Function pt_cuda_allocator_compatibility

pub fn pt_cuda_allocator_compatibility() -> bool

Check if PyTorch CUDA caching allocator has expandable segments enabled.

This function calls the C++ implementation that directly accesses the PyTorch C10 CUDA allocator configuration to check if expandable segments are enabled, which is required for RDMA operations with CUDA tensors.

Returns

true if both CUDA caching allocator is enabled AND expandable segments are enabled, false otherwise.