Geir Storli

In mixed and sparse tensor attributes we are heap allocating the memory needed for each tensor value. These allocations are managed by _vespamalloc_. When such tensor attribute is downsized (e.g....

Currently, only `first-phase` and `second-phase` support `rank-score-drop-limit`. In some contexts, you can discover that the retrieved document was irrelevant after having spent more resources on re-ranking in `global-phase`. This is...