|
|
|
|
|
|
 |
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
Lexra LX5280 Is Fastest RISC-DSP in BDTImark2000[tm] Composite DSP Speed Metric
Lexra LX5280 Is Fastest RISC-DSP in BDTImark2000[tm] Composite DSP Speed Metric
"We are very pleased not only at the result, but also the experience of working with the experts from Berkeley Design Technology Inc. (BDTI)," commented Charlie Cheng, president & CEO of Lexra. "The results of the benchmarks will no doubt influence our future DSP architecture, and make the LX5xxx RISC-DSP even more powerful." Today, most if not all, RISC CPUs have integrated MACs. However, this type of architecture does not provide true DSPs functionality or performance. RISC-plus-MAC machines lack functions such as fractional arithmetic, overflow/underflow protection, dual 16-bit arithmetic, post-modified pointer arithmetic with circular buffers, zero-overhead looping and other specialized DSP ALU instructions. Analyzing the results The BDTImark2000[tm] is a composite DSP speed metric score distilled from a processor's results on BDTI's 12 DSP benchmarks. Therefore, the score provides an excellent balanced view of the overall DSP speed of the processor. BDTImark2000[tm] scores are only issued for processors that have been demonstrated in silicon. Further information about the BDTImark2000[tm] is available on BDTI's web site, www.BDTI.com. The table of comparison roughly divides the groups into two: Long Instruction Word (LIW) and non-LIW DSPs. The LIW DSP architectures generally yield performance above 1,000 BDTImark2000[tm], while non-LIW DSPs are below. The LX5280 ties with Lucent's DSP16xxx as highest performance non-LIW DSP. Lexra believes that achieving a near-1,000 BDTImark2000[tm] gives the LX5280 an outstanding price and performance point for licensees to integrate a DSP core for running multiple algorithms. Applications requiring more than 1,000 BDTImark2000[tm] can almost always be channelized, and divided into multiple LX5280s for execution. The small die area in the LX5280 makes it feasible for Lexra licensees to use multiple instantiations to scale the performance. "Most applications requiring performance in the thousands of BDTImark2000[tm]s, are `channelized'-the same DSP tasks are being performed on many independent channels," said Pat Hays, chief technology officer and co-founder of Lexra. "For example, VoIP gateways require G.723/G.726/G.729 audio algorithms to be run on hundreds of channels. In these cases, a system-on-chip (SOC) design with multiple LX5280 cores achieve near linear performance versus area, while high-end DSPs are far less area efficient." The LX5280, therefore, combines high-end, non-LIW level DSP performance with the ability to run control software and support RTOS. This level of flexibility and performance makes it ideal for convergence products such as home infotainment chips, or integrated access devices (IAD). About the LX5280 The LX5280 is a RISC-DSP, which extends the MIPS®* RISC instruction set into DSP by adding hardware and ISA support for the DSP, Radiax[tm]**. Radiax[tm] includes dual 16-bit arithmetic, post-modified pointer arithmetic with circular buffers, zero-overhead looping, as well as a number of specialized DSP ALU instructions. The LX5280 is a dual-issue superscalar implementation of the Radiax[tm] extensions. Typically, for inner-loops, one 32-bit instruction will load a 64-bit register pair from data memory, and another will execute a SIMD MAC operation on register-based data. Lexra also offers a single pipeline implementation called the LX5180 for the most cost- and power-sensitive applications. Both the LX5280 and the LX5180 are supported worldwide by leading third-party development tool vendors. About the BDTImark2000[tm] The BDTImark2000[tm], developed by BDTI, is the most respected independent measure of DSP performance on actual DSP tasks. Because DSP algorithms are highly specialized, benchmarks oriented to general computing such as Dhrystone or Spec-Int are irrelevant. Raw speed - as measured by MIPS - is also irrelevant because of the wide variations in DSP ISAs. Millions of Multiply-Accumulate per second (MMACs) is a more useful "rule-of-thumb" but still a rough approximation of a DSP's performance. About Lexra Lexra, Inc. is a leading microprocessor developer specializing in 32-bit RISC, DSP and NPU cores for the embedded market. In less than four years, the company has established itself as an innovator in embedded microprocessor technology and intellectual property (IP) licensing business model, with proven track record for customer success. During this short period, Lexra has delivered nine different processors to over 30 licensees in six different countries. Its customers are among the top three network communication companies as well as the top ten semiconductor companies. Lexra is headquartered in San Jose, Calif. Further company information can be found at http://www.lexra.com. *MIPS, MIPS I, MIPS16, R3000, and other MIPS common law marks are trademarks and/or registered trademarks of MIPS Technologies, Inc. Lexra, Inc. is not associated with MIPS Technologies, Inc. in any way. Unaligned loads & stores are not supported in hardware or software. **Radiax is a trademark of Lexra Inc.
Footnotes For Immediate Release Lexra Contact: Press Contact: |
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
|
|
|
|
Home | Feedback | Register | Site Map |
|
|
|
All material on this site Copyright © 2017 Design And Reuse S.A. All rights reserved. |