Line data Source code
1 : /**
2 : * 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 : * 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.
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.
8 : * See LICENSE in the root of the software repository for the full text of the License.
9 : */
10 :
11 : #ifndef AICPUSD_AICPUSD_INFO_H
12 : #define AICPUSD_AICPUSD_INFO_H
13 :
14 : #include <cstdint>
15 : #include <sched.h>
16 : #include <sys/types.h>
17 : #include "../common/type_def.h"
18 :
19 : extern "C" {
20 : struct __attribute__((visibility("default"))) AICPUActiveStream {
21 : uint32_t streamId;
22 : };
23 :
24 : static const uint32_t MAX_CUST_SO_NAME_LEN = 128U;
25 : static const uint32_t OP_NAME_MAX_LEN = 50U;
26 : static const uint32_t FUN_NAME_MAX_LEN = 30U;
27 : static const uint32_t ERROR_KEY_INFO_MAX_LEN = 30U;
28 : static const uint32_t MODULE_NAME_MAX_LEN = 10U;
29 : static const uint32_t FILE_NAME_MAX_LEN = 30U;
30 : static const uint16_t PRIORITY_MSG_CHECKCODE = 0xABCD;
31 : static const int32_t INVALID_ESCAPE_PRI_VALUE = -1;
32 : // loadOpFromBuf task args
33 : struct __attribute__((visibility("default"))) LoadOpFromBufArgs {
34 : uint64_t kernelSoBuf; // the starting address of custom operator so buf
35 : uint32_t kernelSoBufLen; // the length of custom operator so buf
36 : uint64_t kernelSoName; // the starting address of custom operator so name
37 : uint32_t kernelSoNameLen; // the length of custom operator so name
38 : } __attribute__((packed));
39 :
40 : // batchLoadOpFromBuf task args
41 : struct __attribute__((visibility("default"))) BatchLoadOpFromBufArgs {
42 : uint32_t soNum; // the number of so
43 : uint64_t opInfoArgs;
44 : } __attribute__((packed));
45 :
46 : /**
47 : * The mode of profiling
48 : */
49 : enum __attribute__((visibility("default"))) ProfilingMode {
50 : PROFILING_CLOSE = 0,
51 : PROFILING_OPEN,
52 : };
53 :
54 : enum __attribute__((visibility("default"))) AicpuSchedMode {
55 : SCHED_MODE_INTERRUPT = 0,
56 : SCHED_MODE_MSGQ,
57 : SCHED_MODE_INVALID
58 : };
59 :
60 : enum __attribute__((visibility("default"))) AICPUSubEvent {
61 : AICPU_SUB_EVENT_ACTIVE_STREAM = 0,
62 : AICPU_SUB_EVENT_EXECUTE_MODEL,
63 : AICPU_SUB_EVENT_REPEAT_MODEL,
64 : AICPU_SUB_EVENT_RECOVERY_STREAM,
65 : AICPU_SUB_EVENT_UPDATE_PROFILING_MODE,
66 : AICPU_SUB_EVENT_LOAD_SO,
67 : AICPU_SUB_EVENT_END_GRAPH,
68 : AICPU_SUB_EVENT_ACTIVE_MODEL,
69 : AICPU_SUB_EVENT_PREPARE_MEM,
70 : AICPU_SUB_EVENT_TABLE_UNLOCK,
71 : AICPU_SUB_EVENT_SUPPLY_ENQUEUE,
72 : AICPU_SUB_EVENT_MAX_NUM,
73 : };
74 :
75 : enum __attribute__((visibility("default"))) AICPUCustSubEvent {
76 : // sub type begin with 10 for interface event
77 : AICPU_SUB_EVENT_BIND_SD_PID = 10, // cust-sd bind sd pid, Implemented by cust-sd
78 : AICPU_SUB_EVENT_OPEN_CUSTOM_SO, // open costom so file, Implemented by cust-sd
79 : AICPU_SUB_EVENT_CUST_UPDATE_PROFILING_MODE, // update profiling mode, Implemented by cust-sd
80 : AICPU_SUB_EVENT_ABNORMAL_LOG, // print aicpu cust schedule's error log
81 : AICPU_SUB_EVENT_REPORT_CUST_DUMPDATA, // aicpusd do cust datadump
82 : AICPU_SUB_EVENT_REPORT_UDF_DUMPDATA, // aicpusd do udf datadump
83 : AICPU_SUB_EVENT_CUST_LOAD_PLATFORM, // custom scheduler process load platform info event
84 : };
85 :
86 : struct __attribute__((visibility("default"))) AICPUSubEventStreamInfo {
87 : uint32_t streamId;
88 : };
89 :
90 : struct __attribute__((visibility("default"))) AICPUProfilingModeInfo {
91 : uint32_t deviceId;
