mirror of
https://github.com/ruby/ruby.git
synced 2026-01-27 04:24:23 +00:00
Now that we've inlined the eden_heap into the size_pool, we should rename the size_pool to heap. So that Ruby contains multiple heaps, with different sized objects. The term heap as a collection of memory pages is more in memory management nomenclature, whereas size_pool was a name chosen out of necessity during the development of the Variable Width Allocation features of Ruby. The concept of size pools was introduced in order to facilitate different sized objects (other than the default 40 bytes). They wrapped the eden heap and the tomb heap, and some related state, and provided a reasonably simple way of duplicating all related concerns, to provide multiple pools that all shared the same structure but held different objects. Since then various changes have happend in Ruby's memory layout: * The concept of tomb heaps has been replaced by a global free pages list, with each page having it's slot size reconfigured at the point when it is resurrected * the eden heap has been inlined into the size pool itself, so that now the size pool directly controls the free_pages list, the sweeping page, the compaction cursor and the other state that was previously being managed by the eden heap. Now that there is no need for a heap wrapper, we should refer to the collection of pages containing Ruby objects as a heap again rather than a size pool
ruby/benchmark
This directory has benchmark definitions to be run with benchmark_driver.gem.
Normal usage
Execute gem install benchmark_driver and run a command like:
# Run a benchmark script with the ruby in the $PATH
benchmark-driver benchmark/app_fib.rb
# Run benchmark scripts with multiple Ruby executables or options
benchmark-driver benchmark/*.rb -e /path/to/ruby -e '/path/to/ruby --jit'
# Or compare Ruby versions managed by rbenv
benchmark-driver benchmark/*.rb --rbenv '2.5.1;2.6.0-preview2 --jit'
# You can collect many metrics in many ways
benchmark-driver benchmark/*.rb --runner memory --output markdown
# Some are defined with YAML for complex setup or accurate measurement
benchmark-driver benchmark/*.yml
See also:
Usage: benchmark-driver [options] RUBY|YAML...
-r, --runner TYPE Specify runner type: ips, time, memory, once, block (default: ips)
-o, --output TYPE Specify output type: compare, simple, markdown, record, all (default: compare)
-e, --executables EXECS Ruby executables (e1::path1 arg1; e2::path2 arg2;...)
--rbenv VERSIONS Ruby executables in rbenv (x.x.x arg1;y.y.y arg2;...)
--repeat-count NUM Try benchmark NUM times and use the fastest result or the worst memory usage
--repeat-result TYPE Yield "best", "average" or "worst" result with --repeat-count (default: best)
--alternate Alternate executables instead of running the same executable in a row with --repeat-count
--bundler Install and use gems specified in Gemfile
--filter REGEXP Filter out benchmarks with given regexp
--run-duration SECONDS Warmup estimates loop_count to run for this duration (default: 3)
--timeout SECONDS Timeout ruby command execution with timeout(1)
-v, --verbose Verbose mode. Multiple -v options increase visilibity (max: 2)
make benchmark
Using make benchmark, make update-benchmark-driver automatically downloads
the supported version of benchmark_driver, and it runs benchmarks with the downloaded
benchmark_driver.
# Run all benchmarks with the ruby in the $PATH and the built ruby
make benchmark
# Or compare with specific ruby binary
make benchmark COMPARE_RUBY="/path/to/ruby --jit"
# Run vm benchmarks
make benchmark ITEM=vm
# Run some limited benchmarks in ITEM-matched files
make benchmark ITEM=vm OPTS=--filter=block
# You can specify the benchmark by an exact filename instead of using the default argument:
# ARGS = $$(find $(srcdir)/benchmark -maxdepth 1 -name '*$(ITEM)*.yml' -o -name '*$(ITEM)*.rb')
make benchmark ARGS=benchmark/erb_render.yml
# You can specify any option via $OPTS
make benchmark OPTS="--help"
# With `make benchmark`, some special runner plugins are available:
# -r peak, -r size, -r total, -r utime, -r stime, -r cutime, -r cstime
make benchmark ITEM=vm_bigarray OPTS="-r peak"