[adwn]

If anything, GPU architectures are even less suited to immutability than CPUs. To get good performance out of a GPU (otherwise, why use them?), you have to avoid thread divergence caused by data-dependent branches as much as possible (I'm simplifying a bit). The simplest and most effective way to do this is to allocate a couple of large arrays and work on them. Immutable data structures, garbage collection, and frequent allocation of small objects, in contrast, will cause a lot of thread divergence.

In addition, immutable data structures tend to be pointer-based. For GPUs (even more so than for CPUs), you want contiguous memory accesses for acceptable performance.