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 : #include "acl_rt_impl.h"
12 : #include "runtime/rts/rts_kernel.h"
13 : #include "runtime/kernel.h"
14 : #include "runtime/base.h"
15 : #include "runtime/rts/rts_stream.h"
16 : #include "runtime/rts/rts_device.h"
17 : #include "common/log_inner.h"
18 : #include "common/error_codes_inner.h"
19 : #include "common/prof_reporter.h"
20 : #include "utils/data_type_utils.h"
21 :
22 : namespace {
23 : constexpr uint32_t ACL_ERROR_INVALID_EXCEPTION_INFO = 0xFFFFFFFFU;
24 : constexpr char ACLRT_EXCEPTION_CALLBACK_REG_PREFIX[] = "__ACLRT_EXCEPTION_CALLBACK_REGISTER__:";
25 :
26 7 : std::string MakeExceptionCallbackRegName(rtTaskFailCallback callback)
27 : {
28 14 : return std::string(ACLRT_EXCEPTION_CALLBACK_REG_PREFIX) + acl::ACL_MODULE_NAME + ":" +
29 21 : std::to_string(static_cast<uint64_t>(reinterpret_cast<uintptr_t>(callback)));
30 : }
31 : } // namespace
32 :
33 : #ifdef __cplusplus
34 : extern "C" {
35 : #endif
36 :
37 4 : aclError aclrtSubscribeReportImpl(uint64_t threadId, aclrtStream stream)
38 : {
39 4 : ACL_LOG_INFO("start to execute aclrtSubscribeReport, threadId is %lu.", threadId);
40 4 : ACL_REQUIRES_RTS_OK(rtSubscribeReport(threadId, static_cast<rtStream_t>(stream)));
41 2 : ACL_LOG_INFO("successfully execute aclrtSubscribeReport, threadId is %lu.", threadId);
42 2 : return ACL_SUCCESS;
43 : }
44 :
45 3 : aclError aclrtSetExceptionInfoCallbackImpl(aclrtExceptionInfoCallback callback)
46 : {
47 3 : ACL_LOG_INFO("start to execute aclrtSetExceptionInfoCallback.");
48 3 : ACL_REQUIRES_RTS_OK(rtRegTaskFailCallbackByModule(acl::ACL_MODULE_NAME, static_cast<rtTaskFailCallback>(callback)));
49 1 : ACL_LOG_INFO("successfully execute aclrtSetExceptionInfoCallback");
50 1 : return ACL_SUCCESS;
51 : }
52 :
53 6 : aclError aclrtExceptionInfoCallbackRegisterImpl(aclrtExceptionInfoCallback callback)
54 : {
55 6 : ACL_PROFILING_REG(acl::AclProfType::AclrtExceptionInfoCallbackRegister);
56 6 : ACL_LOG_INFO("start to execute aclrtExceptionInfoCallbackRegister.");
57 10 : ACL_REQUIRES_NOT_NULL_RET_INPUT_REPORT(callback, ACL_ERROR_RT_PARAM_INVALID);
58 5 : const auto rtCallback = static_cast<rtTaskFailCallback>(callback);
59 5 : const std::string regName = MakeExceptionCallbackRegName(rtCallback);
60 5 : ACL_REQUIRES_RTS_OK(rtRegTaskFailCallbackByModule(regName.c_str(), rtCallback));
61 4 : ACL_LOG_INFO("successfully execute aclrtExceptionInfoCallbackRegister");
62 4 : return ACL_SUCCESS;
63 6 : }
64 :
65 3 : aclError aclrtExceptionInfoCallbackUnregisterImpl(aclrtExceptionInfoCallback callback)
66 : {
67 3 : ACL_PROFILING_REG(acl::AclProfType::AclrtExceptionInfoCallbackUnregister);
68 3 : ACL_LOG_INFO("start to execute aclrtExceptionInfoCallbackUnregister.");
69 7 : ACL_REQUIRES_NOT_NULL_RET_INPUT_REPORT(callback, ACL_ERROR_RT_PARAM_INVALID);
70 2 : const std::string regName = MakeExceptionCallbackRegName(static_cast<rtTaskFailCallback>(callback));
