texture ref bound to linear memory texture ref. binding limit

Accelerated Computing CUDA CUDA Programming and Performance
1

CUDA spec says that for a texture reference bound to linear memory, the maximum width is 2^27.
Does this maximum width mean that each texture reference can refer to a memory region whose size is up to 2&27, but it does not matter how many (memory region, texture reference) mappings exist? or does this width mean the total memory size that can be bound to texture references?

2

It is the width that can be bound to a particular texture reference, not an aggregate total.

There is an independent limit on the number of texture references, I believe, but it is very high.