Gklee

Li, Guodong; Sawaya, Geof; Gopalakrishnan, Ganesh; Ghosh, Indradeep; Sunitha, R

doi:10.1145/2145816.2145844

Cited by 89 publications

(14 citation statements)

References 15 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…That is, we symbolically execute one thread from one barrier to the other, then switch to the other thread and do the same, etc. The fact that this single sequential schedule [3] is sufficient is argued in [1], [15], [4], [17]. Both GKLEEp and SESA carry this sequential schedule using one symbolic thread.…”

Section: B Parametric Flows and Race Checkingmentioning

confidence: 99%

“…Concolic ("concrete plus symbolic") execution tools such as GKLEE [17], [18] have also recently been proposed, and have helped find bugs in real-world GPU kernels. As the work in this paper is based on concolic execution, we make more detailed comparisons against these tools.…”

Section: Introductionmentioning

confidence: 99%

“…Unfortunately, GKLEEp still suffers from search explosion: even if only two symbolic threads are considered, each thread may create a large number of flows or symbolic states. For instance, SC14, November [16][17][18][19][20][21]2014, New Orleans, USA 978-1-4799-5500-8/14/$31.00 c 2014 IEEE if a thread contains n feasible branches, then O(2 n ) flows or paths may be generated. As shown later, this scenario is not uncommon in realistic CUDA programs.…”

Section: Introductionmentioning

confidence: 99%

“…SESA inherits several features from its predecessors [17], [18]. In particular, it has the ability to do race-checking under standard warp-sizes, or a warp-size of 1.…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Practical Symbolic Race Checking of GPU Programs

Gopalakrishnan

2014

SC14: International Conference for High Performance Computing, Networking, Storage and Analysis

Self Cite

View full text Add to dashboard Cite

Abstract-Even the careful GPU programmer can inadvertently introduce data races while writing and optimizing code. Currently available GPU race checking methods fall short either in terms of their formal guarantees, ease of use, or practicality. Existing symbolic methods: (1) do not fully support existing CUDA kernels; (2) may require user-specified assertions or invariants; (3) often require users to guess which inputs may be safely made concrete; (4) tend to explode in complexity when the number of threads is increased; and (5) explode in the face of thread-ID based decisions, especially in a loop. We present SESA, a new tool combining Symbolic Execution and Static Analysis to analyze C++ CUDA programs that overcomes all these limitations. SESA also scales well to handle non-trivial benchmarks such as Parboil and Lonestar, and is the only tool of its class that handles such practical examples. This paper presents SESA's methodological innovations and practical results.

show abstract

Section: B Parametric Flows and Race Checkingmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

“…SESA inherits several features from its predecessors [17], [18]. In particular, it has the ability to do race-checking under standard warp-sizes, or a warp-size of 1.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Practical Symbolic Race Checking of GPU Programs

Gopalakrishnan

2014

SC14: International Conference for High Performance Computing, Networking, Storage and Analysis

Self Cite

View full text Add to dashboard Cite

show abstract

“…There are also three more typical requirements: (1) strings are converted into numeric values for back-end computations; (2) string values are constrained through regular expressions; and (3) unsatisfiable hybrid constraints should be identified quickly. This poses unique challenges to many symbolic execution tools [5,12,13,17] which usually handle only numeric constraints well.…”

Section: Introductionmentioning

confidence: 99%

PASS: String Solving with Parameterized Array and Interval Automaton

Ghosh

2013

Lecture Notes in Computer Science

Self Cite

View full text Add to dashboard Cite

Abstract. The problem of solving string constraints together with numeric constraints has received increasing interest recently. Existing methods use either bit-vectors or automata (or their combination) to model strings, and reduce string constraints to bit-vector constraints or automaton operations, which are then solved in the respective domain. Unfortunately, they often fail to achieve a good balance between efficiency, accuracy, and comprehensiveness. In this paper we illustrate a new technique that uses parameterized arrays as the main data structure to model strings, and converts string constraints into quantified expressions that are solved through quantifier elimination. We present an efficient and sound quantifier elimination algorithm. In addition, we use an automaton model to handle regular expressions and reason about string values faster. Our method does not need to enumerate string lengths (as bit-vector based methods do), or concrete string values (as automaton based methods do). Hence, it can achieve much better accuracy and efficiency. In particular, it can identify unsatisfiable cases quickly. Our solver (named PASS) supports most of the popular string operations, including string comparisons, string-numeric conversions, and regular expressions. Experimental results demonstrate the advantages of our method.

show abstract

Structural testing for CUDA programming model

Luz,

Souza,

Souza

2024

Concurrency and Computation

View full text Add to dashboard Cite

SummaryGraphic processors offer an accessible solution for high‐performance computing, addressing challenges across various fields. The Compute Unified Device Architecture (CUDA) programming model has emerged to enhance the performance of general‐purpose applications on graphic processors. However, developing CUDA programs is far from straightforward, and developers' lack of experience in parallel programming has led to numerous issues. This article presents a structural testing model and criteria to improve the quality of CUDA programs. These criteria facilitate the selection of test cases and aid in identifying faults. The ValiCUDA tool was developed to implement and validate this testing model and criteria. This tool instruments and analyzes programs, generating the necessary elements for each testing criterion. It also facilitates program execution and evaluation of criterion coverage. A statistical validation experiment assessed these criteria' effectiveness, cost, and strength metrics. The results demonstrate that the criteria can identify nontrivial faults in CUDA programs and assist testers in their testing endeavors for such applications.

show abstract

Gklee

Cited by 89 publications

References 15 publications

Practical Symbolic Race Checking of GPU Programs

Practical Symbolic Race Checking of GPU Programs

PASS: String Solving with Parameterized Array and Interval Automaton

Structural testing for CUDA programming model

Contact Info

Product

Resources

About