Abstract ARMv7 is the present ARM processor architecture highest edition. This article introduced based on ARMv7 newest ARM the microprocessor Cortex series, described this series processor’s technical detail, simultaneously elaborated Cortex-R, Cortex-A, the Cortex-M series processor’s technical characteristic with emphasis. Finally, has made the further elaboration to the different series processor’s application domain.
Key word ARMv7 architecture Cortex-M Cortex-R Cortex-A
Introduction
Along with the embedded technical application domain’s unceasing expansion, is getting higher and higher to embedded system’s request, but takes the embedded system core the microprocessor also to face day by day the severe challenge. ARM Corporation since was established, by the intellectual property rights (IP, Intelligence Property) the tenderer status sell intellectual property rights, development domain has made the breakthrough unceasingly in 32 RISC in the CPU, its design’s microprocessor structure already developed from v3 to present’s v7. The Cortex series processor is based on the ARMv7 construction, divides into Cortcx-M, Cortex-R and the Cortex-A three kinds. The ARM series microprocessor’s core and architecture like table l arranges in order.
1 ARM Correx processor technology characteristic
The ARMv7 construction is born in the ARMv6 construction foundation. This construction has used the Thumb-2 technology, it is develops in the ARM Thumb code compress technique foundation, and maintained to the extant ARM solution complete code compatibility. The Thumb-2 technology little uses 31% memories compared to the pure 32 codes, reduced the system expenses, simultaneously can provide compared to is higher than 38% performance existing based on the Thumb technology solution. The ARMv7 construction also uses Ding the NEON technology, enhanced nearly 4 times DSP and media handling ability. And supports the improvement the floating point calculation, satisfies the next generation 3D graph, the game physics application as well as the traditional embedded control application demand. In addition, ARMv7 also supports the improvement the movement environment, caters to JIT which increases unceasingly (Just In Time) and DAC (DynamicAdaptlve Compilation) the technical use.
With early ARM processor software compatibility but only then, ARMv7 construction in design time considered fully. ARM the Cortex-M series supports the Thumb-2 set of instructions (Thumb set of instructions expansion collection), what may carry out all has saved is the early processor compilation code. Before one to the conversion mode, is the user code which ARM the Cortex-M series processor writes may series microprocessor be completely compatible with ARM the Cortex-R. ARMCortex-M series system code (truthfully when operating system) may transplant very easily to based in the ARMCortex-R series system. ARMCortex-A and the Cortex-R series processor also supports the ARM 32 sets of instructions, backward the completely compatible early time’s ARM processor, including issued the ARM7TDMI processor which issued from 1995 the ARMll processor series which to 2002. Figure 1 is the v5~v7 construction processor technology comparison. As a result of the application domain’s difference, the technology which processor series institute uses based on v7 the construction Cortex is not same.
In the naming way, processor already no longer extended based on ARMv7 the construction ARM with the digital naming way, but is the crown shouts by the Cortex generation. Based on v7A named “Cortex-A is the state”, based on v7R named “Cortex-R series”, based on v7M named “Cortex-M3″.
2 ARM CorteX-M3 processor technology characteristic
ARM the Cortex-M3 processor is develops specially for the memory and processor’s size to the product cost influence enormous each kind of application has not counted, its structure as shown in Figure 2. Its conformity many kinds of technologies, reduced the use memory, and provides the low power loss and the high performance in the minimum RISC essence, may realize by the former code to 32 micro controller’s fast transplants. ARM the Cortex-M3 processor is use least gate number ARM CPU, was opposite reduced the chip area greatly in the past’s design, might reduce the installment the volume or uses the low cost the craft to carry on the production, the only 33000 essence performance might reach l,2DMIPS/MHz. In addition, the basic system peripheral device also has highly the integrated characteristic, integrated many close coupling system peripheral device, has used the chip space reasonably, causes the system to satisfy the next generation product the control demand.
ARM the Cortex-M3 processor unified has carried out the Thumb-2 instruction 32 Harvard element lash construction and the system peripheral device, including Nested Vectored Interrupt Controller and Arbiter main line. This technical program in tests and in the example application displays the high performance: Under the Taiwan mechanical and electrical 180 nm craft, the chip performance reaches 1.2 DMIPS/MHz, the clock rate reaches as high as l00 MHz. The Cortex-M3 processor has also realized the Tail-Chaining interrupt technology. This technology is one item completely based on hardware’s interrupt processing technology, most may reduce 12 clock periodicities, may reduce 70% interrupts in the practical application; Has promoted the new single track debugging technique, avoids using the multi-pins to carry on the JTAG debugging, and supports the RealVicw compiler and the RealView debugging product comprehensively. The Realview tool provides the simulation, the foundation virtual model to the designer, to translate the software, the debugging, the confirmation and the test based on ARMv7 construction functions and so on system.
