Auto-tuning a high-level language targeted to GPU codes

By S. Grauer-Gray, Lifan Xu, Robert Searles, Sudhee Ayalasomayajula, John Cavazos. 2012 Innovative Parallel Computing (InPar), 2012

522 citations recorded by the source corpus.

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

Bones (2014, methodology): “Additionally, PolyBench/GPU [Grauer-Gray et al. 2012] is included, which provides handwritten nonoptimized CUDA code (hand-optimized is not available).”
End-to-End Deep Learning of Optimization Heuristics (2017, unknown): “2 6 8 PolyBench [33] 1.”
Neural Code Comprehension: A Learnable Representation of Code Semantics (2018, unknown): “Lines Size Pairs Machine Learning Tensorflow [1] 2,492 16,943,893 220,554 260,250,973 High-Performance Computing AMD APP SDK [9] 123 1,304,669 4,146 45,081,359 BLAS [22] 300 280,782 566 283,856 Benchmarks NAS [57] 268 572,521 1,793 1,701,968 Parboil [59] 151 118,575 2,175 151,916 PolybenchGPU [27] 4”
Accel-Sim: An Extensible Simulation Framework for Validated GPU Modeling (2018, methodology): “With the exception of a handful of applications with trace-size issues, we run all the applications found in Rodinia 3.1 [9], the CUDA SDK [40], Parboil [65], Polybench [16], CUTLASS [47], our microbenchmark suite, and cutting-edge machine learning workloads from the Deepbench benchmark suite [35].”