Abstract: Optimizing instruction scheduling and memory allocation for tensor and graphical processors using lattice image data structure optimizations is provided. A method of using a loop fusion by a compiler in interconnected accelerator units to simplify a machine learning (ML) graph representing a program to be compiled is provided. The method includes (A) lowering a plurality of operations in an initial first program to an original plurality of at least two loops. The method also includes (B) inferring a fused loop structure from the original plurality of at least two loops in the initial first program thus creating a second program having the fused loop. The fused loop in the second program is pipelining the multiplication and addition operations thus significantly reducing the memory bandwidth requirements and improving cache locality.

Abstract: A received query is analyzed to provide a set of potential bushy tree operation plans that include the creation of one or more “sub-joins” that provide intermediate resultant “sub-tables”, and using these sub-tables as operands in the other predicates of the query. The received query is analyzed to identify “satellite” tables and “seed” tables. A satellite table is a table of the database that is filtered by a selective predicate, and is connected to only one other table in the database via at least one linking predicate, and a seed table is a table of the database that is connected to a satellite table and at least one other table. Multiple candidate operation plans are formulated as bushy trees that include a sub-join of each seed table and its connected satellite table. A least cost operation plan is selected for execution from among these candidate operation plans and the conventional least cost linear operation plan based on the costs associated with each operation plan.

Abstract: A received query is analyzed to provide a set of potential bushy tree operation plans that include the creation of one or more “sub-joins” that provide intermediate resultant “sub-tables”, and using these sub-tables as operands in the other predicates of the query. The received query is analyzed to identify “satellite” tables and “seed” tables. A satellite table is a table of the database that is filtered by a selective predicate, and is connected to only one other table in the database via at least one linking predicate, and a seed table is a table of the database that is connected to a satellite table and at least one other table. Multiple candidate operation plans are formulated as bushy trees that include a sub-join of each seed table and its connected satellite table. A least cost operation plan is selected for execution from among these candidate operation plans and the conventional least cost linear operation plan based on the costs associated with each operation plan.