1# Dynamic Linker Relocation Benchmark 2 3This benchmark measures the time spent in the dynamic linker to load a large set 4of shared objects with many symbols and relocations. It mimics the work involved 5in loading a fixed version of `libandroid_servers.so`. 6 7## Running the benchmark 8 9To run the benchmark, build the `linker-reloc-bench` target, sync `data`, and 10run the benchmark from `/data/benchmarktest[64]/linker-reloc-bench`. 11 12There is also a `run_bench_with_ninja.sh` script that uses the 13`gen_bench.py --ninja` mode to generate a benchmark. It's useful for 14experimentation. The `--cc` and `--linker` flags allow swapping out different 15static and dynamic linkers. 16 17## Regenerating the synthetic benchmark 18 19`regen/dump_relocs.py` scans an ELF file and its dependencies, outputting a JSON 20dump, then `regen/gen_bench.py` processes the JSON file into benchmark code. 21`gen_bench.py` has two modes: 22 23 - (default) generate Android.bp and source files 24 - with `--ninja`: generate a build.ninja instead, and build a set of ELF file 25 outputs 26