Fastpath Benchmark Catalog
This section documents benchmark definitions available from the Fastpath
benchmark library. Each benchmark has its own generated page named from the
suite/name benchmark id, with / replaced by -. For example,
lmbench/lat-mem-rd is documented in lmbench-lat-mem-rd.rst.
This page documents the benchmark catalog. For the plan schema and examples showing how benchmark fragments are included from plans, see Plan Schema.
The fastpath benchmark command inspects benchmark definitions from the
configured Fastpath benchmark libraries. It is separate from
fastpath result commands: fastpath benchmark describes benchmark
definitions and documentation metadata, while fastpath result reads data
that has already been collected into a resultstore.
Listing Benchmarks
Use fastpath benchmark list to list the benchmarks available from the
configured benchmark libraries:
fastpath benchmark list
The output shows each benchmark id in suite/name form, a summary
when documentation metadata is available, and the rendered documentation URL.
Use --id to filter the list by exact id or shell-style glob:
fastpath benchmark list --id "lmbench/*"
Rendering Benchmark Documentation
Use fastpath benchmark show to render the RST documentation for one
benchmark id:
fastpath benchmark show --id lmbench/lat-mem-rd
To write the RST output to a file, use --output with any target RST file
name:
fastpath benchmark show \
--id lmbench/lat-mem-rd \
--output lat-mem-rd-details.rst
The benchmark catalog below contains one page for each benchmark id available from the Fastpath benchmark library.
Benchmark Page Layout
Each benchmark page starts with a short summary and a benchmark profile:
Typedescribes whether the benchmark is a focused micro benchmark or a broader macro workload.Kernel subsystemnames the Linux kernel areas most directly exercised by the benchmark, or the kernel areas most likely to influence the workload when the signal is indirect.Coverageexplains whether the benchmark directly measures the named kernel behavior or infers sensitivity through a larger workload.Topologydescribes the Fastpath role layout required to run the benchmark.Single role: executeris the default Fastpath role used when a benchmark does not declare explicit roles in its YAML definition.
The remaining sections describe the benchmark overview, Fastpath parameters, result-class naming, and the source benchmark metadata used to generate the page.
Benchmarks
- lmbench/lat-mem-rd
- micromm/fork
- micromm/munmap
- micromm/vmalloc
- mmtests/hackbench
- mmtests/kernbench
- mmtests/stressng-af-alg
- mmtests/stressng-bad-altstack
- mmtests/stressng-class-io-parallel
- mmtests/stressng-context
- mmtests/stressng-fork
- mmtests/stressng-get
- mmtests/stressng-getdent
- mmtests/stressng-madvise
- mmtests/stressng-mmap
- mmtests/stressng-vm-splice
- mmtests/stressng-wait
- mmtests/stressng-zombie
- mmtests/sysbench-cpu
- mmtests/sysbench-mutex
- mmtests/sysbench-thread
- perf/futex
- perf/sched
- perf/syscall
- pts/memtier-benchmark
- pts/nginx
- pts/perl-benchmark
- pts/pgbench
- pts/pybench
- pts/redis
- pts/sqlite-speedtest
- repro-collection/mysql-workload
- schbench/thread-contention
- selftest/echo
- sockperf/echo-lat-tcp
- sockperf/echo-lat-udp
- sockperf/packet-tp-tcp
- sockperf/packet-tp-udp
- speedometer/v2.0
- speedometer/v2.1
- syscall/getpid
- syscall/getppid
- syscall/invalid