Cadence Tensilica FloatingPoint DSP family targets compute-intensive applications
Cadence Design Systems has unveiled the Tensilica FloatingPoint DSP family, which has been designed to provide a scalable and configurable solution specifically for floating-point-centric workloads.
These new DSP IP cores extend from small, ultra-low power to very high performance and are well-suited for a broad array of applications ranging from: energy-efficient solutions for battery-operated devices to artificial intelligence/machine learning (AI/ML), motor control, sensor fusion, object tracking and augmented reality/virtual reality (AR/VR) applications in the mobile, automotive, hyperscale computing and consumer markets.
The family is comprised of four cores: the Tensilica FloatingPoint KP1 DSP, the Tensilica FloatingPoint KP6 DSP, the Tensilica FloatingPoint KQ7 DSP and the Tensilica FloatingPoint KQ8 DSP.
Each shares a common instruction set architecture (ISA) with existing Tensilica DSPs’ optional vector floating-point unit (VFPU), promoting software portability and reusability while enabling easy offloading of floating-point workloads.
Scalable from 128-bit SIMD to 1024-bit SIMD on both the Tensilica Xtensa LX and NX platforms, the FloatingPoint DSPs deliver a 25% improvement in fused multiply-add (FMA) operations compared to Tensilica fixed-point DSPs with the VFPU add-on, contributing to greater operational throughput. Performance can be further enhanced and differentiated using the Tensilica Instruction Extension (TIE) language. In addition, the FloatingPoint DSPs offer up to 40% area savings compared to the similar class of fixed-point DSPs with VFPUs.
High-performance software tools provide effective auto-vectorization that helps optimise the scalar code to utilize the vector floating-point units with minimal to no manual effort. Support for optimised Eigen, NatureDSP, simultaneous location and mapping (SLAM) and math software libraries enables easier development of high-performance software.
The Tensilica FloatingPoint DSPs also offer a software development environment thatsupports the seamless migration of the existing floating-point common software stack to the new Tensilica FloatingPoint DSPs and between the FloatingPoint DSPs within the family.
“Floating-point numbers are used widely in modern computations across a broad range of compute-intensive applications, and the need for floating-point processing is growing,” said Larry Przywara, senior group director, Tensilica marketing at Cadence. “Energy-efficient, cost-effective and high-performance DSPs designed specifically for floating-point-centric computation are critical for developing competitive and differentiated products. The scalable Tensilica FloatingPoint DSP family provides optimal PPA for these floating-point computations, regardless of the application. These DSPs are an example of how Cadence is applying our computational software prowess to hardware to solve our customers’ design challenges.”