ChipWrights’ ViSP Technology Excels in Image Processing

ChipWrights' Visual Signal Processor (ViSP) technology excels in 2D processing for still image and video data. ChipWrights' enhanced vector architecture enables the processor to work on up to 64 data items simultaneously, while providing high memory bandwidth via an interleaved on-chip memory. This provides the performance of the fastest VLIW processors while maintaining the power and cost characteristics necessary for battery-operated consumer devices.

High Performance Image Processing at a Low Cost:

High performance image processing at consumer-level price and power points

Low-cost integrated IO for most applications

Rich software feature set from the integrated ARM® core

Flexibility from programmability to support any application

ChipWrights ViSP Architecture
ChipWrights uses a single-instruction-multiple-data (SIMD) architecture to maximize the logic and power that are devoted to handling data. While other vendors add SIMD as an accessory to their older architectures, SIMD is at the core of what the CWvX processors do. The instruction set, address indexing schemes, and data access mechanisms are tuned for image and video processing. Even the system Direct Memory Access engines are architected to move data in 2D patches.

32-bit Parallel DSPs (Parallel Datapath)	Optimized for image processing and programmability Very Dense Instruction Word (VDIW) - can do up to 256 operations in one instruction Available in 8 and 16 datapath configurations
32-bit RISC Processor (Serial Datapath)	Optimized to support parallel datapaths Provides an easy compiler target for non-critical code
High Speed On-Chip Data Memory	Up to 8 accesses per cycle Strided and scatter/gather addressing Burst ports for DMA

ViSP Vs. DSP

"Most general-purpose DSPs are designed for communications applications - they don't exploit the parallelism possible in image processing because that's harder to find in communications.

Our new DSP architecture—the CWvX Visual Signal Processor—applies techniques from vector architectures to achieve high performance while meeting the low cost and low power needs of portable image-processing devices."

- John Redford, CTO