errors while executing fifoCUDA sample

Accelerated Computing CUDA CUDA Setup and Installation
1

Hello,
I have just installed NVIDIA GPU Direct for Video SDK and built fifoCUDA sample without any modifications

Now, stepping with debugger over function Fifo::initFifo:

First, while executing dvpInitCUDAContext it prints
DVP_ERROR: Initialization failed
Fail on line 871
DVP Failed with status FFFFFFFF

Next, while executing dvpGetRequiredConstantsGLCtx:
DVP_ERROR: Initialization failed
Fail on line 877
DVP Failed with status FFFFFFFF

Probably it is not worth doing following steps until these will be solved but I am still interested in the following code:

uint32_t bufferStride = m_width*numChannels*numBytesPerChannel; // here bufferStride becomes 7680 as expected
bufferStride += g_bufferGPUStrideAlignment-1;  // 7679
bufferStride &= ~(g_bufferGPUStrideAlignment-1); // 0

Given g_bufferGPUStrideAlignment is 0, bufferStride also becomes zero at this point. This makes sysMemBuffersDesc.size 0 and folowing VirtualAlloc() fails.

I will appropriate any tips that could help me to pass through this initial sample code

2

Hi Michael,

I have the same problem when I try to run the fifoCUDA example. Could you share how to fix it if you known the solution.

my develop environment: Win7 64bit, Nvidia TESLA K10, CUDA 10.1, GPUDirect for Video SDK 1.70, MSVC 2012.

Regards,
Snaku

3

Hi Michael,

I tried to build the fifoCUDA sample code with CUDA 6.0, and it works, so I think new version CUDA runtime doesn’t support GPUDirect for Video library. (I also tried CUDA 8.0, the result is fail).

Regards,
Snaku

4

Hi snaku,
I will buy a Quadro Rtx A4000 which is seemed only supported by CUDA 11.0 or higher. If new version CUDA runtime doesn’t support GPUDirect for Video library, does CUDA 6.0 work for A4000?