Stéphane Caron
Stéphane Caron
### Is your proposal related to a problem? No. ### Describe the solution you'd like The benchmark only performs cold-start calls. It would be nice to cater to warm and...
Currently we time solvers externally, so that the timing will include both matrix build and solve times. This is done in `qpbenchmark/utils.py` by, essentially: ```python start_time = perf_counter() solution =...
### Is your proposal related to a problem? From https://github.com/qpsolvers/qpsolvers_benchmark/discussions/87: > Unless very efficiently cooled, modern CPU cannot run at max speed for long. This can considerably skew timings. ###...
The current submission process asks for Python code. Alternatively, we could accept matrix files in a specific format (same as the current Maros-Meszaros test set).
As @bodono pointed out in https://github.com/qpsolvers/qpsolvers_benchmark/issues/58#issuecomment-1443914166, [`socp_from_qp`](https://qpsolvers.github.io/qpsolvers/developer-notes.html#qpsolvers.conversions.socp_from_qp) relies on NumPy's implementation, which in turns require the matrix to be positive definite. This requirement comes from the conversion, not from the...
The conversion of SCS problems may be slightly suboptimal, see https://github.com/stephane-caron/qpsolvers_benchmark/issues/19#issuecomment-1313363422 For instance on QETAMACR: ``` - Solving QETAMACR with solver SCS-3.0 - Optimal objective 86760.37 ------------------------------------------------------------------ SCS v3.2.1 -...
### Is your proposal related to a problem? Arose from this discussion: https://github.com/qpsolvers/qpsolvers_benchmark/discussions/65 ### Describe the solution you'd like Use `qpsolvers_benchmark` as a library rather than a command-line script so...
Any measure of the mean value of data is misleading when there is large variance. We should provide some low and high estimates (min and max, or 5% and 95%...