Open xspronken opened 2 years ago
The information currently recorded is:
"group": "Add-2-dense-numpy",
"name": "test_add[2-dense-numpy]",
"fullname": "test_linear_algebra.py::test_add[2-dense-numpy]",
"params": {
"size": 2,
"density": "dense",
"dtype": "numpy"
}
For each group me might want to add a parameter or extra_param named "implementation" that specifies the implementation used, e.g. "v5-dense", "julia", "v5-sparse", "v4-sparse", ... . I think the name of the test sort of gives this information, but not explicitly, and combines it with the group information. Have a variable that clearly names the implementation would be useful.
pytest-benchmark provides an existing mechanism for setting these things that looks quite nice -- https://pytest-benchmark.readthedocs.io/en/latest/usage.html#markers
The groups are currently being changed in #22.
I'd like to keep this open or create new issues for the ideas of adding a decorator (mostly stylistic) and an implementation
extra_info (which would be useful for organizing results by implementation).
As @hodgestar mentioned in #6 the contents of
group
should probably be changed. For the benchmarks in the existing PR i would suggest somehting likeQobj-operation
andQobjEvo-operation
or maybestatic
andtimedependent
instead ofQobj
andQobjEvo
.