tegra_multimedia_api sample memory not release ?

Autonomous Machines Jetson & Embedded Systems Jetson TX2
#1

Hi,

I have seen the trt sample in /home/nvidia/tegra_multimedia_api/samples/common/algorithm/trt/trt_inference.cpp and the code:

void
TRT_Context::buildTrtContext(const string& deployfile,
        const string& modelfile, bool bUseCPUBuf)
{
    if (!parseNet(deployfile))
    {
        cout<<"parse net failed, exit!"<<endl;
        exit(0);
    }
    ifstream trtModelFile("trtModel.cache");
    if (trtModelFile.good())
    {
        // get cache file length
        size_t size = 0;
        size_t i = 0;

        cout<<"Using cached TRT model" <<endl;

        // Get the length
        trtModelFile.seekg(0, ios::end);
        size = trtModelFile.tellg();
        trtModelFile.seekg(0, ios::beg);

        char * buff = new char ;
        while (trtModelFile.get(buff[i])) i++;
        trtModelFile.close();
        runtime = createInferRuntime(*pLogger);
        engine = runtime->deserializeCudaEngine((void *)buff, size, nullptr);
    }
    else
    {
        caffeToTRTModel(deployfile, modelfile);
        cout<<"Create TRT model cache"<<endl;
        ofstream trtModelFile("trtModel.cache");
        trtModelFile.write((char *)trtModelStream->data(), trtModelStream->size());
        trtModelFile.close();
        runtime = createInferRuntime(*pLogger);
        engine = runtime->deserializeCudaEngine(trtModelStream->data(), trtModelStream->size(), nullptr);
        trtModelStream->destroy();
    }
    context = engine->createExecutionContext();
    allocateMemory(bUseCPUBuf);
}

in line 24:

char * buff = new char ;

does this buff can’t be release or you forgot it ?

I have written a demo such as this sample, but I found the demo would leak 16MB per hour.

Thanks.

#2

Hi ClancyLIAN,
It looks to be leak. Do you see 16MB leak disappear after adding ‘delete buff;’?

#3

Hi,

Sorry, this is caused by my program and this buff, but this buff should also free when not used.

Thanks.