bgenc.net/content/publications.md

3.8 KiB

no-ttr
true
## Crafty: Efficient, HTM-Compatible Persistent Transactions
Kaan Genç, Michael D. Bond, and Guoqing Harry Xu
ACM SIGPLAN Conference on Programming Language Design and Implementation (PLDI 2020), Online, June 2020

Crafty is a library for transactional storage, built for new non-volatile memory hardware. Taking advantage of hardware transactional capabilities of modern CPUs, it provides a low-overhead option that also eliminates the need for additional concurrency control.

Talk Paper Extended Paper Implementation [Poster](/extra/Crafty Poster.pdf)

## Dependence Aware, Unbounded Sound Predictive Race Detection
Kaan Genç, Jake Roemer, Yufan Xu, and Michael D. Bond
ACM SIGPLAN International Conference on Object-Oriented Programming, Systems, Languages, and Applications (OOPSLA 2019), Athens, Greece, October 2019

This paper presents 2 data race detection analyses which analyze a single run of a program to predict data races that can happen in other runs. These analyses take advantage of data and control flow dependence to accurately understand how the analyzed program works, expanding what races can be predicted.

Talk Extended Paper (updated version) Paper Corrigendum to paper Implementation [Poster](/extra/DepAware Poster.pdf)

## SmartTrack: Efficient Predictive Race Detection
Jake Roemer, Kaan Genç, and Michael D. Bond
ACM SIGPLAN Conference on Programming Language Design and Implementation (PLDI 2020), Online, June 2020

Predictive data race detection methods greatly improve the number of data races found, but they typically significantly slow down programs compared to their non-predictive counterparts. SmartTrack, through improved analyses and clever algorithms, reduces their overhead to just around non-predictive analyses without impacting their performance.

Paper Extended Paper

## High-Coverage, Unbounded Sound Predictive Race Detection
Jake Roemer, Kaan Genç, and Michael D. Bond
ACM SIGPLAN Conference on Programming Language Design and Implementation (PLDI 2018), Philadelphia, PA, USA, June 2018

Predictive data race detection methods typically walk a tight line between predicting more races and avoiding false races. This paper presents a new analysis that can predict more races, and a method to efficiently eliminate false races.

Paper Extended Paper

Activities

PLDI 2021 Artifact Evaluation Committee member

ASPLOS 2021 Artifact Evaluation Committee member

OOPSLA 2020 Artifact Evaluation Committee member