Cortex-M3 which develops for the micro controller application has the following performance:
◇ realizes the monocycle Flash application optimization;
◇ accurately fast interrupt processing. Never surpasses for 12 cycles, only 6 cyclical tail-chaining (end chain-like);
◇ has the low power loss clock gating (Clock Gating) 3 kind of sleep patterns;
◇ monocycle multiplication and multiplication accumulation instruction;
◇ARM Thumb-2 mix 16/32 inherent set of instructions, non-schema translation;
◇ including data viewpoint and Flash patch high-level debugging function;
◇ the atomic position operation, reads/the revision/compilation in a sole instruction;
◇1.25DMIPS/MHz (compares with 0.9DMIPS/MHz ARM7 and 1.1DMIPS/MHz ARM9).
3 ARM Cortex-R processor technology characteristic
ARM the Cortex-R series processor includes ARM CortexR4 and ARM at present the Cortex-R4F two models, is mainly suitable for real time system’s embedded processor.
3.1 ARM Cortex-R4 processor technology characteristic
Cortex-R4 processor structure as shown in Figure 3. This processor support handset, the hard disk, the printer and the automobile electron design, can assist the new generation embedded product to carry out each kind of complex control algorithm and the real time operation operation fast; May through the memory protect unit (MPU, Memory Protection Unit), the high speed buffer as well as the tight bond memory (TCM, Tightly Coupled Mennory) lets the processor make the optimization adjustment in view of each kind of different embedded application, and does not affect the basic ARM set of instructions compatibility. This kind of design can in extend with in the original procedure code situation, reduces system’s cost and the order of complexity, simultaneously its tight bond memory function can also provide a smaller specification and the higher efficiency conformity, and brings when the fast response heard.
The Cortex-R4 processor uses the ARMv7 architecture, enables it with the existing procedure maintenance complete recollection compatibility, can support nowadays establishes from all over the world several billions systems; And has carried on the optimization design in view of the Thumb-2 instruction. This characteristic brings many benefits, including: The lower clock speed brings province electricity benefit; A higher performance leads each multi-purpose characteristic the mobile phone and the automobile product design; The more complex algorithm support high performance’s digital phantom constructs hard disk’s system with. Using the Thumb-2 set of instructions, the RealView development suite, causes the chip interior memory’s capacity to be most in addition can reduce 30%, reduces the system cost largely, its speed compares the Thumb set of instructions which uses in the ARM9tt6E-S processor to be higher than 40%. Because the memory takes the space in the chip to be getting more and more, therefore this design will save the chip capacity largely, lets the chip manufacturer utilize this model of processor to develop each kind of SoC (System on a Chip) the component.
Compares in the first several generation of processors, the Cortex-R4 processor high efficiency design proposal, enables its to achieve a higher performance by a lower clock; After optimization design Artisan Mctro memory, then further reduces embedded system’s volume and the cost. The processor embarks an advanced micro construction, has the double instruction transmission function, uses the 90nm craft and matches Artisan the Advantage library of subroutines’s module, the floor space does not arrive at lmm2, consumes the electricity most to be lower than 0.27mW/MHz, and can provide surpasses 600 DMIPS the performance.
The Cortex-R4 processor joins the fault-tolerant function and the memory protect mechanism in each kind of safe application, supports the most update OSEK real-time operating system; Supports RealView the Develop series software development tool, RealView Create the series ESL tool and the module, as well as Core Sight debugs with the tracing technology, assists the designer to develop each kind of embedded system rapidly.
3.2 ARM Correx-R4F processor technology characteristic
Cortex-R4F processor structure as shown in Figure 4. This processor has each advanced function which develops in view of the auto market, including debugs the function, to be possible automatically the interconnection the wrong detection mechanism, as well as may choose the floating point calculation unit which optimizes (FPU, Floating-Point Unit). The ECC technology can monitor the memory deposit work, detects and adjusts each kind of mistake. When has the memory mistake, the ECC logic wrong and stops the system operation besides the circular, but can also perform to adjust. It also has Cortex-R4 series each advanced function, can penetrate the high efficiency memory protect unit, the high speed buffer, as well as the tight bond memory, enables the processor to aim at each kind of different application to make the optimization adjustment; Meanwhile extends the traditional processor’s in wrong detection function to entire SoC, the system will scan unceasingly formerly detected the wrong material, lift system’s reliability. Based on to safety performance’s value, the Cortex-R4F processor embarked the high resolution memory protect mechanism specially, can the close control independent software work.
