[quote][size=2][url=forum.php?mod=redirect&goto=findpost&pid=789473&ptid=111147][color=#999999]回复：cjl68 发表于 2021-3-5 10:37[/color][/url][/size] 您好，该问题非常有价值，推荐您到https://gitee.com/ascend/samples/issues 创建一个issue。我们会帮您安排专人[/quote]已提交 issue：https://gitee.com/ascend/samples/issues/I3A4CS?from=project-issue

好的

还是一直有这个问题，与 vdec 无关吧，我在 venc 之前都保存图片了的，图片没有问题

[quote][size=2][url=forum.php?mod=redirect&goto=findpost&pid=789449&ptid=111147][color=#999999]回复：cjl68 发表于 2021-3-5 10:23[/color][/url][/size] 报错里有关于stream的报错，你可以尝试在线程里显示创建context试下，aclrtCreateContext 会创建一个默认Streamhttp[/quote]我每个线程都创了新的context的，并且用的是显示创的stream

补充：我让 dvpp decode 和 model infer 一起跑且不通信，dvpp decode 解码真实视频流，model infer 循环推理假数据，这样模型推理都会出同样的错，感觉像是这两个东西不能一起跑

我好像发现问题了，我出错是因为一个线程做 dvpp decode，另一个线程做 model infer，两线程间用阻塞队列通信。然而我试了试，先让 dvpp decode 线程跑，把解码后的数据全部往队列里塞，thread.join() 后再启动 model infer 线程，这样就不会出错。请问这是什么问题？

