SysPilot: Cross-Library Performance Autotuning via Library Performance Contracts
Organizations: University of Texas at Austin Austin, TX, USA · Microsoft USA
Abstract
Modern scientific and machine-learning applications compose many independently developed libraries. When these libraries run together, their independently managed runtimes can oversubscribe shared resources: using the same CPU cores, fighting over GPUs, or contending for memory bandwidth. Avoiding such contention requires careful configuration across library boundaries. Existing autotuners do not directly investigate what resources are shared between libraries, which runtime signals correlate with interference. As a result, they are inefficient for finding optimal configurations to mitigate contention. This paper presents SysPilot, a cross-library performance autotuning system based on Library Performance Contract (LPC). A Library Performance Contract captures a library's resource demands, assumptions, controls, and observability signals in LPCL, a specification language for describing library resource requirements. SysPilot synthesizes contracts from documentation and code artifacts, composes contracts from the libraries used by an application, derives resource-conflict hypotheses, profiles the application to validate those hypotheses, and selects candidate configurations. This design makes the Library Performance Contract a reusable unit of performance knowledge: a contract for a library can be reused across many applications that depend on that library. We evaluate SysPilot on a benchmark suite of cross-library contention scenarios. The results show that SysPilot matches the performance of autotuners that rely on application-specific manual information while retaining cross-application reuse. Compared with general-purpose autotuners such as OpenTuner and SMAC3, SysPilot improves benchmark performance metrics by up to an additional 38% over the best configuration found by these autotuners, while using 3x fewer runs.
Figures & tables
| ::= | contract | |
| ::= | ||
| ::= | component : evidence= ; | |
| ::= | use by= ; resource= ; ; evidence= ; | |
| ::= | assume resource= ; ; evidence= ; | |
| ::= | control target= ; effect= ; evidence= ; | |
| ::= | signal target= ; probe= ; evidence= ; |
| Resource | Meaning |
| cpu.compute | CPU execution capacity consumed by runnable workers, thread pools, and task executors. |
| cpu.placement | Affinity and core-placement constraints over CPU sets and NUMA domains. |
| gpu.compute | GPU compute units consumed by kernels. |
| gpu.memory | GPU memory capacity consumed by buffers. |
| gpu.bandwidth | GPU memory traffic throughput from reads and writes. |
| mem.capacity | Host memory capacity consumed by buffers. |
| Tuner | Trials to target | Avg. score | ||
| runs | runs | |||
| SysPilot (App Manual-Guided) | 6 | 12 | 0.896 | 0.935 |
| SysPilot (Lib Contract-Guided) | 8 | 12 | 0.879 | 0.960 |
| SMAC3 | 24 | 0.611 | 0.824 | |
| OpenTuner | 0.625 | 0.729 | ||
| Variant | Trials to oracle |
| Full SysPilot | 7 |
| Without probes/profiling | 10 |
| Raw documentation instead of LPCL | 28 |
| Knob-only autotuner | 30+ |
Appendix figures & tables1 asset
Supplementary material from the paper’s appendix.
Appendix
| Benchmark | Libraries and datasets | Shared resources | Configuration space | Best performance gain |
| Joblib/BLAS Nested CPU Contention | Joblib, NumPy, OpenBLAS ( Harris et al., 2020 ; Zhang et al., 2012 ) | cpu.compute | ||
| Dask/DuckDB Columnar Memory-Spill | Dask, PyArrow, DuckDB ( Rocklin and others, 2015 ; Raasveldt and Mühleisen, 2019 ) | cpu.compute , mem.capacity , io.bandwidth | ||
| HF/PyTorch Text Data Pipeline | HF Datasets, HF Tokenizers, PyTorch DataLoader ( Lhoest et al., 2021 ; PyTorch Contributors, 2026 ) | cpu.compute , mem.capacity | ||
| TorchData/S3-DDP Object-Store Pipeline | TorchData, fsspec/s3fs, MinIO, PyTorch DDP ( Paszke et al., 2017 ) | cpu.compute , mem.capacity , io.bandwidth , io.queue | ||
| DALI/PyTorch GPU Decode-Inference | DALI/nvJPEG, PyTorch/cuDNN | gpu.compute , gpu.memory , io.queue | ||
| Forest-Cover Model Selection | Dask-ML, XGBoost, UCI Covertype ( Rocklin and others, 2015 ; Chen and Guestrin, 2016 ; Blackard and Dean, 1998 ) | cpu.compute , mem.capacity |
High-Performance Computing
High-performance computing (HPC) uses parallel systems and optimized software to solve computationally demanding scientific and engineering problems.
See all 152 papers