In the Cortex-R4F processor carries out floating point calculation FPU, provides exceeds the fixed decimal point operand the dynamirange and accurate. This FPU and between ARM other processor core’s FPU maintain with traces the compatibility, and aims at each kind of automobile application common single precision processing work to carry on the optimization. Use single precision form, but non-double precision value material, not can only the data processing speed promotion to 2 times, be able to maintain the essential precision to enhance the SoC design the efficiency.
The Conex-R4F processor uses a set to have the double instruction transmission function advanced micro construction, by ARM Artisan in the Advantage library of subroutines in view of the 90nm craft’s optimization, achieves surpasses 800 DMIPS the performance standards. The Levcl 1 memory loose succession design, enables the module to use the high density, low power loss RAM, causes holds the high proportion memory in the total cost, can look like the processor logic to have equally saves spatial the superiority. Under 90 nm crafts, takes spatial less than 1 mm2, and power consumption less than 0.27 mW/MHz, assists the system exploiter to reduce the cost and the power loss effectively. This processor uses the ARMv7ISA construction, can with already the ARM procedure code maintenance recollection compatibility which is used by the global several billions systems, and aims at the Thumab-2 set of instructions to carry on the optimized design. Using the ThLtmb-2 instruction, matches ARM the RealView development suite, enables the chip interior memory’s capacity most greatly to reduce 30%, reduces the system cost largely. In addition, compares Yu Neijian in ARM946E-S processor’s old version Thumb set of instructions, the new technology can enhance 40% efficiency.
4 ARM Cortex-A8 processor technology characteristic
ARM the Cortex-A8 processor is one section is suitable for the complex operating system and the user application application processor, its structure as shown in Figure 5. The support intelligence energy management (IEM, Inte11igent Energy Manger) technical ARM the Artisan storehouse as well as the advanced divulging control technology, caused the Cortex-A8 processor to realize the extraordinary speed and the power loss efficiency. Under the 65nm craft, ARM the Conex-A8 processor’s power loss does not arrive at 300mw, can provide the high performance and the low power loss. It first time for the low expense, the high-capacity product has brought the table model machine rank performance.
The Conex-A8 processor is the first section based on the next generation ARMv7 construction application processor, used has been able to bring the high performance, a lower power loss and the higher code density Thumb-2 technology. It has used the formidable NEON signal processing expansion collection for the first time, is media and so on H.264 and MP3 arranges the decoding to provide the acceleration. The Cortex-A8 solution also includes the Jazelle-RCTJava acceleration technology, translates (DAC) to real-time (JTT) and the dynamic alignment to provide the optimization, simultaneously reduces the memory to take the space to reach as high as 3 times. This processor has disposed the advanced exceeding the allowed figure quantity system structural flow water line, can simultaneously carry out many instructions, and provides surpasses 2.ODMIPS/MHz the performance. The processor integrated an adjustable size second-level cache memory, can works together with high speed 16KB or the 32KB first-level cache memory, thus achieves the quickest reading speed and the greatest volume of goods handled. The new processor also disposed has used in trading and number space copyright management Trust safely the Zone technology, as well as realized the low power loss management IEM function.
The Cortex-A8 processor has used the advanced branch forecasting technology, and has the special-purpose NEON trueing and the floating point assembly line carries on the media and the signal processing. In the use is smaller than the 4mm2 silicon chip and in the low power loss 65 nm craft situation, the Cortex-A8 processor’s movement frequency will be higher than 600MHz (not including NEON tracing technology and second-level cache memory). In high performance 90nm and under the 65nm craft, the Cortex-A8 processor movement frequency is highest may reach 1GHz, can satisfy the high performance consumption product design the need.
Conclusion
These new ARM the Conex processor series is based on the ARMv7 construction product, looking from the size and the performance aspect, also has is short in 33 000 gate ARM the Cortex-M series, also has the high performance ARMCortex-A series. And, ARM the Cortex-A series aims at grows day by day, moves including Linux, Windows CE and the Symhian operating system’s consumer entertainment and the wireless product design; What ARM Cortex-R series in view of is needs to move the real-time operating system to carry on the control application the system, including automobile electron, network and phantom system; ARM the Cortex-M series is sensitive for these to the development cost at the same time the embedded application which increases unceasingly to the performance requirement (for example micro controller, automobile body control system and each kind of large-scale electrical appliances) designs. Along with in each kind of different domain application demand’s increase, the microprocessor market also tends the diversification. In order to adapt the market development change, processor series will develop own application domain unceasingly based on ARMv7 the construction ARM.