谢谢你的回复，已明白你的意思，我再测测

[quote][size=2][url=forum.php?mod=redirect&goto=findpost&pid=752660&ptid=108244][color=#999999]回复：_xyt 发表于 2021-2-23 19:10[/color][/url][/size] 麻烦上传下完整日志内容，取/var/log/npu/slog/host-0下的日志[/quote][ERROR] GE(6101,cold_river):2021-02-23-08:00:30.963.520 [framework/domi/executor/../graph/load/graph_loader.cc:285]6115 ExecuteModel: ErrorNo: 117702659() Execute model failed, model_id:7. [ERROR] ASCENDCL(6101,cold_river):2021-02-23-08:00:30.963.535 [acl/model/model.cpp:635]6115 ModelExecute: "Execute model failed, ge result[117702659], modelId[7]" [ERROR] ASCENDCL(6101,cold_river):2021-02-23-08:00:30.963.552 [acl/model/model.cpp:667]6115 aclmdlExecuteAsync: "aclmdlExecuteAsync failed, result[500000], modelId[7]" [ERROR] DVPP(6101,cold_river):2021-02-23-08:00:30.966.113 [dvpp/vcodec/omx_component/vdec_hivna/core/../normal/omx_vdec.cpp:2389][VDEC] [event_process:2389] [T89] OMX_ErrorSyntaxOneFrameFail ERROR, chanId = 1, frame index = 154 [ERROR] DVPP(6101,cold_river):2021-02-23-08:00:30.966.162 [dvpp/vcodec/omx_component/vdec_hivna/core/../normal/omx_adapt.cpp:487][VDEC] [EventHandlerErr:487] [T89] chanId = 1, frame index = 154 decode failed, error type = 0xfffe90000006 [ERROR] DVPP(6101,cold_river):2021-02-23-08:00:30.966.180 [dvpp/vcodec/omx_component/vdec_hivna/core/../normal/omx_adapt.cpp:317][VDEC] [ExtendError:317] [T89] report decode one frame failed to user, channelId = 1, errtype = 0x3fff000090000006. [ERROR] DVPP(6101,cold_river):2021-02-23-08:00:30.987.368 [dvpp/vcodec/omx_component/vdec_hivna/core/../normal/omx_vdec.cpp:2389][VDEC] [event_process:2389] [T89] OMX_ErrorSyntaxOneFrameFail ERROR, chanId = 0, frame index = 134 [ERROR] DVPP(6101,cold_river):2021-02-23-08:00:30.987.427 [dvpp/vcodec/omx_component/vdec_hivna/core/../normal/omx_adapt.cpp:487][VDEC] [EventHandlerErr:487] [T89] chanId = 0, frame index = 134 decode failed, error type = 0xfffe90000006 [ERROR] DVPP(6101,cold_river):2021-02-23-08:00:30.987.446 [dvpp/vcodec/omx_component/vdec_hivna/core/../normal/omx_adapt.cpp:317][VDEC] [ExtendError:317] [T89] report decode one frame failed to user, channelId = 0, errtype = 0x3fff000090000006. [ERROR] RUNTIME(6101,cold_river):2021-02-23-08:00:30.995.355 [runtime/feature/src/api_impl.cc:1549]6115 ModelExecute:model is not in current ctx [ERROR] RUNTIME(6101,cold_river):2021-02-23-08:00:30.995.423 [runtime/feature/src/logger.cc:991]6115 ModelExecute:ret=117702659, model not in current ctx [ERROR] GE(6101,cold_river):2021-02-23-08:00:30.995.443 [framework/domi/executor/../graph/load/new_model_manager/davinci_model.cc:3464]6115 NnExecute: ErrorNo: -1(failed) Call rt api failed, ret: 0x7040003 [ERROR] GE(6101,cold_river):2021-02-23-08:00:30.995.461 [framework/domi/executor/../graph/load/graph_loader.cc:285]6115 ExecuteModel: ErrorNo: 117702659() Execute model failed, model_id:7. [ERROR] ASCENDCL(6101,cold_river):2021-02-23-08:00:30.995.477 [acl/model/model.cpp:635]6115 ModelExecute: "Execute model failed, ge result[117702659], modelId[7]" [ERROR] ASCENDCL(6101,cold_river):2021-02-23-08:00:30.995.495 [acl/model/model.cpp:667]6115 aclmdlExecuteAsync: "aclmdlExecuteAsync failed, result[500000], modelId[7]" [ERROR] DVPP(6101,cold_river):2021-02-23-08:00:30.998.045 [dvpp/vcodec/omx_component/vdec_hivna/core/../normal/omx_vdec.cpp:2389][VDEC] [event_process:2389] [T89] OMX_ErrorSyntaxOneFrameFail ERROR, chanId = 1, frame index = 155 [ERROR] DVPP(6101,cold_river):2021-02-23-08:00:30.998.097 [dvpp/vcodec/omx_component/vdec_hivna/core/../normal/omx_adapt.cpp:487][VDEC] [EventHandlerErr:487] [T89] chanId = 1, frame index = 155 decode failed, error type = 0xfffe90000006 [ERROR] DVPP(6101,cold_river):2021-02-23-08:00:30.998.115 [dvpp/vcodec/omx_component/vdec_hivna/core/../normal/omx_adapt.cpp:317][VDEC] [ExtendError:317] [T89] report decode one frame failed to user, channelId = 1, errtype = 0x3fff000090000006. [ERROR] DVPP(6101,cold_river):2021-02-23-08:00:31.023.912 [dvpp/vcodec/omx_component/vdec_hivna/core/../normal/omx_vdec.cpp:2389][VDEC] [event_process:2389] [T89] OMX_ErrorSyntaxOneFrameFail ERROR, chanId = 0, frame index = 135 [ERROR] DVPP(6101,cold_river):2021-02-23-08:00:31.023.971 [dvpp/vcodec/omx_component/vdec_hivna/core/../normal/omx_adapt.cpp:487][VDEC] [EventHandlerErr:487] [T89] chanId = 0, frame index = 135 decode failed, error type = 0xfffe90000006 [ERROR] DVPP(6101,cold_river):2021-02-23-08:00:31.023.989 [dvpp/vcodec/omx_component/vdec_hivna/core/../normal/omx_adapt.cpp:317][VDEC] [ExtendError:317] [T89] report decode one frame failed to user, channelId = 0, errtype = 0x3fff000090000006. [ERROR] RUNTIME(6101,cold_river):2021-02-23-08:00:31.027.136 [runtime/feature/src/api_impl.cc:1549]6115 ModelExecute:model is not in current ctx [ERROR] RUNTIME(6101,cold_river):2021-02-23-08:00:31.027.196 [runtime/feature/src/logger.cc:991]6115 ModelExecute:ret=117702659, model not in current ctx [ERROR] GE(6101,cold_river):2021-02-23-08:00:31.027.215 [framework/domi/executor/../graph/load/new_model_manager/davinci_model.cc:3464]6115 NnExecute: ErrorNo: -1(failed) Call rt api failed, ret: 0x7040003 [ERROR] GE(6101,cold_river):2021-02-23-08:00:31.027.233 [framework/domi/executor/../graph/load/graph_loader.cc:285]6115 ExecuteModel: ErrorNo: 117702659() Execute model failed, model_id:7. [ERROR] ASCENDCL(6101,cold_river):2021-02-23-08:00:31.027.249 [acl/model/model.cpp:635]6115 ModelExecute: "Execute model failed, ge result[117702659], modelId[7]" [ERROR] ASCENDCL(6101,cold_river):2021-02-23-08:00:31.027.267 [acl/model/model.cpp:667]6115 aclmdlExecuteAsync: "aclmdlExecuteAsync failed, result[500000], modelId[7]" [ERROR] DVPP(6101,cold_river):2021-02-23-08:00:31.029.659 [dvpp/vcodec/omx_component/vdec_hivna/core/../normal/omx_vdec.cpp:2389][VDEC] [event_process:2389] [T89] OMX_ErrorSyntaxOneFrameFail ERROR, chanId = 1, frame index = 156 [ERROR] DVPP(6101,cold_river):2021-02-23-08:00:31.029.708 [dvpp/vcodec/omx_component/vdec_hivna/core/../normal/omx_adapt.cpp:487][VDEC] [EventHandlerErr:487] [T89] chanId = 1, frame index = 156 decode failed, error type = 0xfffe90000006 [ERROR] DVPP(6101,cold_river):2021-02-23-08:00:31.029.726 [dvpp/vcodec/omx_component/vdec_hivna/core/../normal/omx_adapt.cpp:317][VDEC] [ExtendError:317] [T89] report decode one frame failed to user, channelId = 1, errtype = 0x3fff000090000006. [ERROR] RUNTIME(6101,cold_river):2021-02-23-08:00:31.057.459 [runtime/feature/src/api_impl.cc:1549]6115 ModelExecute:model is not in current ctx [ERROR] RUNTIME(6101,cold_river):2021-02-23-08:00:31.057.528 [runtime/feature/src/logger.cc:991]6115 ModelExecute:ret=117702659, model not in current ctx [ERROR] GE(6101,cold_river):2021-02-23-08:00:31.057.549 [framework/domi/executor/../graph/load/new_model_manager/davinci_model.cc:3464]6115 NnExecute: ErrorNo: -1(failed) Call rt api failed, ret: 0x7040003 [ERROR] GE(6101,cold_river):2021-02-23-08:00:31.057.569 [framework/domi/executor/../graph/load/graph_loader.cc:285]6115 ExecuteModel: ErrorNo: 117702659() Execute model failed, model_id:7. [ERROR] ASCENDCL(6101,cold_river):2021-02-23-08:00:31.057.584 [acl/model/model.cpp:635]6115 ModelExecute: "Execute model failed, ge result[117702659], modelId[7]" [ERROR] ASCENDCL(6101,cold_river):2021-02-23-08:00:31.057.602 [acl/model/model.cpp:667]6115 aclmdlExecuteAsync: "aclmdlExecuteAsync failed, result[500000], modelId[7]"