92 : pid_t hostpId;
93 : uint32_t flag;
94 : };
95 :
96 : struct __attribute__((visibility("default"))) AICPULoadSoInfo {
97 : uint32_t kernelSoIndex;
98 : };
99 :
100 : struct __attribute__((visibility("default"))) AICPUEndGraphInfo {
101 : uint32_t result;
102 : };
103 :
104 : struct __attribute__((visibility("default"))) AICPUSharderTaskInfo {
105 : uint32_t parallelId;
106 : int64_t shardNum;
107 :
108 185 : bool operator==(const AICPUSharderTaskInfo& sharderInfo) const noexcept
109 : {
110 185 : return (parallelId == sharderInfo.parallelId);
111 : }
112 : };
113 :
114 : struct __attribute__((visibility("default"))) AICPUUnLockTableInfo {
115 : uint32_t tableId;
116 : };
117 :
118 : struct __attribute__((visibility("default"))) AICPUDumpCustInfo {
119 : uint32_t threadIndex; // dump任务的线程号
120 : int32_t retCode; // dump result
121 : uint32_t streamId; // dump key streamId
122 : uint32_t taskId; // dump key taskId
123 : };
124 :
125 : struct __attribute__((visibility("default"))) AICPUDumpUdfInfo {
126 : uint64_t length;
127 : uint64_t udfInfo;
128 : uint32_t udfPid;
129 : char_t rsv[20];
130 : };
131 :
132 : struct __attribute__((visibility("default"))) AICPULoadPlatformCustInfo {
133 : uint64_t length;
134 : uint64_t platformInfo;
135 : uint32_t aicpuPid;
136 : char_t rsv[20];
137 : };
138 :
139 : struct __attribute__((visibility("default"))) AICPUSubEventInfo {
140 : uint32_t modelId;
141 : union {
142 : AICPUSubEventStreamInfo streamInfo;
143 : AICPUProfilingModeInfo modeInfo;
144 : AICPULoadSoInfo loadSoInfo;
145 : AICPUEndGraphInfo endGraphInfo;
146 : AICPUSharderTaskInfo sharderTaskInfo;
147 : AICPUUnLockTableInfo unlockTableInfo;
148 : } para;
149 : };
150 :
151 : struct __attribute__((visibility("default"))) AICPUBindSdPidEventMsg {
152 : int32_t pid;
153 : } __attribute__((packed));
154 :
155 : struct __attribute__((visibility("default"))) AICPUOpenCustomSoEventMsg {
156 : char_t kernelSoName[MAX_CUST_SO_NAME_LEN];
157 : } __attribute__((packed));
158 :
159 : struct __attribute__((visibility("default"))) CpuSchedInitParam {
160 : uint32_t deviceId;
161 : pid_t hostPid;
162 : ProfilingMode profilingMode;
163 : char_t rsv[128];
164 : } __attribute__((packed));
165 :
166 : struct __attribute__((visibility("default"))) ModelQueueInfo {
167 : uint32_t queueId;
168 : uint32_t flag;
169 : } __attribute__((packed));
170 :
171 : struct __attribute__((visibility("default"))) ModelTaskInfo {
172 : uint32_t taskId;
173 : uint64_t kernelName;
174 : uint64_t paraBase; // param地址
175 : } __attribute__((packed));
176 :
177 : struct __attribute__((visibility("default"))) ModelStreamInfo {
178 : uint32_t streamId;
179 : uint32_t streamFlag;
180 : uint16_t taskNum;
181 : ModelTaskInfo* tasks;
182 : } __attribute__((packed));
183 :
184 : struct __attribute__((visibility("default"))) AicpuPriInfo {
185 : uint16_t checkHead;
186 : int32_t pidPriority;
187 : int32_t eventPriority;
188 : } __attribute__((packed));
189 :
190 : struct __attribute__((visibility("default"))) ModelCommOpList {
191 : uint64_t commOpDescsListAddr;
192 : uint32_t opNum;
193 : } __attribute__((packed));
194 :
195 : // 配置类型，用于给AICPU控制不同的加载流程
196 : enum __attribute__((visibility("default"))) ModelType {
197 : kModelWithEmbedding = 0,
198 : kModelWithSyncEvent = 1,
199 : kModelTypeNum
200 : };
201 :
202 : struct __attribute__((visibility("default"))) CommGroup {
203 : const char* groupName;
204 : uint32_t rankNum;
205 : uint32_t* rankIds;
