file except in compliance with the License. |
| 9 | -# THIS SOFTWARE IS PROVIDED ON AN "AS IS" BASIS, WITHOUT WARRANTIES OF ANY KIND, EITHER EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO NON-INFRINGEMENT, MERCHANTABILITY, OR FITNESS FOR A PARTICULAR PURPOSE. | 8 | +# THIS SOFTWARE IS PROVIDED ON AN "AS IS" BASIS, WITHOUT WARRANTIES OF ANY KIND, EITHER EXPRESS OR IMPLIED, |
| 9 | +# INCLUDING BUT NOT LIMITED TO NON-INFRINGEMENT, MERCHANTABILITY, OR FITNESS FOR A PARTICULAR PURPOSE. | ||
| 10 | # See LICENSE in the root of the software repository for the full text of the License. | 10 | # See LICENSE in the root of the software repository for the full text of the License. |
| 11 | # ---------------------------------------------------------------------------- | 11 | # ---------------------------------------------------------------------------- |
| 12 | import sys | 12 | import sys |
| @@ -15,6 +15,8 @@ import numpy as np | |||
| 15 | import re | 15 | import re |
| 16 | import tensorflow as tf | 16 | import tensorflow as tf |
| 17 | 17 | ||
| 18 | +TAIL_MARKER = -999.0 | ||
| 19 | + | ||
| 18 | 20 | ||
| 19 | def parse_str_to_shape_list(shape_str): | 21 | def parse_str_to_shape_list(shape_str): |
| 20 | shape_list = [] | 22 | shape_list = [] |
| @@ -25,44 +27,83 @@ def parse_str_to_shape_list(shape_str): | |||
| 25 | return shape_list | 27 | return shape_list |
| 26 | 28 | ||
| 27 | 29 | ||
| 28 | -def gen_data_and_golden(shape_str, scale_value=2.0, d_type="float32"): | 30 | +def gen_data_and_golden( |
| 31 | + shape_str, scale_value=2.0, d_type="float32", out_shape_str=None | ||
| 32 | +): | ||
| 29 | d_type_dict = { | 33 | d_type_dict = { |
| 30 | "float32": np.float32, | 34 | "float32": np.float32, |
| 31 | "float16": np.float16, | 35 | "float16": np.float16, |
| 32 | "int32": np.int32, | 36 | "int32": np.int32, |
| 33 | - "bfloat16_t": tf.bfloat16.as_numpy_dtype | 37 | + "bfloat16_t": tf.bfloat16.as_numpy_dtype, |
| 34 | } | 38 | } |
| 35 | np_type = d_type_dict[d_type] | 39 | np_type = d_type_dict[d_type] |
| 36 | shape_list = parse_str_to_shape_list(shape_str) | 40 | shape_list = parse_str_to_shape_list(shape_str) |
| 41 | + out_shape_list = ( | ||
| 42 | + parse_str_to_shape_list(out_shape_str) if out_shape_str is not None else None | ||
| 43 | + ) | ||
| 37 | if d_type == "bfloat16_t": | 44 | if d_type == "bfloat16_t": |
| 38 | - scalars = np.array([scale_value for i in range(len(shape_list))]).astype(np.float32) | 45 | + scalars = np.array([scale_value for i in range(len(shape_list))]).astype( |
| 46 | + np.float32 | ||
| 47 | + ) | ||
| 39 | else: | 48 | else: |
| 40 | - scalars = np.array([scale_value for i in range(len(shape_list))]).astype(np_type) | 49 | + scalars = np.array([scale_value for i in range(len(shape_list))]).astype( |
| 50 | + np_type | ||
| 51 | + ) | ||
| 41 | for index, shape in enumerate(shape_list): | 52 | for index, shape in enumerate(shape_list): |
