HiFi 2 Audio DSP Product Brief
Cut DSP Development Time - Get High Performance From C, No Assembly Required
Optimizing a DSP Architecture for Wireless Baseband
A Designers Guide to HD Video Pre- and Post-Processing
Put Low-Power, Low-Overhead, High-Fidelity Digital Sound in Your Next ASIC or SOC
See our complete white paper library
Tensilica DSP Targets LTE Advanced - Microprocessor Report review of ConnX BBE64
Tensilica Plays Baseband - New ConnX Core Aims for Low-Power Wireless Communications - Microprocessor Report review of ConnX BBE16
Tensilica Xtensa LX Processor with Vectra LX - BDTI
The ConnX 545CK is one of the highest performance licensable DSP IP cores. The ConnX 545CK allows system control and industry leading data processing throughput in a single core with a single compiler and single instruction stream. It combines a base CPU controller with a DSP containing eight parallel 16-bit MAC units, allowing it to sustain eight simultaneous MAC operations on independent data pairs per cycle, utilizing the 160-bit vector registers.
As in all Xtensa ISA-based architectures, 16-, 24- and 64-bit VLIW instruction bundles are freely intermixed in the instruction stream with no processor mode switching to decrease performance. All software development tools (compiler, linker, debugger, instruction set simulator) have been enhanced to enable access to DSP-related and control hardware through standard C/C++ source code.
In addition to the data processing gains realized through parallel multiplier hardware, the ConnX 545CK includes support for other DSP-related operations, such as zero-overhead looping, clamps (saturating arithmetic), max/min value, normalize, and sign extend. The ConnX 545CK is not configurable - it is a standard core, ready to use.
The ConnX 545CK is an industry leading combination of an efficient 32-bit RISC controller and an extremely high-performance DSP in a single licensable IP core. It eliminates or greatly reduces the need to develop SOCs with separate control and DSPs.
The ConnX 545CK implements the Xtensa Instruction Set Architecture (ISA), a 32-bit architecture featuring a compact instruction set optimized for embedded designs. The architecture has a 32-bit ALU, 32 general-purpose physical registers, 6 special-purpose registers, and 80 base instructions, including compact 16- and 24-bit (rather than 32-bit) RISC instruction encoding. This reduces code size 10% to 40% compared to conventional RISC cores, depending on the actual code. Reducing code size results in higher performance, smaller memory sizes, and better power dissipation.
The Xtensa ISA's 16- and 24-bit encoding also provides powerful branch instructions, zero-overhead loops, and bit manipulations including funnel shifts and field-extract operations. This architecture is extended with 64-bit VLIW "bundles" - which are composed of multiple instructions-for the ConnX 545CKprocessor. Tensilica's XCC C/C++ compiler generates an optimized combination of 16-, 24- and 64-bit instructions in a single instruction stream, so there's no CPU mode switching between different types. Firmware developers do not need to decide which parts of their code should be compiler with different length instruction encodings.
A full-featured development environment - the Xtensa Xplorer - provides a graphical user interface (GUI) to all code development tools. The compiler toolchain and instruction set simulator (ISS) are available through the GUI in addition to performance modeling tools. Based on the Eclipse framework, Xtensa Xplorer allows developers to quickly evaluate code on the pipeline-accurate ISS and interface to emulation and hardware development boards. Xtensa Xplorer serves as the cockpit for the entire development .
Tensilica's XCC C/C++ compiler is an optimizing compiler that employs sophisticated multi-level optimizations to increase code execution performance and reduce code size. Also included in the Xtensa Xplorer environment are a software project manager, code profiling tools, source code editor, debugger, performance-modeling tool, the Xenergy energy estimation tool, the cache performance explorer, and a number of graphical visualization tools. Tensilica also provides both a C-based modeling environment called XTMP, as well as SystemC models of the ConnX 545CK DSP. For fast-functional simulation, Tensilica offers TurboXim for a 40-80x faster simulation than the ISS.
|
65GP
|
45GS
|
40LP
|
||||
| Speed Optimized | Area Optim |
Speed Optim | Area Optim |
Speed Optim | Area Optim | |
|
Area (mm2) post route
|
1.04 | 0.65 | 0.0688 | 0.395 | 0.705 | 0.403 |
|
Speed (MHz) post Prime Time
|
623 | 58 | 684 | 58 | 457 | 57 |
|
Power (mW/MHz) post route
|
0.129 | 0.076 | 0.082 | 0.045 | 0.113 | 0.064 |

ConnX 545CK is faster than every other licensable DSP core or CPU core tested by BDTI (higher is better)
Tensilica's ConnX 545CK was certified by Berkeley Design Technology, Inc. (BDTI) to have achieved a BDTI speed score of 4070. All processors benchmarked with 16-bit fixed-point data. All scores use worst-case clock speeds for the TSMC CL013G process and ARM Artisan SAGE-X library. Vendors can choose different speed/area/power trade-offs; to understand the trade-offs, please view all BDTI metrics for each core. BDTIsimMark2000TM scores may be based on projected clock speeds. For information, see www.BDTI.com/benchmarks.html Scores copyright © 2010 BDTI.
In addition, the ConnX 545CK achieves a BDTIsimMark2000TM per-mW score of 80. This is over 2x more energy efficient than any other core benchmarked by BDTI to date (as of July 2009).