71 2 : ACL_REQUIRES_RTS_OK(rtRegTaskFailCallbackByModule(regName.c_str(), nullptr));
72 1 : ACL_LOG_INFO("successfully execute aclrtExceptionInfoCallbackUnregister");
73 1 : return ACL_SUCCESS;
74 3 : }
75 :
76 1 : uint32_t aclrtGetTaskIdFromExceptionInfoImpl(const aclrtExceptionInfo* info)
77 : {
78 5 : ACL_REQUIRES_NOT_NULL_RET_INPUT_REPORT(info, static_cast<aclError>(ACL_ERROR_INVALID_EXCEPTION_INFO));
79 0 : return info->taskid;
80 : }
81 :
82 1 : uint32_t aclrtGetStreamIdFromExceptionInfoImpl(const aclrtExceptionInfo* info)
83 : {
84 5 : ACL_REQUIRES_NOT_NULL_RET_INPUT_REPORT(info, static_cast<aclError>(ACL_ERROR_INVALID_EXCEPTION_INFO));
85 0 : return info->streamid;
86 : }
87 :
88 1 : uint32_t aclrtGetThreadIdFromExceptionInfoImpl(const aclrtExceptionInfo* info)
89 : {
90 5 : ACL_REQUIRES_NOT_NULL_RET_INPUT_REPORT(info, static_cast<aclError>(ACL_ERROR_INVALID_EXCEPTION_INFO));
91 0 : return info->tid;
92 : }
93 :
94 2 : uint32_t aclrtGetDeviceIdFromExceptionInfoImpl(const aclrtExceptionInfo* info)
95 : {
96 6 : ACL_REQUIRES_NOT_NULL_RET_INPUT_REPORT(info, static_cast<aclError>(ACL_ERROR_INVALID_EXCEPTION_INFO));
97 1 : return info->deviceid;
98 : }
99 :
100 2 : uint32_t aclrtGetErrorCodeFromExceptionInfoImpl(const aclrtExceptionInfo* info)
101 : {
102 6 : ACL_REQUIRES_NOT_NULL_RET_INPUT_REPORT(info, static_cast<aclError>(ACL_ERROR_INVALID_EXCEPTION_INFO));
103 1 : return info->retcode;
104 : }
105 :
106 8 : aclError aclrtGetArgsFromExceptionInfoImpl(const aclrtExceptionInfo* info, void** devArgsPtr, uint32_t* devArgsLen)
107 : {
108 12 : ACL_REQUIRES_NOT_NULL_RET_INPUT_REPORT(info, static_cast<aclError>(ACL_ERROR_INVALID_EXCEPTION_INFO));
109 7 : ACL_REQUIRES_NOT_NULL_WITH_INPUT_REPORT(devArgsPtr);
110 7 : ACL_REQUIRES_NOT_NULL_WITH_INPUT_REPORT(devArgsLen);
111 :
112 7 : if (info->expandInfo.type == RT_EXCEPTION_AICORE) {
113 1 : *devArgsPtr = info->expandInfo.u.aicoreInfo.exceptionArgs.argAddr;
114 1 : *devArgsLen = info->expandInfo.u.aicoreInfo.exceptionArgs.argsize;
115 6 : } else if (info->expandInfo.type == RT_EXCEPTION_AICPU) {
116 1 : *devArgsPtr = info->expandInfo.u.aicpuInfo.argAddr;
117 1 : *devArgsLen = info->expandInfo.u.aicpuInfo.argsize;
118 5 : } else if (
119 5 : info->expandInfo.type == RT_EXCEPTION_FUSION && info->expandInfo.u.fusionInfo.type == RT_FUSION_AICORE_CCU) {
120 1 : *devArgsPtr = info->expandInfo.u.fusionInfo.u.aicoreCcuInfo.exceptionArgs.argAddr;
121 1 : *devArgsLen = info->expandInfo.u.fusionInfo.u.aicoreCcuInfo.exceptionArgs.argsize;
122 : } else {
123 4 : ACL_LOG_ERROR("exception information type = %d is invalid, get args failed.", info->expandInfo.type);
124 4 : std::string funcName = acl::AclErrorLogManager::GetFuncNameWithoutImplSuffix(__func__);
125 4 : acl::AclErrorLogManager::ReportInputError(
126 8 : acl::INVALID_VALUE_MSG, std::vector<const char*>({"func", "value", "param", "expect"}),
