Will binding of textures endure Memcpy

Accelerated Computing CUDA CUDA Programming and Performance
1

Hi @all!
I’ve got an array of texture references of variable length. So my idea is to allocate this array in host code, set filtermode, adressmode etc. and bind it to cudaArrays that reside in device memory. Then I would like to allocate an array of textureReferences in device memory and do a Memcpy of the texture references from host to device.
Will the tex refs in the device mem have all the desired properties after Memcpy. Or in other words: Are bonding information stored in the textureReference struct?

Kind regards

Navier

2

Does your array of texture references even work? I’m 99.999% positive that this is not supported by CUDA. Each texture reference must be a globally declared variable, and has implicit static storage within the compilation unit.