Benchmarks that benefit from compaction

Ecosystem
1

Do you have benchmarks that benefit from compaction?

There is some ongoing work to improve compaction in the presence of fragmentation. We have some closed source workloads where this benefits programs significantly. We also have some open-source microbenchmarks where we can exhibit this improvement. But we don’t seem to have many open-source benchmarks that benefit from compaction. If you have any, please do share.

2

What do you mean a benchmark that benefits of compaction? There is the locality (when the code is faster because of no cache miss), and the allocation which is faster when all you have to do is to allocate at the extremum of the heap because the heap is compact or encline to be compacted later.

I have made a benchmark inspired by Computer Language Benchmark Game (same problems, but analogous implementation as far as it is possible with the language)

See Language efficiency comparison where binarytrees stresses the standard allocation of multiple implementations.

3

What do you mean a benchmark that benefits of compaction?

@Frederic_Loyer Do you know if binarytrees or other benchmarks in the suite do automatic compaction? Or are there benchmarks in the suite that have a call to Gc.compact() , and without it the programs run slower. That’s what I mean.