Part of the  

Chip Design Magazine


About  |  Contact



This is not your father’s MCU

A lot can happen in four years in the embedded design world, and ARM has responded with the introduction of the ARM® Cortex®-M7 processor, following relatively shortly after the 2010 introduction of the Cortex-M4.
By: Caroline Hayes, Contributing Editor

ARM’s Cortex-M4 introduced DSP into the microprocessor, but according to Richard York, Vice President, Embedded, ARM, the subsequent Cortex-M7’s performance improvement needed a different approach. “The Cortex-M7 is a little less constrained about silicon,” he explains, “The performance has been improved [from the Cortex-M4] but brings in DSP and floating point performance at a similar cost point and a similar silicon area.”

To illustrate this, it is interesting to note that the Cortex-M7 DSP is the same size and the same instruction set as the Cortex-M4, but is faster. The Cortex-M7 DSP is increased to meet sensor fusion and control operations—two characteristics important in the growing IoT market. It operates at 400MHz, compared to 168MHz achieved by the Cortex-M4.

The ARMv7E M six-stage pipeline architecture increases throughput compared to the Cortex-M4 which uses the three-stage pipeline ARMv7E M architecture with the Thumb/Thumb 2 instruction set. The Cortex-M7’s superscalar pipeline allows the processor to execute more instructions/second. “This allows the traditional manufacturer to use a single core for control and DSP functions, which reduces the cost of the design,” adds Bee Hayes-Thakore, Product Marketing Manager, CPU, ARM.

The Cortex-M7 is less silicon-constrained than the Cortex-M4, with the same DSP and instruction set but with a performance uplift.

Figure 1. Cortex overview image: The Cortex-M7 is less silicon-constrained than the Cortex-M4, with the same DSP and instruction set but with a performance uplift.

Ian Johnson, Product Manager, Cortex-M7, ARM, is keen to point out that Cortex-M7 is an evolution and not a replacement for the Cortex-M4. “It is a high performance addition to the Cortex-M processor family – not a replacement for the Cortex-M4”. It has some shared characteristics – i.e. scalable architecture – with some ‘performance uplift’, he says. Hayes-Thakore elaborates: “When the Cortex-M4 was launched, it was in response to demand at the high end, but that bar has significantly changed with always-on and always-aware status devices, and with local processing capability. Partners wanted and demanded more, but with compatibility. The Cortex-M7 is really an extension of the Cortex-M4. It is designed for endpoints in automotive, Internet of Things and portable medical applications that will be expected to deliver 30 years’ of operation.”

The memory interface of the Cortex-M7 allows access to the internal cache for efficiency.

Processing: “More Better”
Higher processing requirements on microcontrollers in automotive and industrial automation are just two examples of why the performance uplift was developed for in the Cortex-M7. Factory floors require an increasing amount of precision and operate on large amounts of data in a short space of time. These application areas also need a good user interface, whether it is in a vehicle dashboard or a factory automation system. This allows the processor to address the fast interrupt controls and processing in almost real-time, and to allow autonomous decisions by the driver or operator, both driving factors, according to Johnson. “The application area is very wide in automotive and industrial control; initial licensees may be creating general purpose processors, but these will be used anywhere.”

He goes on to elaborate that although it may be viewed as a general purpose processor, it is by no means confined to that role. “The Cortex-M7 may be described as general-purpose but it is also used on other applications. . . For example, a Cortex-A partner will see the benefit in using the Cortex-M7 as a companion processor,” he says. “In that setting, it will be an invisible companion to the Cortex-M processor.”

The six-stage pipeline delivers a performance of 2.14DMIPS/MHz, improving the Cortex-M4’s 1.25DIPS/MHz, to fulfil the capability requirements that are normally only seen at the high end of the market. The increase in instructions per cycle have led to a modest improvement in clock rate, says York, for twice the number of instructions per cycle.

The six-stage pipeline increases performance to 2.14DMIPS/MHz to deliver ‘new wave’ capabilities.

“The first devices are on the market and are as energy-efficient as the Cortex-M4,” confirms York. The ability to process audio data and still image data due to the increased processing power addresses a cross-section of customers who are using makeshift solutions [such as a two-core controller and a DSP configuration] for these high-end functions. According to Hayes-Thakore, the single, scalable processor removes the barrier of adapting existing options and also saves development costs.
Another change is the Cortex-M7 memory interface. It is the first Cortex product to integrate the instruction cache and the data cache. The integration reduces the power consumption of the system and allows engineers to execute a large proportion of code from the internal cache to reduce the number of read and write occurrences from the external memory, leading to the power savings.

“This also allows efficient access to a large external memory,” says Johnson. “For example, a large DDR memory with large images to display and audio samples to play. Accessing the internal cache is more efficient.”

All Cortex-M7 processors have a floating point unit, which distinguishes it from other cores. It has the same energy-efficiency modes as the Cortex-M4, points out York, to meet the new wave of capabilities, such as always-on, always-connected, voice processing and image processing. Both share sleep on exit mode, sleep and deep sleep modes. “The comparison we make,” says Johnson, “is that of the user expectation from mobile devices. The pervasiveness of phones and tablets has set user expectations for that experience in the car, and in work-related settings. The Cortex-M7 has been developed to meet those demands—and some we haven’t thought of.”

Energy management
As well as the increased demands of established application areas, such as automotive, factory automation, white goods and medical, the relatively new application of Internet of Things has been a driver for the Cortex-M7, or, as York, the VP, Embedded, expresses it “Adding connectivity to unconnected things,” meeting the increase in what customers are introducing as always-on design features.

Energy-efficiency is another shared feature, with the same power modes offered in the Cortex-M4 and Cortex-M7. Sleep mode, deep-sleep mode and power-down mode, as well as the ability to shut-down and wake up peripherals in a short period of time, contribute to power efficiency. “The Cortex-M architecture responds with minimal latency to interrupts and in waking up the processor and surrounding peripherals,” says Johnson. A capability that can be exploited by the ‘new wave’ of applications such as Internet of Things and wearables which require always-on operation but not at the expense of battery drain. York also comments on the power management for always-on applications: “Monitors, sensors and gyroscopes need to ‘fire up’ a display and can use the Cortex-M7 with wireless communications, displays and interfaces with the same energy efficiency and performance as the Cortex-M4.”

The evolution of the Cortex-M architecture increases performance capabilities without impacting energy efficiency.

In another example, today’s factory automation systems may have several nodes and endpoints across the plant floor. If these can be in deep sleep mode, together with the processor peripherals, to save power, yet wake when triggered by an event in a short period of time, this has advantages across systems using multiple nodes for energy efficiency without time-delay penalties.

Development tools
The shared capabilities between Cortex-M4 and Cortex-M7 processors make it only logical that the development tool legacy is preserved. In addition to ARM’s own DS-5 and Keil® MDK 5 development tools, third party tools, software and RTOS from Atollic, DSP Concepts, Express Logic, FreeRTOS, Green Hills, IAR Systems, Lauterbach, Mentor Graphics, Micrium and SEGGER can also be used across the two processor environments and across the whole Cortex-M family.

As well as preserving the legacy investment of a single software tool, this also reduces the training environment and allows expertise from earlier designs to be reused and code from other Cortex-M processors to be used without modification.

The investment for the future with Cortex-M7 using skills and innovations from earlier generations reinforces the assertion in the title—this is not your father’s MCU.

October 2014

Tags: , , , ,

One Response to “This is not your father’s MCU”

  1. Renuka Says:

    So true for embedding sustainability aspects right from design mode.

Leave a Reply