| 42 | - tmp_input = np.ones(shape)*10 | 53 | + tmp_input = np.ones(shape) * 10 |
| 43 | tmp_input = tmp_input.astype(np_type) | 54 | tmp_input = tmp_input.astype(np_type) |
| 44 | - tmp_tensor1 = np.ones(shape)*10 | 55 | + tmp_tensor1 = np.ones(shape) * 10 |
| 45 | tmp_tensor1 = tmp_tensor1.astype(np_type) | 56 | tmp_tensor1 = tmp_tensor1.astype(np_type) |
| 46 | - tmp_tensor2 = np.ones(shape)*10 | 57 | + tmp_tensor2 = np.ones(shape) * 10 |
| 47 | tmp_tensor2 = tmp_tensor2.astype(np_type) | 58 | tmp_tensor2 = tmp_tensor2.astype(np_type) |
| 48 | - tmp_golden = np.ones(shape)*10 | 59 | + tmp_golden = np.ones(shape) * 10 |
| 49 | tmp_golden = tmp_golden.astype(np_type) | 60 | tmp_golden = tmp_golden.astype(np_type) |
| 50 | 61 | ||
| 51 | for i in range(len(tmp_golden)): | 62 | for i in range(len(tmp_golden)): |
| 52 | - tmp_golden[i] = tmp_input[i] + tmp_tensor1[i] * tmp_tensor2[i] * scalars[index] | 63 | + tmp_golden[i] = ( |
| 53 | - | 64 | + tmp_input[i] + tmp_tensor1[i] * tmp_tensor2[i] * scalars[index] |
| 65 | + ) | ||
| 66 | + | ||
| 67 | + if out_shape_list is not None: | ||
| 68 | + # out 元素数大于 x1 的场景:golden 覆盖 out 全长(前 x1 numel 为计算值,尾部为标记值), | ||
| 69 | + # 另生成 outinit 文件(全长标记值)作为 out 的初始内存——kernel 只应写前 x1 numel 个元素 | ||
| 70 | + out_numel = int(np.prod(out_shape_list[index])) | ||
| 71 | + x_numel = int(np.prod(shape)) | ||
| 72 | + tmp_out_init = np.full((out_numel,), TAIL_MARKER, dtype=np_type) | ||
| 73 | + tmp_golden = np.full((out_numel,), TAIL_MARKER, dtype=np_type) | ||
| 74 | + # 用 flat 索引逐元素取值,兼容输入为多维 shape(如 {4,6})的场景 | ||
| 75 | + tmp_input_flat = tmp_input.flatten() | ||
| 76 | + tmp_tensor1_flat = tmp_tensor1.flatten() | ||
| 77 | + tmp_tensor2_flat = tmp_tensor2.flatten() | ||
| 78 | + for i in range(x_numel): | ||
| 79 | + tmp_golden[i] = ( | ||
| 80 | + tmp_input_flat[i] | ||
| 81 | + + tmp_tensor1_flat[i] * tmp_tensor2_flat[i] * scalars[index] | ||
| 82 | + ) | ||
| 83 | + | ||
| 54 | tmp_golden = tmp_golden.astype(np_type) | 84 | tmp_golden = tmp_golden.astype(np_type) |
| 55 | 85 | ||
| 56 | tmp_input.astype(np_type).tofile(f"{d_type}_input_t_foreach_addcmul{index}.bin") | 86 | tmp_input.astype(np_type).tofile(f"{d_type}_input_t_foreach_addcmul{index}.bin") |
| 57 | - tmp_tensor1.astype(np_type).tofile(f"{d_type}_tensor1_t_foreach_addcmul{index}.bin") | 87 | + tmp_tensor1.astype(np_type).tofile( |
| 58 | - tmp_tensor2.astype(np_type).tofile(f"{d_type}_tensor2_t_foreach_addcmul{index}.bin") | 88 | + f"{d_type}_tensor1_t_foreach_addcmul{index}.bin" |
| 59 | - tmp_golden.astype(np_type).tofile(f"{d_type}_golden_t_foreach_addcmul{index}.bin") | 89 | + ) |
| 90 | + tmp_tensor2.astype(np_type).tofile( | ||
| 91 | + f"{d_type}_tensor2_t_foreach_addcmul{index}.bin" | ||
| 92 | + ) | ||
| 93 | + tmp_golden.astype(np_type).tofile( | ||
| 94 | + f"{d_type}_golden_t_foreach_addcmul{index}.bin" | ||
| 95 | + ) | ||
| 96 | + if out_shape_list is not None: | ||
| 97 | + tmp_out_init.astype(np_type).tofile( | ||
| 98 | + f"{d_type}_outinit_t_foreach_addcmul{index}.bin" | ||
