Abstract
Approximate program transformations such as skipping tasks [29, 30], loop perforation [21, 22, 35], reduction sampling [38], multiple selectable implementations [3, 4, 16, 38], dynamic knobs [16], synchronization elimination [20, 32], approximate function memoization [11], and approximate data types [34] produce programs that can execute at a variety of points in an underlying performance versus accuracy tradeoff space. These transformed programs have the ability to trade accuracy of their results for increased performance by dynamically and nondeterministically modifying variables that control their execution. We call such transformed programs relaxed programs because they have been extended with additional nondeterminism to relax their semantics and enable greater flexibility in their execution. We present language constructs for developing and specifying relaxed programs. We also present proof rules for reasoning about acceptability properties [28], which the program must satisfy to be acceptable. Our proof rules work with two kinds of acceptability properties: relational acceptability properties, which characterize desired relationships between the values of variables in the original and relaxed programs, and unary acceptability properties, which involve values only from a single (original or relaxed) program. The proof rules support a staged reasoning approach in which the majority of the reasoning effort works with the original program. Exploiting the common structure that the original and relaxed programs share, relational reasoning transfers reasoning effort from the original program to prove properties of the relaxed program. We have formalized the dynamic semantics of our target programming language and the proof rules in Coq and verified that the proof rules are sound with respect to the dynamic semantics. Our Coq implementation enables developers to obtain fully machinechecked verifications of their relaxed programs.
| Original language | English (US) |
|---|---|
| Title of host publication | PLDI'12 - Proceedings of the 2012 ACM SIGPLAN Conference on Programming Language Design and Implementation |
| Pages | 169-180 |
| Number of pages | 12 |
| DOIs | |
| State | Published - Jul 9 2012 |
| Externally published | Yes |
| Event | 33rd ACM SIGPLAN Conference on Programming Language Design and Implementation, PLDI'12 - Beijing, China Duration: Jun 11 2012 → Jun 16 2012 |
Publication series
| Name | Proceedings of the ACM SIGPLAN Conference on Programming Language Design and Implementation (PLDI) |
|---|
Other
| Other | 33rd ACM SIGPLAN Conference on Programming Language Design and Implementation, PLDI'12 |
|---|---|
| Country | China |
| City | Beijing |
| Period | 6/11/12 → 6/16/12 |
Fingerprint
Keywords
- Acceptability
- Coq
- Relational hoare logic
- Relaxed programs
ASJC Scopus subject areas
- Software
Cite this
Proving acceptability properties of relaxed nondeterministic approximate programs. / Carbin, Michael; Kim, Deokhwan; Misailovic, Sasa; Rinard, Martin C.
PLDI'12 - Proceedings of the 2012 ACM SIGPLAN Conference on Programming Language Design and Implementation. 2012. p. 169-180 (Proceedings of the ACM SIGPLAN Conference on Programming Language Design and Implementation (PLDI)).Research output: Chapter in Book/Report/Conference proceeding › Conference contribution
}
TY - GEN
T1 - Proving acceptability properties of relaxed nondeterministic approximate programs
AU - Carbin, Michael
AU - Kim, Deokhwan
AU - Misailovic, Sasa
AU - Rinard, Martin C.
PY - 2012/7/9
Y1 - 2012/7/9
N2 - Approximate program transformations such as skipping tasks [29, 30], loop perforation [21, 22, 35], reduction sampling [38], multiple selectable implementations [3, 4, 16, 38], dynamic knobs [16], synchronization elimination [20, 32], approximate function memoization [11], and approximate data types [34] produce programs that can execute at a variety of points in an underlying performance versus accuracy tradeoff space. These transformed programs have the ability to trade accuracy of their results for increased performance by dynamically and nondeterministically modifying variables that control their execution. We call such transformed programs relaxed programs because they have been extended with additional nondeterminism to relax their semantics and enable greater flexibility in their execution. We present language constructs for developing and specifying relaxed programs. We also present proof rules for reasoning about acceptability properties [28], which the program must satisfy to be acceptable. Our proof rules work with two kinds of acceptability properties: relational acceptability properties, which characterize desired relationships between the values of variables in the original and relaxed programs, and unary acceptability properties, which involve values only from a single (original or relaxed) program. The proof rules support a staged reasoning approach in which the majority of the reasoning effort works with the original program. Exploiting the common structure that the original and relaxed programs share, relational reasoning transfers reasoning effort from the original program to prove properties of the relaxed program. We have formalized the dynamic semantics of our target programming language and the proof rules in Coq and verified that the proof rules are sound with respect to the dynamic semantics. Our Coq implementation enables developers to obtain fully machinechecked verifications of their relaxed programs.
AB - Approximate program transformations such as skipping tasks [29, 30], loop perforation [21, 22, 35], reduction sampling [38], multiple selectable implementations [3, 4, 16, 38], dynamic knobs [16], synchronization elimination [20, 32], approximate function memoization [11], and approximate data types [34] produce programs that can execute at a variety of points in an underlying performance versus accuracy tradeoff space. These transformed programs have the ability to trade accuracy of their results for increased performance by dynamically and nondeterministically modifying variables that control their execution. We call such transformed programs relaxed programs because they have been extended with additional nondeterminism to relax their semantics and enable greater flexibility in their execution. We present language constructs for developing and specifying relaxed programs. We also present proof rules for reasoning about acceptability properties [28], which the program must satisfy to be acceptable. Our proof rules work with two kinds of acceptability properties: relational acceptability properties, which characterize desired relationships between the values of variables in the original and relaxed programs, and unary acceptability properties, which involve values only from a single (original or relaxed) program. The proof rules support a staged reasoning approach in which the majority of the reasoning effort works with the original program. Exploiting the common structure that the original and relaxed programs share, relational reasoning transfers reasoning effort from the original program to prove properties of the relaxed program. We have formalized the dynamic semantics of our target programming language and the proof rules in Coq and verified that the proof rules are sound with respect to the dynamic semantics. Our Coq implementation enables developers to obtain fully machinechecked verifications of their relaxed programs.
KW - Acceptability
KW - Coq
KW - Relational hoare logic
KW - Relaxed programs
UR - http://www.scopus.com/inward/record.url?scp=84863438138&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=84863438138&partnerID=8YFLogxK
U2 - 10.1145/2254064.2254086
DO - 10.1145/2254064.2254086
M3 - Conference contribution
AN - SCOPUS:84863438138
SN - 9781450312059
T3 - Proceedings of the ACM SIGPLAN Conference on Programming Language Design and Implementation (PLDI)
SP - 169
EP - 180
BT - PLDI'12 - Proceedings of the 2012 ACM SIGPLAN Conference on Programming Language Design and Implementation
ER -