【正文】 ance puting and core network. Third, support both fixed and floating point calculation. The trend keeps the same direction with the development of general processors. The first processors separated the arithmetic units and floating point units, but present ones integrate the two. As to the products available on the market, TMS320C66x serial multicore DSP produced by the Texas Instruments have the function of both fixed and floating point units. Ding Gang, Business Development Manager of Semiconductor Division, Texas Instruments said: “with the same high performance, unified fixed and floating point support is more convenient for programming. TI has pleted it, and we believe that the industry will follow the step soon. Zhangtie Hu, ADI39。s Technology and Market Manager agrees: in the next few years, we can see DSP develops toward the 32bit DSP direction, while supporting integer and floating point arithmetic instruction set. Fourth, singlechip DSP. Singlechip DSP is also known as DSC (digital signal controller, digital signal controller), which is the concept of SOC. It can reduce overall board cost, power consumption and size through integrating more functions and interfaces with advanced technology, including a variety of simulation Digital onchip peripherals, hardware acceleration and generalpurpose processing core. TI 39。s DaVinci platform is a typical example of a SOC, in addition to Freescale39。s MC56F82xx series. MC56F82xx is a member of DSC family designed on the basis of 56800E core. It bines microcontroller (MCU) function and digital signal processor (DSP) processing power and providing a wide range of peripherals, including PWM device with NanoEdge highspeed layout techniques and two ultrahighspeed analog (ADC). Charlie Wu, senior systems engineer of Freescale Microcontroller Division said: The DSC will be widely used in the field of new energy, such as digital power, solar power and the management of battery charging of hybrid electric vehicles. Fifth, DSP + MCU, which is the integration of microcontroller. Low cost microcontroller is the main generalpurpose processor for the task of smart directional control. Although it works well, its digital signal processing is relatively poor. DSP is just to pensate for this dysfunction. Many applications require both intelligent control and digital signal processing, therefore, the bination of DSP and microprocessors in a single chip will greatly accelerate development of the personal munications products, smart phones and wireless networking products, thus simplifying design, reducing PCB size, reducing power consumption and lowering overall system cost. “32bit MCU integrating DSP instruction is the development direction. This development is out of the cost requirements. Many analog control circuits will be replaced by digital control (software). Charlie Wu fromFreescale39。s, explained “generally speaking, the 32 bit MCU will increase multiplyaccumulate instructions of DSP, but only integer instructions. As a result, MCU also has some DSP functions and can calculate be some of the less demanding realtime filter. MCU39。