I'd like to share with you a new concept of transforming data. I'll try to desribe the subject as well as I can and as short as I can so it doesn't end being an essay.
The engine is called Lustro. It's an experimental, deterministic diffusion mechanism with an unusual property: the control parameters are dynamically derived from two co-evolving state channels. It's a single pass engine, no multiple rounds of transformation. All current functions (Hash, PRNG, XOF) use the exact same engine. Unlike most solutions available right now Lustro is not a repeating, layered pipeline. It's a dynamic, highly coupled system. The main mechanism is called ERD (Evolving Representation Dynamics). My observations suggest that the information propagation is approx isotropic. On my PC I was able to exceed 13GB/s on an i5-11600K at 4.5GHz constant speed.
Within V1 there is a division of responsibility in data processing between two mechanisms. I like to call these - a worker and an conductor (IDM + ERD). My design philosophy was to build something different. Something that behaves more like a system that "flows", for the lack of a better word. I belive it's due to my educational backgroud. I'm not a programmer by education, I'm a chemist (hence the specific vocabulary at times). A bit of an obessive hermit by nature and due to some extraordinary circumstances I decided to approach the problem of data processing from a different angle. In case of Lustro V1 the parameters are rotation values that define the final permutation.
ERD has a specific method of linking the data channels. The data basically controls itself, so to say. Sort of like a chemical reaction, that's how I like to think about it ;-)
I'd really appreciate your take on the subject. I've done a lot of testing to investigate the bahaviour of the system. Everything is included in the repo to which I'm linking.
Ps. I'm Polish and in my native language Lustro means "mirror". It is a project of deep personal meaning for me.
Ligatum•47m ago
I'd like to share with you a new concept of transforming data. I'll try to desribe the subject as well as I can and as short as I can so it doesn't end being an essay.
The engine is called Lustro. It's an experimental, deterministic diffusion mechanism with an unusual property: the control parameters are dynamically derived from two co-evolving state channels. It's a single pass engine, no multiple rounds of transformation. All current functions (Hash, PRNG, XOF) use the exact same engine. Unlike most solutions available right now Lustro is not a repeating, layered pipeline. It's a dynamic, highly coupled system. The main mechanism is called ERD (Evolving Representation Dynamics). My observations suggest that the information propagation is approx isotropic. On my PC I was able to exceed 13GB/s on an i5-11600K at 4.5GHz constant speed.
Within V1 there is a division of responsibility in data processing between two mechanisms. I like to call these - a worker and an conductor (IDM + ERD). My design philosophy was to build something different. Something that behaves more like a system that "flows", for the lack of a better word. I belive it's due to my educational backgroud. I'm not a programmer by education, I'm a chemist (hence the specific vocabulary at times). A bit of an obessive hermit by nature and due to some extraordinary circumstances I decided to approach the problem of data processing from a different angle. In case of Lustro V1 the parameters are rotation values that define the final permutation.
ERD has a specific method of linking the data channels. The data basically controls itself, so to say. Sort of like a chemical reaction, that's how I like to think about it ;-)
I'd really appreciate your take on the subject. I've done a lot of testing to investigate the bahaviour of the system. Everything is included in the repo to which I'm linking.
Ps. I'm Polish and in my native language Lustro means "mirror". It is a project of deep personal meaning for me.