Optimal coding is an information store and transmission issue. Large volumes of information lead to the need to compress it. Among various algorithms, Huffman coding has been the most popular choice due to its lack of redundancy. There is a known two stage sequential algorithm for Huffman codes calculating. This paper describes a developed algorithm for generating Huffman codes. The algorithm does not require generating codes in two stages, but instead generates codes in a single stage. The codes are generated in parallel. The algorithm was experimentally studied by simulating it in the C programming language and using natural language texts for testing. The computational experiment was carried out using a cluster consisting of four nodes. Test results demonstrate the adequacy and effectiveness of the proposed algorithm. We report our comparative study results. Our comparison is performed on two types of data sets: English texts and Russian ones. The developed algorithm can be used in systems for processing symbolic information, for implementing algorithms for efficient encoding and compression of symbolic data.