The detail algorithm implemention is illustrated as below:
General Similarity contains most of modules in Sparse Similarity and Dense Similarity. It have two DataLoader which can process sparse and dense input by config. It is shown in the aboved API picture, each PE have four AXI port to store input data. For sparse mode, each PE have 3 valid AXI input which are corresponding to offset, indice and weight, so that there is a dangling port. In dense mode, the partitioned weight data are stored in each AXI and it can better improve the data loading speed which can significantly impact the finnal performance. After DataLoader, the input data is transform to COO stream internally so that it can share most of calculation logic between dense and sparse mode. The overall diagram of general similarity kernel have a insert sort module which return the top K number of similarity values. The maximum number of K is a template number which can be changed by rebuilding the xclbin. The default value of top K is 32.