Composable Computing
Abstract program intent and expected behaviors
Composable Computing abstracts algorithm details, provides CPU-like programmability, and delivers the performance and efficiency of customized, single-purpose ASICs based on the terms of several key behaviors.
Create a novel compiler to compose the program into expected behaviors
Implementing a novel compiler that can intelligently compose the program into those behaviors, integrating it directly into the backends for TensorFlow, ONNX, and PyTorch. This enables existing AI code bases to integrate seamlessly.
Modify hardware in real-time to execute the program
Executing the program on physically modified hardware in real time enables behavioral integration—as if the hardware were customized for the program—which allows directly manipulating the program and understanding its properties.
ADVANTAGES
Composable compute platform for accelerating AI and ML environments
Flexibility
Future-ready for emerging models and diverse workloads
Faster Time-to-Results
Runs up to 64 different models simultaneously
Ease-of-Use
Supports a wide range of different use cases
Faster ROI
Rapid deployment, reduced TCO
A solution that meets you where you are
As the world increasingly becomes algorithmic and data-driven, today’s data analytics are continually evolving to address dynamic analytic requirements. With our solution’s programmability and time scalability, engineers can develop applications and efficiently execute software for various computing needs across many industries.
Click to see Deployment
On Cloud
Composable Computing is compatible with applications in the cloud…
On Premise
…or it can work and integrate seamlessly with software on-premise.