I want to directly transfer video frames from a framegrabber to GPU memory. The suggested methods are
1)Using GPUDirect RDMA by referring GitHub - NVIDIA/jetson-rdma-picoevb: Minimal HW-based demo of GPUDirect RDMA on NVIDIA Jetson AGX Xavier running L4T and
2)Another method is making use of below APIs
glGenBuffers, glBindBuffer, glBufferData,glMapbuffer and pass the address to kernel in User_ptr method.
I want to difference between these two methods
GPUDirect requires PCIe hardware but EGL doesn’t.
If without GPUDirect and PCIe , i can able access GPU memory using EGL means then what’s the advantage of going to GPUDirect ? why nvidia releases such feature like GPUDirect?
GPUDirect is used for two device transfer buffers through PCIe.
But you cannot transfer a video frame to another Jetson with the EGL approach.
Can i make use of a GPUDirect RDMA in an X86 system where a nvidia’s GPU inserted in one PCIe slot and framegrabber in another slot?
Do you mean you want to transfer buffer between two dGPU cards?
It should work: GPUDirect | NVIDIA Developer
For more information, please file a new topic to the corresponding category.
This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.