Tensilica’s Diamond Standard and Xtensa processors are ideal for digital signal processing tasks.

Diamond Standard DSPs

Several Diamond Standard processors can be used as DSPs. Even the Diamond 212GP has basic DSP capabilities, with a single-cycle 16bitX16bit MAC unit adding four dedicated 32-bit registers and a 40-bit accumulator. Additionally, there is support for zero-overhead looping, clamps (saturating arithmetic), max/min value, normalize, and sign extend.

The Diamond 545CK is the highest performing licensable DSP core, according to BDTI. The Diamond 545CK, which combines a base CPU controller and a DSP containing 8 parallel 18-bit single-cycle MAC units, is unique in the industry, allowing system control and industry leading data processing throughput in a single core with a single compiler and single instruction stream. SIMD capability allows eight simultaneous MAC operations on independent data pairs per cycle, utilizing the 160-bit vector registers.

The Diamond 330HiFi is optimized for audio processing. Specialized audio instructions designed into the Diamond 330HiFi by Tensilica increase code density (reducing memory requirements) and reduce Mhz requirements (lowering power). Techniques used to define these custom instructions include Single Instruction, Multiple Data (SIMD - parallelization), VLIW (multiple operations per cycle), and flexible data path widths (optimum data widths to reduce power).

The Diamond 38xVDO family of video engines are optimized for multi-standard, multi-resolution video. Targeted at mobile handsets and personal media players (PMPs), these video subsystems are fully programmable to support all popular VGA and standard definition (SD, also known as D1) video codecs with resolutions up to 720x480 (NTSC) and 720x576 (PAL) including H.264 Main Profile, VC-1 Main Profile, MPEG-4 Advanced Simple Profile (ASP) and MPEG_2 Main Profile, each of which is available from Tensilica.

Xtensa LX2 Processors as DSPs

Some customers use Xtensa processors as classic single-MAC DSPs, while others push the performance limits. The sections below discuss how Xtensa processors are used in DSP applications. For additional information, please check the DSP sections in our Xtensa 7 and Xtensa LX2 product pages.

Classic Single-MAC Xtensa LX2 DSP

DSPs are increasingly used in SOC applications due to pressures to decrease costs and increase performance, especially in consumer markets. High computational requirements in applications such as cellular phones, disk drives, audio codecs, and others cannot be met with control CPUs. However, these DSP cores must maintain a small die area since they are used in cost sensitive markets.

Tensilica’s Xtensa LX2 is an excellent technology with which to base a small, efficient DSP. A typical DSP based on an Xtensa LX2 processor would include two load/store units with separate X-Y data, zero-overhead looping, and other basic DSP instructions. These features are included with the base CPU functionality of the Xtensa LX2, which means a single processor core can function as both an efficient control CPU and a high performance DSP utilizing a single development tool environment.

Configuration Requirements

• 60K-70K gates, not including memory
• Simultaneous X-Y data fetching
• Basic DSP instructions
• 200-300Mhz worst-case frequency

Xtensa Benefits

• Core can run control code or DSP intensive application
• Single application development tool set
• Much smaller die size than competitors
• Easier application development than heterogeneous core system

Typical Application: DSP in Bluetooth SOC

High-Performance DSP

Extremely high performance DSPs are used in areas such as wireless infrastructure, high-definition video, and high-end imaging. These types of processors commonly execute multiple MACs per processor cycle. Additionally, some have the capability of operating in single-instruction, multiple data (SIMD) mode. Similar to high-performance CPU control processors, the complexity of DSPs has increased tremendously.

Tensilica’s Xtensa LX2 can easily outperform other licensable DSPs using VLIW and SIMD techniques. Utilizing Tensilica Instruction Extensions (TIE), Xtensa LX2 has achieved the highest scores ever on these benchmarks.

Xtensa Configuration Highlights

• Highest performing DSP, featuring the Vectra LX DSP engine
• 200K-300K gates, not including memory
• 4 to 8 MACs per cycle
• Over 300Mhz worst-case frequency

Xtensa Benefits

• Performance headroom allowing more software application flexibility
• Techniques such as VLIW and SIMD allow more processing done per cycle
• Much smaller die size than other high performance DSPs
• Flexible architecture, user may target many applications for performance improvement

Vectra LX DSP Engine Block Diagram

Read more about Xtensa LX2 DSP capabilities in our Products section.