ECP Proxy and Polybench/C Source-Optimized IRs

This dataset includes source code and LLVM IRs for the following Polybench/C and Exascale Computing Project Proxy Applications:Polybench/C: 3MM, Heat3d, Syr2k(*)ECP PA: AMG, RSBench, SW4Lite (GPU-accelerated) Each LLVM IR is compiled using Clang v13.0.0 with experimental compiler extensions provided by SOLLVE's Polyhedral Loop Optimizers.The source code of each IR corresponds to a particular input size (when expressed at compile time, as in Polybench/C applications) and tuned source code configuration (selected from an expert-defined optimization space expressed solely at the source-code level). The mapping between LLVM IR filenames and expressed optimizations are tracked per-benchmark in the CSVs in the collations dataset.