Performing OPC on hardware or software platforms with GPU
Optical proximity correction techniques performed on one or more graphics processors improve the masks used for the printing of microelectronic circuit designs. Execution of OPC techniques on hardware or software platforms utilizing graphics processing units. GPUs may share the computation load with the system CPUs to efficiently and effectively execute the OPC method steps.
1. A method comprising:
providing a system comprising a plurality of computing nodes, wherein each computing node comprises at least one processor selected from the group consisting of central processing unit and graphics processing unit;
using at least one processor, separating an optical proximity correction process into tasks depending on a type of computation required;
allocating the tasks of the optical proximity correction process to the central processing unit or the graphics processing unit; and
delivering output of the central processing unit and the graphics processing unit as a result of the optical proximity corrections (OPC) process,
wherein the separating an optical proximity correction process into tasks depending on a type of computation required comprises:
analyzing a mask pattern to be written to a target material; and
splitting the mask pattern into a plurality of two-dimensional subregions, wherein allocating the tasks of the optical proximity correction process to the central processing unit or the graphics processing unit comprises:
assigning each subregion to one or more computing nodes in the plurality of computing nodes, wherein OPC results are output for each subregion; and
assembling the OPC results for the plurality of two-dimensional subregions;
wherein the mask pattern is used to manufacture semiconductor devices.
2. The method of claim 1 wherein adjacent subregions overlap, and wherein assembling the OPC results comprises removing output from overlapping subregions.
3. The method of claim 1 wherein assembling the OPC results comprises transferring output results for a subregion from the subregion's assigned computing node to a designated computing node in the plurality of computing nodes.
4. The method of claim 3 wherein the designated computing node runs a master program.
5. The method of claim 1 wherein each subregion is assigned to one computing node.
6. The method of claim 1 wherein while processing a given subregion, a computing node has available mask pattern data from overlapping regions, where an overlapping region comprises portions of a subregion adjacent to the given subregion.
7. The method of claim 6 wherein processing the given subregion comprises outputting results only for the given subregion, but not for the overlapping regions.
8. The method of claim 1 wherein the computing nodes to which subregions have been assigned do not share information.