LLMCodeEval: An Execution-Based Multilingual Multitask Multidimensional Benchmark for Evaluating Large Language Models on Code Understanding and Generation - View it on GitHub