How Well Do LLMs Generate Code for Different Application Domains? Benchmark and Evaluation