Fine-grained software-directed data prefetching using integrated high-level and low-level code analysis optimizations

A computer program product for minimizing effective memory latency, the computer program product comprising:

• a non-transitory computer recordable-type media having computer executable instructions stored thereon, the computer executable instructions comprising:
• computer executable instructions for analyzing a portion of code that contains one or more loops to form a loop table;
• computer executable instructions for identifying at least one candidate loop within the one or more loops of the loop table for prefetch efficiency optimization;
• and
• computer executable instructions for inserting prefetch control instructions and directives to optimize the at least one candidate loop, wherein the prefetch control instructions and directives are inserted at selected locations in an instruction stream, in advance of data use, according to a stream type, and wherein a stream is a sequence of addresses which depend upon an induction variable of the at least one candidate loop;
• computer executable instructions for determining whether pipelining is available;
• computer executable instructions responsive to a determination that pipelining is not available, for adjusting prefetch addresses in the prefetch control instructions forward to fetch cache lines in advance of data use;
• computer executable instructions for gathering memory references of the at least one loop;
• and
• computer executable instructions for performing at least one of data dependency analysis and reuse analysis on the at least one candidate loop to identify unique streams within the at least one candidate loop.