Static load classification for improving the value predictability of data-cache misses

Abstract

While caches are effective at avoiding most main-memory accesses, the few remaining memory references are still expensive. Even one cache miss per one hundred accesses can double a program's execution time. To better tolerate the data-cache miss latency, architects have proposed various speculation mechanisms, including load-value prediction. A load-value predictor guesses the result of a load so that the dependent instructions can immediately proceed without having to wait for the memory access to complete.To use the prediction resources most effectively, speculation should be restricted to loads that are likely to miss in the cache and that are likely to be predicted correctly. Prior work has considered hardware-and profile-based methods to make these decisions. Our work focuses on making these decisions at compile time. We show that a simple compiler classification is effective at separating the loads that should be speculated from the loads that should not. We present results for a number of C and Java programs and demonstrate that our results are consistent across programming languages and across program inputs.

Identifying the sources of cache misses in Java programs without relying on hardware counters

“…Burtscher et al [5] classified load instructions based on the region of memory (stack, heap, or global), the kind of reference (array, field, or scalar) and the type of data (pointer or value). They showed that load instructions for certain classes caused more cache misses than others in C and Java programs.…”

Static identification of delinquent loads

“…Burtscher, Diwan, and Hauswirth [3] (which we shall call the 'BDH-method' in the rest of the paper) proposed a static classification system for improving value prediction. In the BDH-method, load instructions are classified according to the following criteria:…”

“…The heuristic that was suggested [3] is that loads belonging to the union of the classes GAN, HSN, HFN, HAN, HFP and HAP account for most of the misses in a program. It should be noted that their study was performed on traces.…”

“…This gives an indication of the damage done if say prefetching is applied to loads wrongly identified as delinquent. Table 12 reports the values of and using the heuristics proposed by Ozawa, Kimura and Nishizaki [10] (the 'OKN-method') and the BDH-method [3]. The same binary and cache configurations were used in all the tests.…”

“…Note that in original paper [3] this classification was done over the execution trace. The simulator was instru-mented to produce type information for loads.…”