Mercurial > repos > public > sbplib_julia

comparison README.md @ 1378:2b5480e2d4bf tooling/benchmarks

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
Fix some errors in README.md
author Jonatan Werpers <jonatan@werpers.com>
date Wed, 31 May 2023 08:59:34 +0200
parents 68dd0c9f5cac
children 654a2b7e6824
comparison
equal deleted inserted replaced
1366:6bdb6fed6ed3 1378:2b5480e2d4bf
27 Pkg.test(test_args=["*/lazy_tensor_operations_test.jl", "Grids/*"]) 27 Pkg.test(test_args=["*/lazy_tensor_operations_test.jl", "Grids/*"])
28 ``` 28 ```
29 will run any file named `lazy_tensor_operations_test.jl` and all the files in the `Grids` folder. 29 will run any file named `lazy_tensor_operations_test.jl` and all the files in the `Grids` folder.
30 30
31 ## Running benchmarks 31 ## Running benchmarks
32 Benchmarks are defined in `benchmark/` and use the tools for benchmark suites in BenchmarkTools.jl. The format is compatible with PkgBenchmark.jl which helps with running the suite, comparing results and presenting the results in a readable way. There are custom functions included for running the benchmarks in this Mercurial repository. 32 Benchmarks are defined in `benchmark/` and use the tools for benchmark suites
33 in BenchmarkTools.jl. The format is compatible with PkgBenchmark.jl which
34 helps with running the suite, comparing results and presenting the results in
35 a readable way. There are custom functions included for running the benchmarks
36 in this Mercurial repository.
33 37
34 `benchmark/` contains a julia environment with the necessary packages for working with the benchmarks. 38 `benchmark/` contains a julia environment with the necessary packages for
39 working with the benchmarks.
35 40
36 To run the benchmarks, either use `make` run them manually from the REPL, as explained further below. 41 To run the benchmarks, either use `make` or run them manually from the REPL, as
42 explained further below.
37 43
38 Using `make` there are four targets for benchmarks 44 Using `make` there are four targets for benchmarks
39 ```shell 45 ```shell
40 make benchmark # Runs the suite for the current working directory 46 make benchmark # Runs the suite for the current working directory
41 make benchmarkrev REV=rev # Runs the suite at the specified revision 47 make benchmarkrev REV=rev # Runs the suite at the specified revision
42 make benchmarkcmp TARGET=target BASELINE=baseline # Compares two revisions 48 make benchmarkcmp TARGET=target BASELINE=baseline # Compares two revisions
43 make cleanbenchmark # Cleans up benchmark tunings and results 49 make cleanbenchmark # Cleans up benchmark tunings and results
44 ``` 50 ```
45 Here `rev`, `target` and `baseline` are any valid Mercurial revision specifiers. Note that `make benchmarkrev` and `make benchmarkcmp` will fail if you have pending changes in your repository. 51 Here `rev`, `target` and `baseline` are any valid Mercurial revision
52 specifiers.
46 53
47 54 Alternatively, the benchmarks can be run from the REPL. To do this, first
48 Alternatively, the benchmarks can be run from the REPL. To do this, first activate the environment in `benchmark/` then include the file `benchmark_utils.jl`. The suite can then be run using the function `main` in one of the following ways 55 activate the environment in `benchmark/` then include the file
56 `benchmark_utils.jl`. The suite can then be run using the function `main` in
57 one of the following ways
49 58
50 ```julia 59 ```julia
51 main() # Runs the suite for the current working directory 60 main() # Runs the suite for the current working directory
52 main(rev) # Runs the suite at the specified revision 61 main(rev="...") # Runs the suite at the specified revision
53 main(target, baseline) # Compares two revisions 62 main(target="...", baseline="...") # Compares two revisions
54 ``` 63 ```
55 64
56 Again, `rev`, `target` and `baseline` are any valid Mercurial revision specifiers. Note that `main(rev)` and `main(target, baseline)` will fail if you have pending changes in your repository. 65 Again, `rev`, `target` and `baseline` are any valid Mercurial revision
66 specifiers.
57 67
58 PkgBenchmark can also be used directly. 68 PkgBenchmark can also be used directly.
59 69
60 ```julia 70 ```julia
61 using PkgBenchmark 71 using PkgBenchmark