| 99 | + ) | ||
| 60 | 100 | ||
| 61 | 101 | ||
| 62 | if __name__ == "__main__": | 102 | if __name__ == "__main__": |
| 63 | - if len(sys.argv) != 4: | 103 | + if len(sys.argv) not in (4, 5): |
| 64 | - print("Param num must be 4.") | 104 | + print("Param num must be 4 or 5 (with optional out_shape).") |
| 65 | exit(1) | 105 | exit(1) |
| 66 | # 清理bin文件 | 106 | # 清理bin文件 |
| 67 | os.system("rm -rf *.bin") | 107 | os.system("rm -rf *.bin") |
| 68 | - gen_data_and_golden(sys.argv[1], sys.argv[2], sys.argv[3]) | 108 | + out_shape_str = sys.argv[4] if len(sys.argv) == 5 else None |
| 109 | + gen_data_and_golden(sys.argv[1], sys.argv[2], sys.argv[3], out_shape_str) | ||
| @@ -1,12 +1,11 @@ | |||
| 1 | /** | 1 | /** |
| 2 | - * This program is free software, you can redistribute it and/or modify. | 2 | + * Copyright (c) 2025-2026 Huawei Technologies Co., Ltd. |
| 3 | - * Copyright (c) 2025 Huawei Technologies Co., Ltd. | 3 | + * This program is free software, you can redistribute it and/or modify it under the terms and conditions of |
| 4 | - * This file is a part of the CANN Open Software. | 4 | + * CANN Open Software License Agreement Version 2.0 (the "License"). |
| 5 | - * Licensed under CANN Open Software License Agreement Version 2.0 (the "License"). | ||
| 6 | * 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. |
| 7 | - * THIS SOFTWARE IS PROVIDED ON AN "AS IS" BASIS, WITHOUT WARRANTIES OF ANY KIND, EITHER EXPRESS OR IMPLIED, INCLUDING | 6 | + * THIS SOFTWARE IS PROVIDED ON AN "AS IS" BASIS, WITHOUT WARRANTIES OF ANY KIND, EITHER EXPRESS OR IMPLIED, |
| 8 | - * BUT NOT LIMITED TO NON-INFRINGEMENT, MERCHANTABILITY, OR FITNESS FOR A PARTICULAR PURPOSE. See LICENSE in the root of | 7 | + * INCLUDING BUT NOT LIMITED TO NON-INFRINGEMENT, MERCHANTABILITY, OR FITNESS FOR A PARTICULAR PURPOSE. |
| 9 | - * 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. |
| 10 | */ | 9 | */ |
| 11 | 10 | ||
| 12 | /*! | 11 | /*! |
| @@ -193,3 +192,45 @@ TEST_F(foreach_addcmul_scalar_test, test_case_bfloat16_4) | |||
| 193 | 192 | ||
| 194 | system("cd ./addcmul_scalar_data/ && python3 compare_data.py 'bfloat16_t'"); | 193 | system("cd ./addcmul_scalar_data/ && python3 compare_data.py 'bfloat16_t'"); |
| 195 | } | 194 | } |
| 195 | + | ||
| 196 | +// out 元素数大于 x1(numel 32 >= 24)的执行级用例:验证 kernel 只写前 x1 numel 个元素、out 尾部保持初始值 | ||
| 197 | +TEST_F(foreach_addcmul_scalar_test, test_case_out_larger) | ||
| 198 | +{ | ||
| 199 | + std::vector<std::vector<uint64_t>> shapeInfos = {{4, 6}}; | ||
| 200 | + std::vector<std::vector<uint64_t>> outShapeInfos = {{32}}; | ||
| 201 | + system("cp -rf " | ||
| 202 | + "../../../../foreach/foreach_addcmul_scalar/tests/ut/op_kernel/addcmul_scalar_data ./"); | ||
| 203 | + system("chmod -R 755 ./addcmul_scalar_data/"); | ||
| 204 | + system("cd ./addcmul_scalar_data/ && python3 gen_data.py '{{4, 6}}' 3 'float32' '{32}'"); | ||
| 205 | + AscendC::SetKernelMode(KernelMode::AIV_MODE); | ||
