A broadcast-based parallel R-tree method on UPMEM PIM hardware delivers up to 3.66x kernel speedup and 3.4x lower energy than CPU execution on datasets with up to 16M rectangles.
The true processing in memory accelerator
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.DB 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Parallel R-tree-based Spatial Query Processing on a Commercial Processing-in-Memory System
A broadcast-based parallel R-tree method on UPMEM PIM hardware delivers up to 3.66x kernel speedup and 3.4x lower energy than CPU execution on datasets with up to 16M rectangles.