LLM-Aided Compilation for Tensor Accelerators