Specification and Verification of Atomic Operations in GPGPU Programs

Amighi, Afshin; Darabi, Saeed; Blom, Stefan; Huisman, Marieke

doi:10.1007/978-3-319-22969-0_5

Cited by 11 publications

(11 citation statements)

References 15 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…VerCors takes programs as input that are annotated with logical specifications, and can automatically verify whether the code implementation adheres to these specifications. The specifications are formulated in a Concurrent Separation Logic (CSL) that supports permission accounting, and are annotated as pre/postconditions for functions and threads, and invariants for loops and locks [2,7]. However, to keep the paper accessible, this paper describes the formalization independent of specific knowledge of CSL, and explains any further necessary details whenever needed.…”

Section: The Vercors Verifiermentioning

confidence: 99%

Automated Verification of the Parallel Bellman–Ford Algorithm

Safari

Oortwijn²,

Huisman

2021

Static Analysis

Self Cite

View full text Add to dashboard Cite

Many real-world problems such as internet routing are actually graph problems. To develop efficient solutions to such problems, more and more parallel graph algorithms are proposed. This paper discusses the mechanized verification of a commonly used parallel graph algorithm, namely the Bellman-Ford algorithm, which provides an inherently parallel solution to the Single-Source Shortest Path problem.Concretely, we verify an unoptimized GPU version of the Bellman-Ford algorithm, using the VerCors verifier. The main challenge that we had to address was to find suitable global invariants of the graph-based properties for automated verification. This case study is the first deductive verification to prove functional correctness of the parallel Bellman-Ford algorithm. It provides the basis to verify other, optimized implementations of the algorithm. Moreover, it may also provide a good starting point to verify other parallel graph-based algorithms.

show abstract

Section: The Vercors Verifiermentioning

confidence: 99%

Automated Verification of the Parallel Bellman–Ford Algorithm

Safari

Oortwijn²,

Huisman

2021

Static Analysis

Self Cite

View full text Add to dashboard Cite

show abstract

“…83 The content of this chapter is based on the following publications of the author: "Specification and verification of atomic operations in GPGPU programs" [ADBH15]. 1 To be precise, the GPU execution model is Single Instruction Multiple Thread (SIMT), which extends SIMD with more flexibility in the control flow.…”

Section: Discussionmentioning

confidence: 99%

“…Chapter 6 (Verification of GPGPU Programs): gives a short introduction to GPGPU programming and presents how we reason about the data race freedom and functional correctness of GPGPU kernels that use atomic operations. This chapter is based on the paper "Specification and Verification of Atomic Operations in GPGPU Programs", which was published at SEFM 2015 [ADBH15].…”

mentioning

confidence: 99%

“…However, because they are necessary for understanding the content of this chapter, we discuss the other parts of the specification briefly. For further examples and discussions, we refer to[BHM14] 4. As introduced earlier in Chapter 2, resources are permission formulas in permission-based separation logic specifying the ownership of threads over the memory locations.…”

mentioning

confidence: 99%

See 1 more Smart Citation

Verification of Program Parallelization

Darabi¹

Self Cite

View full text Add to dashboard Cite

Second, we propose a novel technique to reason about deterministic parallel programs. We first formally define the Parallel Programming Language (PPL), a simple core language that captures the main forms of deterministic parallel programs. This language distinguishes three kinds of basic blocks: parallel, vectorized and sequential blocks, which can be composed using three different composition operators: sequential, parallel and fusion composition. We show that it is sufficient to have contracts for the basic blocks to prove the correctness of the PPL program, and moreover that the functional correctness of the sequential program implies the correctness of the parallelized program. We formally prove the correctness of our approach. In addition, we define a widelyused subset of OpenMP that can be encoded into our core language, thus effectively enabling verification of OpenMP compiler directives, and we discuss automated tool support for this verification process. Third, we propose a specification and verification technique to reason about data race freedom and functional correctness of GPGPU kernels that use atomic operations as a synchronization mechanism. We exploit the notion of resource invariant from Concurrent Separation Logic to specify the behaviour of atomic operations. To capture the GPGPU memory model, we adapt this notion of resource invariant such that group resource invariants capture the behaviour of atomic operations that access locations in local memory, which are accessible only to the threads in the same work group, while kernel resource invariants capture the behaviour of atomic operations that access locations in global memory, which are accessible to all threads in all work groups. We show the soundness of our approach and we demonstrate the application of the technique in our toolset. This thesis presents a set of verification techniques based on permissionbased separation logic to reason about the data race freedom and functional correctness of program parallelizations. Our reasoning techniques address different forms of high-level and low-level parallelization. For high-level parallel programs, we first formalize the main features of deterministic parallel programming in PPL and discuss how PPL programs and consequently, real-world deterministic parallel programs (e.g. OpenMP programs) are verified. For low-level parallel programs, we specifically focus on reasoning about GPGPU kernels. At the end we discuss how the presented verification techniques are chained together to reason about the semantical equivalence of high-level parallel programs where they are automatically transformed into low-level

show abstract

“…Note that the barrier always synchronizes threads inside the inner block. We refer to [BHM14,ADBH15] for more details about reasoning of GPU programs with barriers and atomic operations.…”

Section: Cuda To Pvl Transformationmentioning

confidence: 99%

Correct optimized GPU programs

Safari¹

View full text Add to dashboard Cite

It was amazing to work with all of you in the project. I would like to thank all my (ex-)colleagues at the FMT group. I had four years of memorable moments with all of you in the group. I really enjoyed the times we had for social activities such as group lunches, social dinners, FMT outgoings and BOCOM! Rom, it was a pleasure to assist you for Discrete Structures and Efficient Algorithms in module 7. I enjoyed the way you organize it. I am also thankful for taking me to the boxing, even though the pandemic did not allow us to continue it.

show abstract

Specification and Verification of Atomic Operations in GPGPU Programs

Cited by 11 publications

References 15 publications

Automated Verification of the Parallel Bellman–Ford Algorithm

Automated Verification of the Parallel Bellman–Ford Algorithm

Verification of Program Parallelization

Correct optimized GPU programs

Contact Info

Product

Resources

About