Skip to content

Precise

Capable of detecting 1% difference in runtime in ideal conditions

julia
julia> f(n) = sum(rand() for _ in 1:n)
f (generic function with 1 method)

julia> @b f(1000)
1.074 μs

julia> @b f(1000)
1.075 μs

julia> @b f(1000)
1.076 μs

julia> @b f(1010)
1.086 μs

julia> @b f(1010)
1.087 μs

julia> @b f(1010)
1.087 μs

Efficient

ChairmarksBenchmarkToolsRatio
TTFX3.4s13.4s4x
TTFX excluding precompilation43ms1118ms26x
Load time4.2ms131ms31x
minimum runtime34μs459ms13,500x
default runtime0.1s5s50x
proportion of time spent benchmarking90%-99%13%-65%1.5-7x

See https://github.com/LilithHafner/Chairmarks.jl/blob/main/contrib/ttfx_rm_rf_julia.sh for methodology on the first four entries and https://github.com/LilithHafner/Chairmarks.jl/blob/main/contrib/efficiency.jl for the last.

Concise

Chairmarks uses a concise pipeline syntax to define benchmarks. When providing a single argument, that argument is automatically wrapped in a function for higher performance and executed

julia
julia> @b sort(rand(100))
1.500 μs (3 allocs: 2.625 KiB)

When providing two arguments, the first is setup code and only the runtime of the second is measured

julia
julia> @b rand(100) sort
1.018 μs (2 allocs: 1.750 KiB)

You may use _ in the later arguments to refer to the output of previous arguments

julia
julia> @b rand(100) sort(_, by=x -> exp(-x))
5.521 μs (2 allocs: 1.750 KiB)

A third argument can run a "teardown" function to integrate testing into the benchmark and ensure that the benchmarked code is behaving correctly

julia
julia> @b rand(100) sort(_, by=x -> exp(-x)) issorted(_) || error()
ERROR:
Stacktrace:
 [1] error()
[...]

julia> @b rand(100) sort(_, by=x -> exp(-x)) issorted(_, rev=true) || error()
5.358 μs (2 allocs: 1.750 KiB)

See @be for more info

Truthful

On versions of Julia prior to 1.8, Chairmarks automatically computes a checksum based on the results of the provided computations and returns that checksum to the user along with benchmark results. This makes it impossible for the compiler to elide any part of the computation that has an impact on its return value.

While the checksums are fast, one negative side effect of this is that they add a bit of overhead to the measured runtime, and that overhead can vary depending on the function being benchmarked. In versions of Julia 1.8 and later, these checksums are emulated using the function Base.donotdelete which is designed and documented to ensure that necessary computation is not elided without adding extra overhead. You can disable all of this on all versions of Julia by passing the checksum=false keyword argument, possibly in combination with a custom teardown function that verifies computation results. Be aware that as the compiler improves, it may become better at eliding benchmarks whose results are not saved.

julia
julia> @b rand hash
2.276 ns

julia> @b rand hash checksum=false
0 ns

You may experiment with custom reductions using the internal _map and _reduction keyword arguments. The default maps and reductions (Chairmarks.default_map and Chairmarks.default_reduction) are internal and subject to change and/or removal in the future.

Innate qualities

Chairmarks is inherently narrower than BenchmarkTools by construction. It also has more reliable back support. Back support is a defining feature of chairs while benches are known to sometimes lack back support.