The reason for not recommending networkx is the choice it offers:
reply fast on the GPU: nx-cugraph
slow on the CPU: networkx (no arrow, see linked graph-benchmark)
Now that nvidia is becoming a major CPU vendor, they may like the idea too.
agreed. gfql already gives this best-of-both for oss cpu+gpu cypher as we've been steadily working through benchmarks like ldbc, pokec, etc.
replyThe HPC lessons for graph algorithms is less obvious as the substrate the OP is working through is close but not quite what HPC folks figured out, so the trade-off of fast to write and maintainable (dataframe/db-based) vs at the achievable magnitude of performance is tricky. LLMs change the calculus here too IMO, so I've been thinking a lot about more NUMA-exposed ideas that before were relegated to PhD land.