127 4 : std::vector<const char*>(
128 4 : {funcName.c_str(), acl::GetExceptionExpandTypeDesc(info->expandInfo.type), "info->expandInfo.type",
129 8 : "RT_EXCEPTION_AICORE, RT_EXCEPTION_AICPU or RT_EXCEPTION_FUSION"}));
130 4 : return ACL_ERROR_INVALID_EXCEPTION_INFO;
131 4 : }
132 :
133 3 : return ACL_SUCCESS;
134 : }
135 :
136 3 : aclError aclrtGetFuncHandleFromExceptionInfoImpl(const aclrtExceptionInfo* info, aclrtFuncHandle* func)
137 : {
138 7 : ACL_REQUIRES_NOT_NULL_RET_INPUT_REPORT(info, static_cast<aclError>(ACL_ERROR_INVALID_EXCEPTION_INFO));
139 2 : ACL_REQUIRES_NOT_NULL_WITH_INPUT_REPORT(func);
140 2 : ACL_REQUIRES_RTS_OK(rtGetFuncHandleFromExceptionInfo(info, func));
141 :
142 1 : return ACL_SUCCESS;
143 : }
144 :
145 3 : aclError aclrtBinarySetExceptionCallbackImpl(
146 : aclrtBinHandle binHandle, aclrtOpExceptionCallback callback, void* userData)
147 : {
148 3 : ACL_LOG_INFO("start to execute aclrtBinarySetExceptionCallback.");
149 3 : ACL_REQUIRES_NOT_NULL_WITH_INPUT_REPORT(binHandle);
150 2 : ACL_REQUIRES_NOT_NULL_WITH_INPUT_REPORT(callback);
151 :
152 2 : ACL_REQUIRES_RTS_OK(rtBinarySetExceptionCallback(binHandle, callback, userData));
153 :
154 1 : return ACL_SUCCESS;
155 : }
156 :
157 6 : aclError aclrtLaunchCallbackImpl(aclrtCallback fn, void* userData, aclrtCallbackBlockType blockType, aclrtStream stream)
158 : {
159 6 : ACL_PROFILING_REG(acl::AclProfType::AclrtLaunchCallback);
160 6 : ACL_LOG_INFO("start to execute aclrtLaunchCallback.");
161 12 : ACL_CHECK_INVALID_VALUE_WITH_DESC(
162 : (blockType == ACL_CALLBACK_BLOCK || blockType == ACL_CALLBACK_NO_BLOCK),
163 : acl::GetCallbackBlockTypeDesc(blockType), "blockType", "ACL_CALLBACK_BLOCK or ACL_CALLBACK_NO_BLOCK",
164 : ACL_ERROR_INVALID_PARAM);
165 4 : const bool isBlock = (blockType == ACL_CALLBACK_BLOCK);
166 4 : ACL_REQUIRES_RTS_OK(
167 : rtCallbackLaunch(static_cast<rtCallback_t>(fn), userData, static_cast<rtStream_t>(stream), isBlock));
168 2 : ACL_LOG_INFO("successfully execute aclrtLaunchCallback");
169 2 : return ACL_SUCCESS;
170 6 : }
171 :
172 2 : aclError aclrtLaunchHostFuncImpl(aclrtStream stream, aclrtHostFunc fn, void* args)
173 : {
174 2 : ACL_PROFILING_REG(acl::AclProfType::AclrtLaunchHostFunc);
175 2 : ACL_LOG_INFO("start to execute aclrtLaunchHostFunc.");
176 2 : ACL_REQUIRES_RTS_OK(rtsLaunchHostFunc(static_cast<rtStream_t>(stream), static_cast<rtCallback_t>(fn), args));
177 1 : ACL_LOG_INFO("successfully execute aclrtLaunchHostFunc");
178 1 : return ACL_SUCCESS;
179 2 : }
180 :
181 10 : aclError aclrtProcessReportImpl(int32_t timeout)
182 : {
183 10 : ACL_PROFILING_REG(acl::AclProfType::AclrtProcessReport);
184 10 : ACL_LOG_INFO("start to execute aclrtProcessReport, timeout is %dms.", timeout);
185 : // -1 represents infinite wait, timeout value greater than 0 represents waiting for a fixed time.
186 : // other value is invalid.
