Custom 8-bit floating point value format for reducing shared memory bank conflict in approximate nearest neighbor search.
Hiroyuki OotomoAkira NarusePublished in: CoRR (2023)
Keyphrases
- floating point
- shared memory
- approximate nearest neighbor search
- address space
- message passing
- parallel algorithm
- parallel computing
- nearest neighbor
- hash functions
- distributed memory
- fixed point
- instruction set
- locality sensitive hashing
- nearest neighbor search
- similarity search
- kd tree
- floating point arithmetic
- parallel machines
- metric space
- application specific
- memory access
- object recognition
- general purpose
- data structure
- computer vision
- quadtree
- fine grained
- similarity measure