Glossary

Custom Compute

What is Custom Compute?

Teams designing SoCs need new ways to create differentiated products as our industry faces the challenge of new applications requiring new demanding algorithms. How? Through architecture optimization, hardware/software co-optimization, and domain-specific acceleration. In other words: through Custom Compute.

Our vision at Codasip is to enable Custom Compute: the ability to tune the hardware to the type of software workloads to reach unparalleled efficiency. This makes it possible to use much less power for a given class of algorithms, or to increase performance in a given power budget, with gains of even 10x to 100x possible depending on the type of application.

Why Custom Compute?

One of the key benefits of RISC-V is the possibility to tailor both the ISA and the microarchitecture of the processor to meet specific application requirements. This customization capability extends to Custom Compute solutions, enabling developers to create hardware optimized for their workloads.

Codasip glossary - image for custom compute definition

Custom Compute helps you meet the demand for greater computational performance while semiconductor scaling is showing its limits. By innovating architecturally and creating more specialized processing units, you customize your compute for your unique needs. Custom Compute goes beyond adding custom instructions to a processor and includes optimization such as increasing the data throughput between the processor and interfaces or other processing elements, parallelizing some operations, managing specific data types, or adding application-related features.

 

Getting started with Codasip