[quote][size=2][url=forum.php?mod=redirect&goto=findpost&pid=751507&ptid=108244][color=#999999]回复：_xyt 发表于 2021-2-23 14:24[/color][/url][/size] 这个是不是表示，即使我把 aclmdlExecuteAsync 的调用放到互斥锁的临界区内也是会出问题的？是的。可以把aclmdlExecuteAsy[/quote]我试了下推理加锁，多线程调这个推理，会出现 aclError:500000 错误。 pthread_mutex_lock(&mutex);  CHECK_ACL(aclmdlExecuteAsync(model_id, input_dataset, output_dataset, stream));  CHECK_ACL(aclrtSynchronizeStream(stream));  pthread_mutex_unlock(&mutex);

[quote][size=2][url=forum.php?mod=redirect&goto=findpost&pid=751507&ptid=108244][color=#999999]回复：_xyt 发表于 2021-2-23 14:24[/color][/url][/size] 这个是不是表示，即使我把 aclmdlExecuteAsync 的调用放到互斥锁的临界区内也是会出问题的？是的。可以把aclmdlExecuteAsy[/quote]把 aclmdlExecuteAsync 和 aclrtSynchronizeStream 都放到互斥锁的临界区内会不划算，仅仅是因为串行了吗，如果仅这个原因的话，在我的应用场景中还是适用的。