This repository evaluates the proposed Shark sytem using the Koala benchmarks for the shell. Most of the transformations outlined in the original paper has been manually implemented in each of the benchmarks. All transformations (and their size in loc) are outlined as follows:
| Benchmark | LoC | Notes |
|---|---|---|
aurpkg |
4 | run loop in parallel, remove cat |
bio |
8 | identify parallelizable region, export as command, run outer loop in parallel with while, remove cat |
covid-mts |
4 | remove cat |
file-enc |
4 | par/ize with & and wait |
log-analysis |
4 | par/ize with & and wait |
makeself |
||
max-temp |
4 | par/ize with &, wait, use tee |
media-conv |
4 | par/ize with &, wait, use tee |
nlp |
3–5 | par/ize with &, wait, remove cat (~90 loc total) |
oneliners |
13 | remove cat, & |
riker |
56 | par/ize with & and wait |
sklearn |
10 | par/ize independent runs |
unix50 |
36 | remove cat per script |
vps-audit |
65 | use {} & and wait |
web-index |
4 | use tee and pipelines to remove cat |
The top-level main.sh script is a quick script for downloading dependencies and inputs, running, profiling, and verifying a single Koala benchmark.
./main.sh <BENCHMARK_NAME> [OPTIONS] [<args passed to execute.sh>]For more information and context on the benchmarks used, please visit the Koala repository.
The Koala Benchmarks are licensed under the MIT License. See the LICENSE file for more information.