i have a project to implement in llvm i would be grateful if you guide me about the idea and its implementation.
basically it is intended to do just in time compilation of large amount of data (data intensive) for the openmp target
like we have a simple program of adding 2 numbers, and the input comes from a text file containing like 20 million rows (file size in gbs). each row has 2 numbers that are required to be added.
this file is required to be compiled at runtime along with the bytecode emitted by add program. also some adaptive optimizations are required to be done during compilation. and the final executable should emit openmp code to be executed by openmp programming model, it should contain multiple sections such that 1000 numbers addition is done by 1 thread (1 core), thousand by other thread (2nd core) and so on… obviously memory is less than 20gb so there should be such optimization efficient replacement which achieves this task by less memory.
i have to implement it through LLVM. how is it possible?
looking forward to your response