harfbuzz/perf
Behdad Esfahbod 98fbe87a26 [benchmark-font] Disable quadratic callback
We are interested in the quadratic-to-cubic codepath benchmarking.
2022-06-27 12:10:16 -06:00
..
fonts [perf] Make subset benchmark data driven. 2022-05-10 14:30:08 -06:00
texts [benchmark-shape] Remove a few tests 2022-06-05 02:04:02 -06:00
Makefile.am [perf] Add benchmark-ot 2022-05-17 13:12:17 -06:00
README.md [perf] Add missing ninja command in profiling instructions. 2022-04-21 11:16:12 -06:00
benchmark-font.cc [benchmark-font] Disable quadratic callback 2022-06-27 12:10:16 -06:00
benchmark-map.cc Include cassert to fix bots 2022-05-30 05:09:53 -06:00
benchmark-ot.cc [perf/benchmark-ot] Add zh-hans 2022-05-18 11:16:46 -06:00
benchmark-set.cc [benchmark-set] Another Pause/Resume 2022-05-30 05:54:20 -06:00
benchmark-shape.cc [benchmark-shape,hb-shape-threads] Fix argument parsing order 2022-06-05 06:53:42 -06:00
benchmark-subset.cc Include cassert to fix bots 2022-05-30 05:09:53 -06:00
meson.build [perf/benchmark-shape] Test ft font backend as well 2022-05-23 14:36:06 -06:00
run.sh [perf] More rename 2019-12-10 12:43:42 -06:00

README.md

Building and Running

Benchmarks are implemented using Google Benchmark.

To build the benchmarks in this directory you need to set the benchmark option while configuring the build with meson:

meson configure build -Dbenchmark=enabled --buildtype=release

Then build the benchmark binaries with ninja:

ninja -Cbuild perf/benchmark-set

Finally, to run one of the benchmarks:

./build/perf/benchmark-set

It's possible to filter the benchmarks being run and customize the output via flags to the benchmark binary. See the Google Benchmark User Guide for more details.

Profiling

Configure the build to include debug information for profiling:

CXXFLAGS="-fno-omit-frame-pointer" meson --reconfigure build -Dbenchmark=enabled --buildtype=debug
ninja -Cbuild

Then run the benchmark with perf:

perf record -g -o ~/tmp/profiles/subset.prof ./build/perf/benchmark-subset --benchmark_filter="BM_subset_codepoints/subset_notocjk/100000" --benchmark_repetitions=5

You probably want to filter to a specific benchmark of interest and set the number of repititions high enough to get a good sampling of profile data.

Finally view the profile with:

perf report -i ~/tmp/profiles/subset.prof