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 : AICPU_SUB_EVENT_CUST_CLOSE_MONITOR, // notify cust-sd to close aicpu kernel timeout monitor
85 : };
86 :
87 : struct __attribute__((visibility("default"))) AICPUSubEventStreamInfo {
88 : uint32_t streamId;
89 : };
90 :
91 : struct __attribute__((visibility("default"))) AICPUProfilingModeInfo {
92 : uint32_t deviceId;
93 : pid_t hostpId;
94 : uint32_t flag;
95 : };
96 :
97 : struct __attribute__((visibility("default"))) AICPULoadSoInfo {
98 : uint32_t kernelSoIndex;
99 : };
100 :
101 : struct __attribute__((visibility("default"))) AICPUEndGraphInfo {
102 : uint32_t result;
103 : };
104 :
105 : struct __attribute__((visibility("default"))) AICPUSharderTaskInfo {
106 : uint32_t parallelId;
107 : int64_t shardNum;
108 :
109 185 : bool operator==(const AICPUSharderTaskInfo& sharderInfo) const noexcept
110 : {
111 185 : return (parallelId == sharderInfo.parallelId);
112 : }
113 : };
114 :
115 : struct __attribute__((visibility("default"))) AICPUUnLockTableInfo {
116 : uint32_t tableId;
117 : };
118 :
119 : struct __attribute__((visibility("default"))) AICPUDumpCustInfo {
120 : uint32_t threadIndex; // dump任务的线程号
121 : int32_t retCode; // dump result
122 : uint32_t streamId; // dump key streamId
123 : uint32_t taskId; // dump key taskId
124 : };
125 :
126 : struct __attribute__((visibility("default"))) AICPUDumpUdfInfo {
127 : uint64_t length;
128 : uint64_t udfInfo;
129 : uint32_t udfPid;
130 : char_t rsv[20];
131 : };
132 :
133 : struct __attribute__((visibility("default"))) AICPULoadPlatformCustInfo {
134 : uint64_t length;
135 : uint64_t platformInfo;
136 : uint32_t aicpuPid;
137 : char_t rsv[20];
138 : };
139 :
140 : struct __attribute__((visibility("default"))) AICPUSubEventInfo {
141 : uint32_t modelId;
142 : union {
143 : AICPUSubEventStreamInfo streamInfo;
144 : AICPUProfilingModeInfo modeInfo;
145 : AICPULoadSoInfo loadSoInfo;
146 : AICPUEndGraphInfo endGraphInfo;
147 : AICPUSharderTaskInfo sharderTaskInfo;
148 : AICPUUnLockTableInfo unlockTableInfo;
149 : } para;
150 : };
151 :
152 : struct __attribute__((visibility("default"))) AICPUBindSdPidEventMsg {
153 : int32_t pid;
154 : } __attribute__((packed));
155 :
156 : struct __attribute__((visibility("default"))) AICPUOpenCustomSoEventMsg {
157 : char_t kernelSoName[MAX_CUST_SO_NAME_LEN];
158 : } __attribute__((packed));
159 :
160 : struct __attribute__((visibility("default"))) AICPUCloseMonitorEventMsg {
161 : uint8_t closeFlag;
162 : } __attribute__((packed));
163 :
164 : struct __attribute__((visibility("default"))) CpuSchedInitParam {
165 : uint32_t deviceId;
166 : pid_t hostPid;
167 : ProfilingMode profilingMode;
168 : char_t rsv[128];
169 : } __attribute__((packed));
170 :
171 : struct __attribute__((visibility("default"))) ModelQueueInfo {
172 : uint32_t queueId;
173 : uint32_t flag;
174 : } __attribute__((packed));
175 :
176 : struct __attribute__((visibility("default"))) ModelTaskInfo {
177 : uint32_t taskId;
178 : uint64_t kernelName;
179 : uint64_t paraBase; // param地址
180 : } __attribute__((packed));
181 :
182 : struct __attribute__((visibility("default"))) ModelStreamInfo {
183 : uint32_t streamId;
184 : uint32_t streamFlag;
185 : uint16_t taskNum;
186 : ModelTaskInfo* tasks;
187 : } __attribute__((packed));
188 :
189 : struct __attribute__((visibility("default"))) AicpuPriInfo {
190 : uint16_t checkHead;
191 : int32_t pidPriority;
192 : int32_t eventPriority;
193 : } __attribute__((packed));
194 :
195 : struct __attribute__((visibility("default"))) ModelCommOpList {
196 : uint64_t commOpDescsListAddr;
197 : uint32_t opNum;
198 : } __attribute__((packed));
199 :
200 : // 配置类型，用于给AICPU控制不同的加载流程
201 : enum __attribute__((visibility("default"))) ModelType {
