Dask added an additional keyword-only argument ``method``. method : {'sequential', 'blelloch'}, optional Choose which method to use to perform the cumsum. Default is 'sequential'. * 'sequential' performs the cumsum of each prior block before the current block. * 'blell
(x, axis, dtype=None, out=None, *, method="sequential")
| 213 | |
| 214 | @derived_from(np) |
| 215 | def nancumsum(x, axis, dtype=None, out=None, *, method="sequential"): |
| 216 | """Dask added an additional keyword-only argument ``method``. |
| 217 | |
| 218 | method : {'sequential', 'blelloch'}, optional |
| 219 | Choose which method to use to perform the cumsum. Default is 'sequential'. |
| 220 | |
| 221 | * 'sequential' performs the cumsum of each prior block before the current block. |
| 222 | * 'blelloch' is a work-efficient parallel cumsum. It exposes parallelism by |
| 223 | first taking the sum of each block and combines the sums via a binary tree. |
| 224 | This method may be faster or more memory efficient depending on workload, |
| 225 | scheduler, and hardware. More benchmarking is necessary. |
| 226 | """ |
| 227 | return cumreduction( |
| 228 | chunk.nancumsum, |
| 229 | operator.add, |
| 230 | 0, |
| 231 | x, |
| 232 | axis, |
| 233 | dtype, |
| 234 | out=out, |
| 235 | method=method, |
| 236 | preop=np.nansum, |
| 237 | ) |
| 238 | |
| 239 | |
| 240 | @derived_from(np) |
nothing calls this directly
no test coverage detected