s main structure is the single bus architecture, while it has some Harvard bus architecture, that is, a program bus plus a set of data bus. The nest step of MCU is toward the Harvard bus architecture. However, limitation of this bus structure makes it less strong than DSP in realtime digital signal processing. The earliest platform with both the DSP and MCU functions was used in the field of engine control, and then it extended to voice processing, sensor processing and other applications, replacing the original digital analog filter. Today, this platform is more and more applied to the puter, telephone equipment and Ethernetrelated areas, while also in the medical care, electrical appliances, air conditioners, uninterruptible power supply, switching power supplies and semiconductor lighting. When the nextgeneration of MCU with DSP capabilities is available, the difference between entrylevel DSP and general MCU will bee smaller and not clear. In the foreseeable period of time, wireless applications will continue to be marketdriven engine of programmable DSP, and the embedded DSP market is more potential. Whatever form, DSP technology is still a hub for the emerging IP telephony market. Without DSP technology, the access to Internet and multimedia is impossible, so is the wireless munications. Therefore, DSP industry will continue to be the driving force of semiconductor industry . 參考譯文 DSP未來(lái)發(fā)展五大趨勢(shì)窗體底端在4G無(wú)線通信領(lǐng)域，數(shù)據(jù)吞吐量已經(jīng)達(dá)到了3G時(shí)代的100到1000倍；在多媒體處理器領(lǐng)域，每天都有新的標(biāo)準(zhǔn)涌現(xiàn)出來(lái)；在更多其它DSP應(yīng)用領(lǐng)域，密集的數(shù)據(jù)處理器需求不斷給處理器設(shè)計(jì)團(tuán)隊(duì)帶來(lái)新的挑戰(zhàn)。隨著更多新應(yīng)用的更高要求，未來(lái)的DSP發(fā)展將呈現(xiàn)五大發(fā)展趨勢(shì)： 第一，DSP內(nèi)核指令并行處理能力增強(qiáng)，SIMD(單指令多重?cái)?shù)據(jù))和VLIW(極長(zhǎng)指令字)將會(huì)在新一代高性能處理器中占據(jù)主導(dǎo)地位?，F(xiàn)在，幾乎所有的DSP廠商都使用了VLIW構(gòu)架，這一指令構(gòu)架大大加強(qiáng)了DSP處理器的指令并行能力，而SIMD指令集能夠大幅度提高數(shù)據(jù)并行處理能力，在通用處理器上也得到了大量使用。例如，CEVA最新的XC323內(nèi)核就包括8路VLIW、512位SIMD操作、每周期32次MAC乘加運(yùn)算，以及固有的復(fù)雜算術(shù)運(yùn)算支持。這是CEVA針對(duì)4G應(yīng)用推出的全新構(gòu)架，直接針對(duì)高吞吐量高性能的應(yīng)用需求。 第二，在某些苛求高性能的應(yīng)用中，多核DSP方案會(huì)越來(lái)越多。這點(diǎn)得益于半導(dǎo)體制造工藝的發(fā)展，單顆芯片上可以集成越來(lái)越多的晶體管，也符合處理器整體多核化的趨勢(shì)，因?yàn)榭紤]到功耗的原因，單一內(nèi)核通過(guò)提高主頻的方式已經(jīng)難以為繼。依靠多核并行處理提高性能，已經(jīng)成為整個(gè)行業(yè)的選擇。例如德州儀器的C66x高性能多核DSP系列，最高可集成八個(gè)C66x內(nèi)核，總共可達(dá)到320 GMAC和160 GLOP定點(diǎn)和浮點(diǎn)性能，非常適合醫(yī)療和高端成像、測(cè)試和自動(dòng)化、高性能計(jì)算和核心網(wǎng)絡(luò)等高端應(yīng)用。 第三，同時(shí)支持定點(diǎn)和浮點(diǎn)計(jì)算。該趨勢(shì)與通用處理器的發(fā)展方向保持一致，最早處理器也是整數(shù)運(yùn)算單元和浮點(diǎn)運(yùn)算單元分離的，現(xiàn)在的處理器都集成了整數(shù)和浮點(diǎn)計(jì)算單元。從目前面世的產(chǎn)品來(lái)看，德州儀器的TMS320C66x系列多核DSP已經(jīng)融合了定點(diǎn)和浮點(diǎn)功能。德州儀器半導(dǎo)體事業(yè)部業(yè)務(wù)拓展經(jīng)理丁剛表示：“在保證高性能的同時(shí)，統(tǒng)一定點(diǎn)和浮點(diǎn)支持更便于編程。TI已經(jīng)完成了這點(diǎn)，我們認(rèn)為業(yè)界也會(huì)跟進(jìn)。”ADI的技術(shù)市場(chǎng)經(jīng)理張鐵虎對(duì)此表示認(rèn)同：“今后幾年，我們可以見到DSP向著32位的方向發(fā)展，同時(shí)支持整數(shù)和浮點(diǎn)算術(shù)指令集?！?第四，單片DSP。單片DSP又被稱為DSC(digital signal controller，數(shù)字信號(hào)控制器)，這也是SOC的概念，通過(guò)先進(jìn)技術(shù)集成更多功能和接口來(lái)降低整體板級(jí)成本、功耗和大小，包括各種模擬數(shù)字片上外設(shè)、硬件加速、通用處理內(nèi)核。德州儀器的達(dá)芬奇平臺(tái)就是一個(gè)SOC的典型例子，此外還有飛思卡爾的MC56F82xx系列。MC56F82xx是基于56800E內(nèi)核DSC系列的成員，它結(jié)合了微控制器(MCU)的功能和數(shù)字信號(hào)處理器(DSP)的處理能力，提供廣泛的外設(shè)，包括帶有NanoEdge布局技術(shù)的高速脈寬調(diào)制器以及兩個(gè)超高速模數(shù)轉(zhuǎn)換器(ADC)。飛思卡爾微控制器事業(yè)部高級(jí)系統(tǒng)工程師Charlie Wu表示：“這種DSC會(huì)大量應(yīng)用在新能源領(lǐng)域，如數(shù)字電源、太陽(yáng)能發(fā)、混合動(dòng)力汽車的電池充電管理等。” 第五，DSP + MCU，也就是微控制器的融合。微控制器的成本低，是主要執(zhí)行智能定向控制任務(wù)的通用處理器，能很好執(zhí)行智能控制任務(wù)，但是它的缺點(diǎn)就是數(shù)字信號(hào)處理功能比較差，而DSP正好能彌補(bǔ)這一功能缺陷。許多應(yīng)用都需要兼具智能控制和數(shù)字信號(hào)處理兩種功能，因而，把DSP和微處理器進(jìn)行結(jié)合，用單一芯片實(shí)現(xiàn)這兩