message

 

 

 

 

 float* output1;   
 output1 = (float* )malloc(18*4*sizeof(float));     
 int* output2;     
 output2 = (int* )malloc(18 * 4 * sizeof(int));     
 CHECK(cudaMemcpyAsync(buffers[0], InputDatas[0].data(), mInputParams.BatchSize * mInputParams.ImgC * mInputParams.ImgH * mInputParams.ImgW * sizeof(float), cudaMemcpyHostToDevice, stream));       
 mContext->enqueue(mInputParams.BatchSize, buffers,stream, nullptr);      
 CHECK(cudaMemcpyAsync(output1, buffers[1], 1 * 18 * 4*sizeof(float), cudaMemcpyDeviceToHost, stream));     
 CHECK(cudaMemcpyAsync(output2, buffers[2], 1 * 18 * 4*sizeof(int), cudaMemcpyDeviceToHost, stream));