You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
This directory contains a benchmark suite for TensorKit.
4
+
Most of the benchmarks are designed to capture performance characteristics of the library, and are not intended to be used as a benchmark suite for comparing different libraries.
5
+
In particular, the main goal here is to catch performance regressions and/or improvements between different versions of TensorKit.
6
+
7
+
## Running the benchmarks
8
+
9
+
The benchmarks are written using `BenchmarkTools.jl`, and the full suite can be found in the `SUITE` global variable defined in `benchmarks.jl`.
10
+
Sometimes, it is useful to run only a subset of the benchmarks.
11
+
To do this, you can use the `--modules` flag to specify which modules to run.
12
+
Alternatively, you can use the `TensorKitBenchmarks` module directly, which is designed after `BaseBenchmarks` to allow for conditional loading of the benchmarks.
13
+
14
+
For a more streamlined CLI experience, you can use [`AirspeedVelocity.jl`](https://github.com/MilesCranmer/AirspeedVelocity.jl) to run the benchmarks.
15
+
The following command will run the benchmarks and compare with the current master branch:
16
+
17
+
```bash
18
+
benchpkg TensorKit \
19
+
--rev=dirty,master \
20
+
-o benchmark/results/ \
21
+
-exeflags="--threads=4"
22
+
```
23
+
24
+
To compare with previous results, the following command can be used:
0 commit comments