Questions about runtime

[addingIce]

Hi, I ran test with Intel(R) Xeon(R) Gold 5218 CPU @ 2.30GHz and 64GB of RAM, the sizes of dataset are both 10000, and intersection size is 50. It took about 41 seconds for client to get the result, excluding the key generation time. But in the paper it said that for 10000 inputsize only cost 7.47 seconds. Is this performance gap as expected?

[addingIce]

And most of the time was cost by paillier-enc