Performance bottleneck in `MergedOperator.parse`
Yuriyzabegaev opened this issue · comments
Yury Zabegaev commented
According to the line profiler, 90% of time spent in Operator.evaluate
(value without Jacobian) is in MergedOperator.parse
, see profile.md. Inside, the majority of time is spent in scipy.sparse.block_diag
. A more efficient implementation may be considered.