206 : } __attribute__((packed));
207 :
208 : struct __attribute__((visibility("default"))) CommGroups {
209 : uint32_t groupNum;
210 : CommGroup* groups;
211 : } __attribute__((packed));
212 :
213 : struct __attribute__((visibility("default"))) ModelCfgInfo {
214 : uint64_t inBuffPoolSizeAddr; // input buffer block num array's address, the array is uint16_t[]
215 : uint64_t outBuffPoolSizeAddr; // output buffer block num array's address, the array is uint16_t[]
216 : uint64_t inBuffSizeAddr; // input block size array's address, the array is uint64_t[]
217 : uint64_t outBuffSizeAddr; // output block size array's address, the array is uint64_t[]
218 : uint32_t inputNum;
219 : uint32_t outputNum;
220 : int32_t tagId; // tag id for hccl
221 : int32_t rankId; // rank id for hccl
222 : uint64_t rankTableLen; // rank table length
223 : uint64_t rankTableAddr; // rank table ptr
224 : uint64_t roleTableLen; // cluster spec length
225 : uint64_t roleTableAddr; // role table ptr
226 : uint64_t modelCommOpListAddr; // ModelCommOpList ptr
227 : uint32_t modelType;
228 : uint64_t commGroupsAddr; // communication groups ptr
229 : int32_t psId; // ps id
230 : bool supportCounterFilter; // counter filter flag
231 : bool memoryRegister;
232 : uint64_t clientRankNum; // clientRank array's length
233 : uint64_t clientRankAddr; // clientRank array's address, the array is uint32_t[]
234 : uint64_t hcclCommNameAddr; // hcclComm name's addr, the name is char*
235 : int32_t hcclTimeOut; // 0 means default, -1 means never timeout
236 : bool associateWorker; // true means we need to associate worker when load model, default false
237 : char rsv[53UL];
238 : } __attribute__((packed));
239 :
240 : struct __attribute__((visibility("default"))) ModelInfo {
241 : uint32_t modelId;
242 : uint16_t aicpuStreamNum;
243 : ModelStreamInfo* streams;
244 : uint16_t queueNum;
245 : ModelQueueInfo* queues;
246 : int32_t abnormalBreak;
247 : int32_t abnormalEnqueue;
248 : AicpuPriInfo aicpuPriInfo;
249 : uint64_t cfgInfoPtr = 0;
250 : int32_t abnormalEnable = 1;
251 : char rsv[98];
252 : } __attribute__((packed));
253 :
254 : struct __attribute__((visibility("default"))) ReDeployConfig {
255 : uint64_t modelIdsAddr; // ptr which point to modelIds(uint32_t)
256 : uint32_t modelIdNum; // modelIdNum
257 : char rsv[4];
258 : } __attribute__((packed));
259 :
260 : struct __attribute__((visibility("default"))) CheckKernelSupportedConfig {
261 : uint64_t kernelNameAddr; // ptr which point to kernelName
262 : uint32_t kernelNameLen;
263 : uint64_t checkResultAddr; // int32: 0 is supported, others are not supported
264 : uint32_t checkResultLen;
265 : } __attribute__((packed));
266 :
267 : struct __attribute__((visibility("default"))) DataFlowExceptionNotify {
268 : uint64_t transId;
269 : uint32_t type; // 0:Exception occured, 1:Exception expired
270 : uint32_t modelIdNum;
271 : uint64_t modelIdsAddr; // ptr which point to modelIds(uint32_t)
272 : char rsv[40];
273 : } __attribute__((packed));
274 :
275 : struct __attribute__((visibility("default"))) SomaMemMng {
276 : uint64_t size;
277 : uint64_t va; // Virtual address
278 : uint64_t mempoolId; // Id of memory pool
279 : uint32_t deviceId;
280 : int32_t memAsyncOpType; // Malloc(0) Free(1)
281 : int32_t memAsyncSubCMD; // Distinguish between Malloc reuse strategies and Free explicit/implicit reuse methods
282 : char rsv[4];
283 : } __attribute__((packed));
284 : }
285 : #endif // AICPUSD_AICPUSD_INFO_H
|
