How amineux/hysmap works
HySMap is a C++20 SNN-to-mesh mapper that models each neuron's axon as a multicast hyperedge and optimizes activity-weighted routing cost. Its principal workflow is to load or generate a network, represent it as a directed hypergraph, partition neurons and place logical cores on a mesh, refine the mapping against multicast cost, then report or export results. The README documents this workflow and the mapper variants; source-index import lists support some internal dependencies, but several implementation details were not excerpted.
Main entry points: mapper.cpp Mapper orchestration · hypergraph.cpp Directed hypergraph · cli.cpp CLI · bindings.cpp Python bindings
Components 17
User Interface
- CLI
src/cli.cpp - Python bindings
python/bindings.cpp
Network Model
- Network JSON
src/io.cpp - SNN generator
src/snn_generator.cpp - Directed hypergraph
src/hypergraph.cpp
Mapping
- Mesh NoC
src/mesh.cpp - Neuron partitioning
src/partition.cpp - Core placement
src/placement.cpp - Mapper orchestration
src/mapper.cpp
Evaluation
- Incremental evaluator
src/incremental.cpp - Multicast cost
src/cost.cpp - Parallel work
include/hysmap/parallel.hpp - Mapping and metrics
include/hysmap/types.hpp
Results
- Metrics and results
src/io.cpp - Loihi-style export
src/export.cpp
Other components
- User
- Mapping results
Connections 19
- User → CLI: runs commands
README.md - User → Python bindings: uses bindings
README.md - Network JSON → Directed hypergraph: loads network
README.md - SNN generator → Directed hypergraph: generates network
README.md - Directed hypergraph → Mapper orchestration: maps network
README.md - Mesh NoC → Mapper orchestration: provides topology
README.md - Mapper orchestration → Neuron partitioning: seeds and refines
README.md - Mapper orchestration → Core placement: seeds and places
README.md - Mapper orchestration → Incremental evaluator: refines by ΔJ
README.md - Mapper orchestration → Multicast cost: evaluates mapping
README.md - Incremental evaluator → Multicast cost: updates cost
README.md - Mapper orchestration → Parallel work: parallelizes gains
README.md - Mapper orchestration → Mapping and metrics: returns mapping
- Mapper orchestration → Metrics and results: reports metrics
README.md - Mapper orchestration → Loihi-style export: exports mapping
README.md - Metrics and results → Mapping results: writes results
README.md - Loihi-style export → Mapping results: writes export
README.md - Python bindings → Mapper orchestration: runs mapper
README.md - CLI → Mapper orchestration: runs mapping
README.md