202 : kModelWithEmbedding = 0,
203 : kModelWithSyncEvent = 1,
204 : kModelTypeNum
205 : };
206 :
207 : struct __attribute__((visibility("default"))) CommGroup {
208 : const char* groupName;
209 : uint32_t rankNum;
210 : uint32_t* rankIds;
211 : } __attribute__((packed));
212 :
213 : struct __attribute__((visibility("default"))) CommGroups {
214 : uint32_t groupNum;
215 : CommGroup* groups;
216 : } __attribute__((packed));
217 :
218 : struct __attribute__((visibility("default"))) ModelCfgInfo {
219 : uint64_t inBuffPoolSizeAddr; // input buffer block num array's address, the array is uint16_t[]
220 : uint64_t outBuffPoolSizeAddr; // output buffer block num array's address, the array is uint16_t[]
221 : uint64_t inBuffSizeAddr; // input block size array's address, the array is uint64_t[]
222 : uint64_t outBuffSizeAddr; // output block size array's address, the array is uint64_t[]
223 : uint32_t inputNum;
224 : uint32_t outputNum;
225 : int32_t tagId; // tag id for hccl
226 : int32_t rankId; // rank id for hccl
227 : uint64_t rankTableLen; // rank table length
228 : uint64_t rankTableAddr; // rank table ptr
229 : uint64_t roleTableLen; // cluster spec length
230 : uint64_t roleTableAddr; // role table ptr
231 : uint64_t modelCommOpListAddr; // ModelCommOpList ptr
232 : uint32_t modelType;
233 : uint64_t commGroupsAddr; // communication groups ptr
234 : int32_t psId; // ps id
235 : bool supportCounterFilter; // counter filter flag
236 : bool memoryRegister;
237 : uint64_t clientRankNum; // clientRank array's length
238 : uint64_t clientRankAddr; // clientRank array's address, the array is uint32_t[]
239 : uint64_t hcclCommNameAddr; // hcclComm name's addr, the name is char*
240 : int32_t hcclTimeOut; // 0 means default, -1 means never timeout
241 : bool associateWorker; // true means we need to associate worker when load model, default false
242 : char rsv[53UL];
243 : } __attribute__((packed));
244 :
245 : struct __attribute__((visibility("default"))) ModelInfo {
246 : uint32_t modelId;
247 : uint16_t aicpuStreamNum;
248 : ModelStreamInfo* streams;
249 : uint16_t queueNum;
250 : ModelQueueInfo* queues;
251 : int32_t abnormalBreak;
252 : int32_t abnormalEnqueue;
253 : AicpuPriInfo aicpuPriInfo;
254 : uint64_t cfgInfoPtr = 0;
255 : int32_t abnormalEnable = 1;
256 : char rsv[98];
257 : } __attribute__((packed));
258 :
259 : struct __attribute__((visibility("default"))) ReDeployConfig {
260 : uint64_t modelIdsAddr; // ptr which point to modelIds(uint32_t)
261 : uint32_t modelIdNum; // modelIdNum
262 : char rsv[4];
263 : } __attribute__((packed));
264 :
265 : struct __attribute__((visibility("default"))) CheckKernelSupportedConfig {
266 : uint64_t kernelNameAddr; // ptr which point to kernelName
267 : uint32_t kernelNameLen;
268 : uint64_t checkResultAddr; // int32: 0 is supported, others are not supported
269 : uint32_t checkResultLen;
270 : } __attribute__((packed));
271 :
272 : struct __attribute__((visibility("default"))) DataFlowExceptionNotify {
273 : uint64_t transId;
274 : uint32_t type; // 0:Exception occured, 1:Exception expired
275 : uint32_t modelIdNum;
276 : uint64_t modelIdsAddr; // ptr which point to modelIds(uint32_t)
277 : char rsv[40];
278 : } __attribute__((packed));
279 :
280 : struct __attribute__((visibility("default"))) SomaMemMng {
281 : uint64_t size;
282 : uint64_t va; // Virtual address
283 : uint64_t mempoolId; // Id of memory pool
284 : uint32_t deviceId;
285 : int32_t memAsyncOpType; // Malloc(0) Free(1)
286 : int32_t memAsyncSubCMD; // Distinguish between Malloc reuse strategies and Free explicit/implicit reuse methods
287 : char rsv[4];
288 : } __attribute__((packed));
289 : }
290 : #endif // AICPUSD_AICPUSD_INFO_H
|