| 206 | + uint32_t blockDim = 1; | ||
| 207 | + size_t sysWorkspaceSize = 16 * 1024 * 1024; | ||
| 208 | + uint8_t* workspace = (uint8_t*)AscendC::GmAlloc(sysWorkspaceSize); | ||
| 209 | + size_t tilingSize = sizeof(ForeachCommonTilingData); | ||
| 210 | + uint8_t* tiling = (uint8_t*)AscendC::GmAlloc(tilingSize); | ||
| 211 | + | ||
| 212 | + optiling::ForeachCommonTiling tilingFuncObj; | ||
| 213 | + tilingFuncObj.Init(shapeInfos, 1, 4); | ||
| 214 | + tilingFuncObj.RunBigKernelTiling(blockDim); | ||
| 215 | + tilingFuncObj.FillTilingData(reinterpret_cast<ForeachCommonTilingData*>(tiling)); | ||
| 216 | + | ||
| 217 | + uint8_t* x1 = CreateTensorListForeachAddcmulScalar<float>(shapeInfos, "float32", "input"); | ||
| 218 | + uint8_t* tensor1 = CreateTensorListForeachAddcmulScalar<float>(shapeInfos, "float32", "tensor1"); | ||
| 219 | + uint8_t* tensor2 = CreateTensorListForeachAddcmulScalar<float>(shapeInfos, "float32", "tensor2"); | ||
| 220 | + uint8_t* out = CreateTensorListForeachAddcmulScalar<float>(outShapeInfos, "float32", "outinit"); | ||
| 221 | + uint8_t* scalar = (uint8_t*)AscendC::GmAlloc(sizeof(float)); | ||
| 222 | + *((float*)scalar) = 3; | ||
| 223 | + | ||
| 224 | + ICPU_SET_TILING_KEY(2); | ||
| 225 | + ICPU_RUN_KF(foreach_addcmul_scalar, blockDim, x1, tensor1, tensor2, scalar, out, workspace, tiling); | ||
| 226 | + | ||
| 227 | + FreeTensorListForeachAddcmulScalar<float>(out, outShapeInfos, "float32"); | ||
| 228 | + AscendC::GmFree((void*)x1); | ||
| 229 | + AscendC::GmFree((void*)scalar); | ||
| 230 | + AscendC::GmFree((void*)tensor1); | ||
| 231 | + AscendC::GmFree((void*)tensor2); | ||
| 232 | + AscendC::GmFree((void*)workspace); | ||
| 233 | + AscendC::GmFree((void*)tiling); | ||
| 234 | + | ||
| 235 | + system("cd ./addcmul_scalar_data/ && python3 compare_data.py 'float32'"); | ||
| 236 | +} | ||
| @@ -149,8 +149,8 @@ aclnnStatus aclnnForeachAsin( | |||
| 149 | <td>传入的x或out是空指针。</td> | 149 | <td>传入的x或out是空指针。</td> |
| 150 | </tr> | 150 | </tr> |
| 151 | <tr> | 151 | <tr> |
| 152 | - <td rowspan="6">ACLNN_ERR_PARAM_INVALID</td> | 152 | + <td rowspan="7">ACLNN_ERR_PARAM_INVALID</td> |
| 153 | - <td rowspan="6">161002</td> | 153 | + <td rowspan="7">161002</td> |
| 154 | <td>x或out的数据类型不在支持的范围之内。</td> | 154 | <td>x或out的数据类型不在支持的范围之内。</td> |
| 155 | </tr> | 155 | </tr> |
| 156 | <tr> | 156 | <tr> |
| @@ -169,8 +169,6 @@ aclnnStatus aclnnForeachAsin( | |||
| 169 | <td>x与out中对应位置的Tensor的shape不一致。</td> | 169 | <td>x与out中对应位置的Tensor的shape不一致。</td> |
| 170 | </tr> | 170 | </tr> |
| 171 | <tr> | 171 | <tr> |
| 172 | - <td>ACLNN_ERR_INNER_TILING_ERROR</td> | ||
| 173 | - <td>561002</td> | ||
| 174 | <td>x或out中单个Tensor列表的Tensor数量超过256。</td> | 172 | <td>x或out中单个Tensor列表的Tensor数量超过256。</td> |
| 175 | </tr> | 173 | </tr> |
| 176 | </tbody></table> | 174 | </tbody></table> |
| @@ -151,24 +151,24 @@ aclnnStatus aclnnForeachAsin( | |||
| 151 | <td>x or out is a null pointer.</td> | 151 | <td>x or out is a null pointer.</td> |
