Oak DSP moves image processing to peripherals
Oak DSP moves image processing to peripherals
By Anthony Cataldo, EE Times
October 10, 2001 (5:31 a.m. EST)
URL: http://www.eetimes.com/story/OEG20011008S0041
SAN JOSE, Calif. Oak Technology Inc. has fashioned a highly parallel digital signal processor for a new breed of digital imaging systems that work independently of a PC. With the help of an ARM9 processor core, the company has created a system-on-chip platform with supporting software development tools that it hopes will displace hardwired ASICs. The Quatro DSP was built specifically for processing data generated from image sensors such as charge-coupled devices. Applications such as scanners and printers often use digital signal processing, or equivalent hardwired algorithms, for operations like gamma correction, color shading, color-space conversion and compression, and other functions usually outside the realm of an embedded control processor. For this purpose, Oak chose a single-instruction, dual-issue processing engine, but extended the architecture so that it has four parallel data paths in the pipeline. In this way, the DSP can eff ectively process data from four pixels simultaneously. "Other people in the industry may be doing more traditional DSPs with one or two multiply-accumulate functions or VLIW [very long instruction words], which can do several instructions at the same time but not operate on multiple pixels in parallel," said Roger Pennington, vice president of engineering for Oak's imaging group (Sunnyvale, Calif.). Part of the instruction Another feature, already used in other Oak DSPs, is an extraction/insertion pipeline stage, which pulls out partial word values coming off the A/D converter from the CCD and puts them back into 32-bit form. Usually that would require a longer sequence of mask, shift and arithmetic steps; here, it's part of the instruction pipeline. "We still store [data] in 32-bit words but we don't do data alignment and special operations; it's part of the instruction," Pennington said. "And the hardware can keep track of where it's at." To keep the processing elements runn ing efficiently, Oak engineered the system-on-chip architecture so that the inner loops running on the DSP operate separately from the outer control loops that run on the ARM9. The DSP will perform 1 billion multiply-accumulate operations every second, Pennington said. For these inner-loop operations to really shine, programmers will have to hand-code the DSP using assembly language. The good news, said Pennington, is that the assembly code is based on an algebraic language that is similar to a C syntax. "It's assembly language but it reads like C," Pennington said. "And we can help customers map their algorithms into the Quatro DSP." Moreover, Oak is providing a special Windows-based simulator for the DSP inner loops so that designers can look at the results and estimate the performance while running. There's also a simulator for software running on the ARM processor, which will handle housekeeping protocols, data movement, interfaces, and motor and CCD control, he said. The Quatro architec ture is built around the Virtual Component Interface bus, an industry-standard on-chip bus. A converter allows the bus to interface with ARM's Amba CPU bus. Pennington said Oak chose the VCI bus because it allowed direct connectivity between different modules. The architecture takes aim at a new class of imaging devices like scanners and inkjet multifunction peripherals, which can print, scan, copy and fax documents. These products are designed to work independently from a PC, though often they work slowly because of their lack of processing horsepower, Pennington said. It's estimated that the market for "PC-independent" imaging devices will grow from 11 million units in 2000 to 44 million units by 2004, according to market research firm International Data Corp. Oak will add special hardware features for volume customers but in general the company wants to avoid providing design services and concentrate on building standard devices, Pennington said. The company hopes that with the added performa nce of the DSP/ARM9 combination, there will be little need for extra hardwired gates. And Oak said its devices will be price competitive with any ASIC solution, even without considering nonrecurring-engineering charges. Oak plans to deliver first samples of devices based on its Quatro architecture by the first quarter of 2002, and will begin general sampling the following quarter. Pricing has not been determined.
Related News
- VeriSilicon VIP9000 and ZSP are Adopted by iCatch Next Generation AI-powered Automotive Image Processing SoC
- General Processor Technologies Announces AI Accelerator and DSP for Digital and Image Processing
- KYOCERA Document Solutions Uses Synopsys' Application-Specific Instruction-Set Processor Tool to Accelerate Design of High-Performance Image Processing DSP
- Almalence Ports Optimized Image Processing Software onto Tensilica's New IVP Imaging/Video DSP
- Xylon's logicBRICKS HDR ISP IP Suite Gets an RGB-IR Image Processing Upgrade
Breaking News
- TSMC drives A16, 3D process technology
- Frontgrade Gaisler Unveils GR716B, a New Standard in Space-Grade Microcontrollers
- Blueshift Memory launches BlueFive processor, accelerating computation by up to 50 times and saving up to 65% energy
- Eliyan Ports Industry's Highest Performing PHY to Samsung Foundry SF4X Process Node, Achieving up to 40 Gbps Bandwidth at Unprecedented Power Levels with UCIe-Compliant Chiplet Interconnect Technology
- CXL Fabless Startup Panmnesia Secures Over $60M in Series A Funding, Aiming to Lead the CXL Switch Silicon Chip and CXL IP
Most Popular
- Cadence Unveils Arm-Based System Chiplet
- CXL Fabless Startup Panmnesia Secures Over $60M in Series A Funding, Aiming to Lead the CXL Switch Silicon Chip and CXL IP
- Esperanto Technologies and NEC Cooperate on Initiative to Advance Next Generation RISC-V Chips and Software Solutions for HPC
- Eliyan Ports Industry's Highest Performing PHY to Samsung Foundry SF4X Process Node, Achieving up to 40 Gbps Bandwidth at Unprecedented Power Levels with UCIe-Compliant Chiplet Interconnect Technology
- Arteris Selected by GigaDevice for Development in Next-Generation Automotive SoC With Enhanced FuSa Standards
E-mail This Article | Printer-Friendly Page |