Innlegg

John Pham
John Pham
One of the biggest challenges in Fully Homomorphic Encryption (FHE) has always been performance. Traditional FHE processes computations through individual gates. Even a simple 64-bit operation could require thousands of gates and bootstrappings, creating a major bottleneck for confidential onchain applications. This is where @primus_labs is taking an interesting approach with AutoHoG. Instead of treating every gate as an isolated operation, AutoHoG automatically designs custom multi-input, multi-output gates and combines compatible subcircuits into a single bootstrapping process. The performance improvements are particularly impressive: Around 90 ms for one 64-bit confidential transfer. More than 210 transfers per second. Tested across 8 consumer RTX 5090 GPUs. Integer-based computation delivers byte-identical results across nodes. Why does this matter? Because privacy technology is only useful at scale if it can deliver both confidentiality and practical performance. Primus Labs is working to address one of the fundamental barriers to institutional onchain finance: making encrypted computation faster, more efficient, and deterministic.

Ansvarsfraskrivelse: OKX Orbit-innholdet er kun gitt for informasjonsformål. Finn ut mer

Svar

Ingen kommentarer ennå. Vær den første til å svare!