| 152 | </tr> | 152 | </tr> |
| 153 | <tr> | 153 | <tr> |
| 154 | - <td rowspan="2">ACLNN_ERR_PARAM_INVALID</td> | 154 | + <td rowspan="3">ACLNN_ERR_PARAM_INVALID</td> |
| 155 | - <td rowspan="2">161002</td> | 155 | + <td rowspan="3">161002</td> |
| 156 | <td>The data type of x or out is not supported.</td> | 156 | <td>The data type of x or out is not supported.</td> |
| 157 | </tr> | 157 | </tr> |
| 158 | <tr> | 158 | <tr> |
| 159 | <td>The data types of x and out are inconsistent.</td> | 159 | <td>The data types of x and out are inconsistent.</td> |
| 160 | </tr> | 160 | </tr> |
| 161 | <tr> | 161 | <tr> |
| 162 | - <td rowspan="3">ACLNN_ERR_INNER_TILING_ERROR</td> | 162 | + <td>The dimensions of the tensors in x or out exceed 8.</td> |
| 163 | - <td rowspan="3">561002</td> | 163 | + </tr> |
| 164 | + <tr> | ||
| 165 | + <td rowspan="2">ACLNN_ERR_INNER_TILING_ERROR</td> | ||
| 166 | + <td rowspan="2">561002</td> | ||
| 164 | <td>The shapes of x and out do not meet the constraints.</td> | 167 | <td>The shapes of x and out do not meet the constraints.</td> |
| 165 | </tr> | 168 | </tr> |
| 166 | <tr> | 169 | <tr> |
| 167 | <td>The data types of the tensors in x or out are inconsistent.</td> | 170 | <td>The data types of the tensors in x or out are inconsistent.</td> |
| 168 | </tr> | 171 | </tr> |
| 169 | - <tr> | ||
| 170 | - <td>The dimensions of the tensors in x or out exceed 8.</td> | ||
| 171 | - </tr> | ||
| 172 | </tbody></table> | 172 | </tbody></table> |
| 173 | 173 | ||
| 174 | ## aclnnForeachAsin | 174 | ## aclnnForeachAsin |
| @@ -149,24 +149,24 @@ aclnnStatus aclnnForeachAtan( | |||
| 149 | <td>传入的x或out是空指针。</td> | 149 | <td>传入的x或out是空指针。</td> |
| 150 | </tr> | 150 | </tr> |
| 151 | <tr> | 151 | <tr> |
| 152 | - <td rowspan="2">ACLNN_ERR_PARAM_INVALID</td> | 152 | + <td rowspan="3">ACLNN_ERR_PARAM_INVALID</td> |
| 153 | - <td rowspan="2">161002</td> | 153 | + <td rowspan="3">161002</td> |
| 154 | <td>x或out的数据类型不在支持的范围之内。</td> | 154 | <td>x或out的数据类型不在支持的范围之内。</td> |
| 155 | </tr> | 155 | </tr> |
| 156 | <tr> | 156 | <tr> |
| 157 | <td>x和out的数据类型不一致。</td> | 157 | <td>x和out的数据类型不一致。</td> |
| 158 | </tr> | 158 | </tr> |
| 159 | <tr> | 159 | <tr> |
| 160 | - <td rowspan="3">ACLNN_ERR_INNER_TILING_ERROR</td> | 160 | + <td>x或out中的Tensor维度超过8维。</td> |
| 161 | - <td rowspan="3">561002</td> | 161 | + </tr> |
| 162 | + <tr> | ||
| 163 | + <td rowspan="2">ACLNN_ERR_INNER_TILING_ERROR</td> | ||
| 164 | + <td rowspan="2">561002</td> | ||
| 162 | <td>x与out的shape不满足约束。</td> | 165 | <td>x与out的shape不满足约束。</td> |
| 163 | </tr> | 166 | </tr> |
| 164 | <tr> | 167 | <tr> |
| 165 | <td>x或out中的Tensor的数据类型不一致。</td> | 168 | <td>x或out中的Tensor的数据类型不一致。</td> |
| 166 | </tr> | 169 | </tr> |
| 167 | - <tr> | ||
| 168 | - <td>x或out中的Tensor维度超过8维。</td> | ||
| 169 | - </tr> | ||
| 170 | </tbody></table> | 170 | </tbody></table> |
| 171 | 171 | ||
| 172 | ## aclnnForeachAtan | 172 | ## aclnnForeachAtan |
| @@ -151,25 +151,26 @@ aclnnStatus aclnnForeachAtan( | |||