187 16 : ACL_CHECK_INVALID_PARAM_WITH_REASON_RET(
188 : (timeout < -1 || timeout == 0), timeout,
189 : "-1 represents infinite wait, timeout value greater than 0 represents waiting for a fixed time",
190 : ACL_ERROR_INVALID_PARAM);
191 8 : const rtError_t rtErr = rtProcessReport(timeout);
192 8 : if (rtErr != RT_ERROR_NONE) {
193 6 : if (rtErr == ACL_ERROR_RT_THREAD_SUBSCRIBE) {
194 2 : ACL_LOG_INFO("no subscribereport info, runtime errorCode = %d", static_cast<int32_t>(rtErr));
195 4 : } else if (rtErr == ACL_ERROR_RT_REPORT_TIMEOUT) {
196 2 : ACL_LOG_INFO("wait subscribereport timeout, runtime errorCode = %d", static_cast<int32_t>(rtErr));
197 : }
198 6 : return ACL_GET_ERRCODE_RTS(rtErr);
199 : }
200 2 : ACL_LOG_INFO("successfully aclrtProcessReport, timeout is %dms.", timeout);
201 2 : return ACL_SUCCESS;
202 10 : }
203 :
204 4 : aclError aclrtUnSubscribeReportImpl(uint64_t threadId, aclrtStream stream)
205 : {
206 4 : ACL_LOG_INFO("start to execute aclrtUnSubscribeReport, threadId is %lu.", threadId);
207 4 : ACL_REQUIRES_RTS_OK(rtUnSubscribeReport(threadId, static_cast<rtStream_t>(stream)));
208 2 : ACL_LOG_INFO("successfully execute aclrtUnSubscribeReport, threadId is %lu.", threadId);
209 2 : return ACL_SUCCESS;
210 : }
211 :
212 4 : aclError aclrtRegStreamStateCallbackImpl(const char* regName, aclrtStreamStateCallback callback, void* args)
213 : {
214 4 : ACL_PROFILING_REG(acl::AclProfType::AclrtRegStreamStateCallback);
215 4 : ACL_LOG_INFO("start to execute aclrtRegStreamStateCallback");
216 4 : ACL_REQUIRES_NOT_NULL_WITH_INPUT_REPORT(regName);
217 3 : ACL_REQUIRES_RTS_OK(rtsRegStreamStateCallback(regName, reinterpret_cast<rtsStreamStateCallback>(callback), args));
218 2 : ACL_LOG_INFO("successfully execute aclrtRegStreamStateCallback");
219 2 : return ACL_SUCCESS;
220 4 : }
221 :
222 4 : aclError aclrtRegDeviceStateCallbackImpl(const char* regName, aclrtDeviceStateCallback callback, void* args)
223 : {
224 4 : ACL_PROFILING_REG(acl::AclProfType::AclrtRegDeviceStateCallback);
225 4 : ACL_LOG_INFO("start to execute aclrtRegDeviceStateCallback");
226 4 : ACL_REQUIRES_NOT_NULL_WITH_INPUT_REPORT(regName);
227 :
228 3 : ACL_REQUIRES_RTS_OK(rtsRegDeviceStateCallback(regName, reinterpret_cast<rtsDeviceStateCallback>(callback), args));
229 2 : ACL_LOG_INFO("successfully execute aclrtRegDeviceStateCallback");
230 2 : return ACL_SUCCESS;
231 4 : }
232 :
233 4 : aclError aclrtSetDeviceTaskAbortCallbackImpl(const char* regName, aclrtDeviceTaskAbortCallback callback, void* args)
234 : {
235 4 : ACL_PROFILING_REG(acl::AclProfType::AclrtSetDeviceTaskAbortCallback);
236 4 : ACL_LOG_INFO("start to execute aclrtSetDeviceTaskAbortCallback");
237 4 : ACL_REQUIRES_NOT_NULL_WITH_INPUT_REPORT(regName);
238 :
239 3 : ACL_REQUIRES_RTS_OK(
240 : rtsSetDeviceTaskAbortCallback(regName, reinterpret_cast<rtsDeviceTaskAbortCallback>(callback), args));
241 2 : ACL_LOG_INFO("successfully execute aclrtSetDeviceTaskAbortCallback");
242 2 : return ACL_SUCCESS;
243 4 : }
244 : #ifdef __cplusplus
245 : }
246 : #endif
|