MEMORY LATENCY TOLERANCE SEVERAL SIMPLE TECHNIQUES CAN MAKE RUNAHEAD EXECUTION MORE EFFICIENT BY REDUCING THE NUMBER OF INSTRUCTIONS EXECUTED AND THEREBY REDUCING THE ADDITIONAL ENERGY CONSUMPTION TYPICALLY

Today’s high-performance processors face main-memory latencies on the order of hundreds of processor clock cycles. As a result, even the most aggressive processors spend a significant portion of their execution time stalling and waiting for main-memory accesses to return data to the execution core. Previous research has shown that runahead execution significantly increases a high-performance processor’s ability to tolerate long main-memory latencies. 1, 2 Runahead execution improves a processor’s performance by speculatively preexecuting the application program while the processor services a long-latency (L2) data cache miss, instead of stalling the processor fo