| 151 | <td>x or out is a null pointer.</td> | 151 | <td>x or out is a null pointer.</td> |
| 152 | </tr> | 152 | </tr> |
| 153 | <tr> | 153 | <tr> |
| 154 | - <td rowspan="2">ACLNN_ERR_PARAM_INVALID</td> | 154 | + <td rowspan="3">ACLNN_ERR_PARAM_INVALID</td> |
| 155 | - <td rowspan="2">161002</td> | 155 | + <td rowspan="3">161002</td> |
| 156 | <td>The data type of x or out is not in the supported range.</td> | 156 | <td>The data type of x or out is not in the supported range.</td> |
| 157 | </tr> | 157 | </tr> |
| 158 | <tr> | 158 | <tr> |
| 159 | <td>The data types of x and out are inconsistent.</td> | 159 | <td>The data types of x and out are inconsistent.</td> |
| 160 | </tr> | 160 | </tr> |
| 161 | <tr> | 161 | <tr> |
| 162 | - <td rowspan="3">ACLNN_ERR_INNER_TILING_ERROR</td> | 162 | + <td>The dimensions of tensors in x or out exceed 8.</td> |
| 163 | - <td rowspan="3">561002</td> | 163 | + </tr> |
| 164 | + <tr> | ||
| 165 | + <td rowspan="2">ACLNN_ERR_INNER_TILING_ERROR</td> | ||
| 166 | + <td rowspan="2">561002</td> | ||
| 164 | <td>The shapes of x and out do not meet the constraints.</td> | 167 | <td>The shapes of x and out do not meet the constraints.</td> |
| 165 | </tr> | 168 | </tr> |
| 166 | <tr> | 169 | <tr> |
| 167 | <td>The data types of Tensors in x or out are inconsistent.</td> | 170 | <td>The data types of Tensors in x or out are inconsistent.</td> |
| 168 | </tr> | 171 | </tr> |
| 169 | <tr> | 172 | <tr> |
| 170 | - <td>The dimensions of Tensors in x or out exceed 8.</td> | 173 | + </tbody></table> |
| 171 | - </tr> | ||
| 172 | - </tbody></table> | ||
| 173 | 174 | ||
| 174 | ## aclnnForeachAtan | 175 | ## aclnnForeachAtan |
| 175 | 176 | ||
| @@ -107,7 +107,7 @@ aclnnStatus aclnnForeachDivScalar( | |||
| 107 | <td>out(aclTensorList*)</td> | 107 | <td>out(aclTensorList*)</td> |
| 108 | <td>输出</td> | 108 | <td>输出</td> |
| 109 | <td>表示张量x除以张量scalar的输出张量列表,对应公式中的`y`。</td> | 109 | <td>表示张量x除以张量scalar的输出张量列表,对应公式中的`y`。</td> |
| 110 | - <td><ul><li>支持空Tensor。</li><li>该参数中所有Tensor的数据类型保持一致。</li><li>数据类型和数据格式与入参`x`的数据类型和数据格式一致,shape size大于等于入参`x`的shape size。</li></ul></td> | 110 | + <td><ul><li>支持空Tensor。</li><li>该参数中所有Tensor的数据类型保持一致。</li><li>数据类型和数据格式与入参`x`的数据类型和数据格式一致,shape与入参`x`一致。</li></ul></td> |
| 111 | <td>BFLOAT16、FLOAT16、FLOAT32</td> | 111 | <td>BFLOAT16、FLOAT16、FLOAT32</td> |
| 112 | <td>ND</td> | 112 | <td>ND</td> |
| 113 | <td>0-8</td> | 113 | <td>0-8</td> |
| @@ -161,16 +161,14 @@ aclnnStatus aclnnForeachDivScalar( | |||
| 161 | <td>传入的x、scalar、out是空指针。</td> | 161 | <td>传入的x、scalar、out是空指针。</td> |
| 162 | </tr> | 162 | </tr> |
| 163 | <tr> | 163 | <tr> |
| 164 | - <td rowspan="2">ACLNN_ERR_PARAM_INVALID</td> | 164 | + <td rowspan="5">ACLNN_ERR_PARAM_INVALID</td> |
| 165 | - <td rowspan="2">161002</td> | 165 | + <td rowspan="5">161002</td> |
| 166 | <td>x、scalar、out的数据类型不在支持的范围之内。</td> | 166 | <td>x、scalar、out的数据类型不在支持的范围之内。</td> |
| 167 | </tr> | 167 | </tr> |
