Memory Efficient Parallel SAT Solving with Inprocessing

Iser, Markus; Balyo, Tomáš; Sinz, Carsten

doi:10.1109/ictai.2019.00018

Cited by 5 publications

(5 citation statements)

References 26 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In the parallel track, a single 32-core node was employed for up to 5000 s per instance while the cloud track was evaluated on 100 8-core nodes for up to 1000 s per instance. These different modes of operation require different solver architectures: For modest parallelism in shared memory, high concurrency and memory consumption can become a considerable issue [20]. On a larger scale, concurrency can be less of an issue while good diversification and communication efficiency becomes critical.…”

Section: Related Workmentioning

confidence: 99%

See 1 more Smart Citation

Scalable SAT Solving in the Cloud

Schreiber

Sanders

2021

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Previous efforts on making Satisfiability (SAT) solving fit for high performance computing (HPC) have led to super-linear speedups on particular formulae, but for most inputs cannot make efficient use of a large number of processors. Moreover, long latencies (minutes to days) of job scheduling make large-scale SAT solving on demand impractical for most applications. We address both issues with Mallob, a framework for job scheduling in the context of SAT solving which exploits malleability, i.e., the ability to add or remove processing power from a job during its computation. Mallob includes a massively parallel, distributed, and malleable SAT solving engine based on HordeSat with a more succinct and communication-efficient approach to clause sharing and numerous further improvements over its precursor. Experiments with up to 2560 cores show that Mallob outperforms an improved version of HordeSat and scales significantly better. Moreover, Mallob can solve many formulae in parallel while dynamically adapting the assigned resources, and jobs arriving in the system are usually initiated within a fraction of a second.

show abstract

Section: Related Workmentioning

confidence: 99%

“…Memory Usage. The memory consumption of parallel SAT solvers is a known issue [20]: As each solver commonly maintains its own clause database, memory requirements increase proportionally with the number of spawned solvers. As such, large formulae can cause out-of-memory errors.…”

Section: Performance Improvementsmentioning

confidence: 99%

Scalable SAT Solving in the Cloud

Schreiber

Sanders

2021

Lecture Notes in Computer Science

View full text Add to dashboard Cite

show abstract

“…The parallel track was evaluated on a single node with 32 cores and a time limit of 5000s per instance while the cloud track was evaluated on 100 nodes with 8 cores each and a considerably lower time limit of 1000s per instance. These different modes of operation have far-reaching consequences on the design of solvers: For modest levels of (shared-memory) parallelism, high memory consumption and high concurrency can become a considerable issue [19]. In a large-scale cloud environment, shared-memory concurrency can be less of an issue while good diversification and communication efficiency becomes critical.…”

Section: Related Workmentioning

confidence: 99%

“…Memory Usage. The memory consumption of parallel SAT solvers is a known issue [19]: As each solver commonly maintains its own clause database, the memory requirements increase proportionally with the number of spawned solvers. As such, large formulae can cause Out-Of-Memory errors.…”

Section: Practical Improvementsmentioning

confidence: 99%

Scalable SAT Solving in the Cloud

Schreiber¹,

Sanders²

2022

Preprint

View full text Add to dashboard Cite

Previous efforts on making Satisfiability (SAT) solving fit for high performance computing (HPC) have lead to super-linear speedups on particular formulae, but for most inputs cannot make efficient use of a large number of processors. Moreover, long latencies (minutes to days) of job scheduling make large-scale SAT solving on demand impractical for most applications. We address both issues with Mallob, a framework for job scheduling in the context of SAT solving which exploits malleability, i.e., the ability to add or remove processing power from a job during its computation. Mallob includes a massively parallel, distributed, and malleable SAT solving engine based on Hordesat with a more succinct and communication-efficient approach to clause sharing and numerous further improvements over its precursor. For example, Mallob on 640 cores outperforms an updated and improved configuration of Hordesat on 2560 cores. Moreover, Mallob can also solve many formulae in parallel while dynamically adapting the assigned resources, and jobs arriving in the system are usually initiated within a fraction of a second.

show abstract

“…Works on targeted algorithm engineering for SAT-solvers are extensive. Just to name a few examples, there is work on exploiting features such as optimizing memory footprints for the architecture [10], on implementing cache-aware [13], on using huge pages [22], on how to benefit from parallel solving [35] or employing inprocessing. Inprocessing particularly takes advantage of modern hardware as one can execute much more instructions on a modern CPU than accessing bytes on memory [31,51].…”

Section: Related Workmentioning

confidence: 99%