Precise calling context encoding

Abstract

Calling contexts are very important for a wide range of applications such as profiling, debugging, and event logging. Most applications perform expensive stack walking to recover contexts. The resulting contexts are often explicitly represented as a sequence of call sites and hence bulky. We propose a technique to encode the current calling context of any point during an execution. In particular, an acyclic call path is encoded into one number through only integer additions. Recursive call paths are divided into acyclic subsequences and encoded independently. We leverage stack depth in a safe way to optimize encoding: if a calling context can be safely and uniquely identified by its stack depth, we do not perform encoding. We propose an algorithm to seamlessly fuse encoding and stack depth based identification. The algorithm is safe because different contexts are guaranteed to have different IDs. It also ensures contexts can be faithfully decoded. Our experiments show that our technique incurs negligible overhead (1.89% on average). For most medium-sized programs, it can encode all contexts with just one number. For large programs, we are able to encode most calling contexts to a few numbers.

Valence: variable length calling context encoding

“…For baseline comparison to our approach, we implement and employ the current state-of-the-art technique for precise calling context encoding (PCCE) [37].…”

“…Consider the example in Figure 1, which is taken from [37]. The boxes next to each node show their numCC(n) values, while the edge labels show the actual instrumentation for each call site.…”

“…Recursive context decoding is straightforward: the algorithm simply processes one acyclic sub-path of the recursive context at a time, until all values on the stack are fully decoded. Additional and more specific details about the decoding process are available in [37].…”

“…Consider again the example call graph from [37], now shown in Figure 2. In this figure, the boxed numbers show the level of each node, while the edge labels show the Valence instrumentation, including the range of bits within id and the bit codes that are used at each call site.…”

“…Some other encoding strategies sacrifice compactness for accurate and reliable decoding. Sumner et al adapted the Ball-Larus path profiling technique [5] to implement precise calling context encoding (PCCE) [37]. Their approach employs static call graph analysis to construct instrumentation that always generates unique encodings for every possible calling context.…”