| 168 | <tr> | 168 | <tr> |
| 169 | <td>x和out的数据类型不一致。</td> | 169 | <td>x和out的数据类型不一致。</td> |
| 170 | </tr> | 170 | </tr> |
| 171 | <tr> | 171 | <tr> |
| 172 | - <td rowspan="4">ACLNN_ERR_INNER_TILING_ERROR</td> | ||
| 173 | - <td rowspan="4">561002</td> | ||
| 174 | <td>x与out的shape不满足约束。</td> | 172 | <td>x与out的shape不满足约束。</td> |
| 175 | </tr> | 173 | </tr> |
| 176 | <tr> | 174 | <tr> |
| @@ -180,6 +178,8 @@ aclnnStatus aclnnForeachDivScalar( | |||
| 180 | <td>x或out中的Tensor的数据类型不一致。</td> | 178 | <td>x或out中的Tensor的数据类型不一致。</td> |
| 181 | </tr> | 179 | </tr> |
| 182 | <tr> | 180 | <tr> |
| 181 | + <td>ACLNN_ERR_INNER_TILING_ERROR</td> | ||
| 182 | + <td>561002</td> | ||
| 183 | <td>scalar元素个数不为1。</td> | 183 | <td>scalar元素个数不为1。</td> |
| 184 | </tr> | 184 | </tr> |
| 185 | </tbody></table> | 185 | </tbody></table> |
| @@ -109,7 +109,7 @@ aclnnStatus aclnnForeachDivScalar( | |||
| 109 | <td>out (aclTensorList*)</td> | 109 | <td>out (aclTensorList*)</td> |
| 110 | <td>Output</td> | 110 | <td>Output</td> |
| 111 | <td>Output tensor list of tensor x divided by tensor scalar, corresponding to `y` in the formula.</td> | 111 | <td>Output tensor list of tensor x divided by tensor scalar, corresponding to `y` in the formula.</td> |
| 112 | - <td><ul><li>Empty tensors are supported.</li><li>The data types of all tensors in this parameter are the same.</li><li>The data type and data format are the same as those of input parameter `x`, and the shape size is greater than or equal to that of input parameter `x`.</li></ul></td> | 112 | + <td><ul><li>Empty tensors are supported.</li><li>The data types of all tensors in this parameter are the same.</li><li>The data type and data format are the same as those of input parameter `x`, and the shape is the same as that of input parameter `x`.</li></ul></td> |
| 113 | <td>BFLOAT16, FLOAT16, FLOAT32</td> | 113 | <td>BFLOAT16, FLOAT16, FLOAT32</td> |
| 114 | <td>ND</td> | 114 | <td>ND</td> |
| 115 | <td>0-8</td> | 115 | <td>0-8</td> |
| @@ -163,16 +163,14 @@ aclnnStatus: return status code. For details, see [aclnn Return Codes](../../../ | |||
| 163 | <td>x, scalar, or out is a null pointer.</td> | 163 | <td>x, scalar, or out is a null pointer.</td> |
| 164 | </tr> | 164 | </tr> |
| 165 | <tr> | 165 | <tr> |
| 166 | - <td rowspan="2">ACLNN_ERR_PARAM_INVALID</td> | 166 | + <td rowspan="5">ACLNN_ERR_PARAM_INVALID</td> |
| 167 | - <td rowspan="2">161002</td> | 167 | + <td rowspan="5">161002</td> |
| 168 | <td>The data types of x, scalar, or out are not within the supported range.</td> | 168 | <td>The data types of x, scalar, or out are not within the supported range.</td> |
| 169 | </tr> | 169 | </tr> |
| 170 | <tr> | 170 | <tr> |
| 171 | <td>The data types of x and out are inconsistent.</td> | 171 | <td>The data types of x and out are inconsistent.</td> |
| 172 | </tr> | 172 | </tr> |
| 173 | <tr> | 173 | <tr> |
| 174 | - <td rowspan="4">ACLNN_ERR_INNER_TILING_ERROR</td> | ||
| 175 | - <td rowspan="4">561002</td> | ||
