TY - GEN
T1 - Access-based localization with bypassing
AU - Oh, Hakjoo
AU - Yi, Kwangkeun
PY - 2011
Y1 - 2011
N2 - We present an extension of access-based localization technique to mitigate a substantial inefficiency in handling procedure calls. Recently, access-based localization was proposed as an effective way of tightly localizing abstract memories. However, it has a limitation in handling procedure calls: the localized input memory for a procedure contains not only memory locations accessed by the procedure but also those accessed by transitively called procedures. The weakness is especially exacerbated in the presence of recursive call cycles, which is common in analysis of realistic programs. In this paper, we present a technique, called bypassing, that mitigates the problem. Our technique localizes input memory states only with memory locations that the procedure directly accesses. Those parts not involved in analysis of the procedure are bypassed to transitively called procedures. In experiments with an industrial-strength global C static analyzer, the technique reduces the average analysis time by 42%. In particular, the technique is especially effective for programs that extensively use recursion: it saves analysis time by 77% on average.
AB - We present an extension of access-based localization technique to mitigate a substantial inefficiency in handling procedure calls. Recently, access-based localization was proposed as an effective way of tightly localizing abstract memories. However, it has a limitation in handling procedure calls: the localized input memory for a procedure contains not only memory locations accessed by the procedure but also those accessed by transitively called procedures. The weakness is especially exacerbated in the presence of recursive call cycles, which is common in analysis of realistic programs. In this paper, we present a technique, called bypassing, that mitigates the problem. Our technique localizes input memory states only with memory locations that the procedure directly accesses. Those parts not involved in analysis of the procedure are bypassed to transitively called procedures. In experiments with an industrial-strength global C static analyzer, the technique reduces the average analysis time by 42%. In particular, the technique is especially effective for programs that extensively use recursion: it saves analysis time by 77% on average.
UR - http://www.scopus.com/inward/record.url?scp=84055193118&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=84055193118&partnerID=8YFLogxK
U2 - 10.1007/978-3-642-25318-8_7
DO - 10.1007/978-3-642-25318-8_7
M3 - Conference contribution
AN - SCOPUS:84055193118
SN - 9783642253171
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 50
EP - 65
BT - Programming Languages and Systems - 9th Asian Symposium, APLAS 2011, Proceedings
T2 - 9th Asian Symposium on Programming Languages and Systems, APLAS 2011
Y2 - 5 December 2011 through 7 December 2011
ER -