| 176 | <td>The shapes of x and out do not meet the constraints.</td> | 174 | <td>The shapes of x and out do not meet the constraints.</td> |
| 177 | </tr> | 175 | </tr> |
| 178 | <tr> | 176 | <tr> |
| @@ -182,6 +180,8 @@ aclnnStatus: return status code. For details, see [aclnn Return Codes](../../../ | |||
| 182 | <td>The data types of tensors in x or out are inconsistent.</td> | 180 | <td>The data types of tensors in x or out are inconsistent.</td> |
| 183 | </tr> | 181 | </tr> |
| 184 | <tr> | 182 | <tr> |
| 183 | + <td>ACLNN_ERR_INNER_TILING_ERROR</td> | ||
| 184 | + <td>561002</td> | ||
| 185 | <td>The number of elements in scalar is not 1.</td> | 185 | <td>The number of elements in scalar is not 1.</td> |
| 186 | </tr> | 186 | </tr> |
| 187 | </tbody></table> | 187 | </tbody></table> |
| @@ -183,16 +183,14 @@ aclnnStatus aclnnForeachLerpScalar( | |||
| 183 | <td>传入的x1、x2、out是空指针。</td> | 183 | <td>传入的x1、x2、out是空指针。</td> |
| 184 | </tr> | 184 | </tr> |
| 185 | <tr> | 185 | <tr> |
| 186 | - <td rowspan="2">ACLNN_ERR_PARAM_INVALID</td> | 186 | + <td rowspan="5">ACLNN_ERR_PARAM_INVALID</td> |
| 187 | - <td rowspan="2">161002</td> | 187 | + <td rowspan="5">161002</td> |
| 188 | <td>x1、x2、weight、out的数据类型不在支持的范围之内。</td> | 188 | <td>x1、x2、weight、out的数据类型不在支持的范围之内。</td> |
| 189 | </tr> | 189 | </tr> |
| 190 | <tr> | 190 | <tr> |
| 191 | <td>x1、x2、weight、out的数据类型不一致。</td> | 191 | <td>x1、x2、weight、out的数据类型不一致。</td> |
| 192 | </tr> | 192 | </tr> |
| 193 | <tr> | 193 | <tr> |
| 194 | - <td rowspan="3">ACLNN_ERR_INNER_TILING_ERROR</td> | ||
| 195 | - <td rowspan="3">561002</td> | ||
| 196 | <td>x1、x2、out的shape不满足对应关系。</td> | 194 | <td>x1、x2、out的shape不满足对应关系。</td> |
| 197 | </tr> | 195 | </tr> |
| 198 | <tr> | 196 | <tr> |
| @@ -185,16 +185,14 @@ aclnnStatus aclnnForeachLerpScalar( | |||
| 185 | <td>x1, x2, or out is a null pointer.</td> | 185 | <td>x1, x2, or out is a null pointer.</td> |
| 186 | </tr> | 186 | </tr> |
| 187 | <tr> | 187 | <tr> |
| 188 | - <td rowspan="2">ACLNN_ERR_PARAM_INVALID</td> | 188 | + <td rowspan="5">ACLNN_ERR_PARAM_INVALID</td> |
| 189 | - <td rowspan="2">161002</td> | 189 | + <td rowspan="5">161002</td> |
| 190 | <td>The data types of x1, x2, weight, and out are not within the supported range.</td> | 190 | <td>The data types of x1, x2, weight, and out are not within the supported range.</td> |
| 191 | </tr> | 191 | </tr> |
| 192 | <tr> | 192 | <tr> |
| 193 | <td>The data types of x1, x2, weight, and out are inconsistent.</td> | 193 | <td>The data types of x1, x2, weight, and out are inconsistent.</td> |
| 194 | </tr> | 194 | </tr> |
| 195 | <tr> | 195 | <tr> |
| 196 | - <td rowspan="3">ACLNN_ERR_INNER_TILING_ERROR</td> | ||
| 197 | - <td rowspan="3">561002</td> | ||
| 198 | <td>The shapes of x1, x2, and out do not satisfy the corresponding relationship.</td> | 196 | <td>The shapes of x1, x2, and out do not satisfy the corresponding relationship.</td> |
| 199 | </tr> | 197 | </tr> |
| 200 | <tr> | 198 | <tr> |