This blog is about everything to do with microprocessors. While we discuss architecture, tools, best practices, new innovations, performance, and system implementation, we also look at software, applications, and the larger industry trends. The processor market is a fascinating place today, and even though it has been around for more than 30 years, it is in many ways just getting started. We would love to hear from you on our blog posts, and please don't hesitate to let us know if you think we are wrong — it does happen occasionally.
About the Authors
At the age of 10 Mike begged his father to get him a computer. Never mind that at the time computers were the size of a large office and cost millions of dollars. Yes, Mike is no spring chicken and he didn’t get the computer, although his father did give him an abacus telling him that it would enable him to use the computer that he already had between his ears, which was not appreciated. Whether it was due to the trauma that resulted from using an abacus or just Mike’s love of anything electronic he has spent the last 30 years or so designing, building, and programming computers, microprocessors, and microcontrollers and developing applications that run on them. And his fascination continues with the definition of new processors and architectures in his search for the holy grail of computing: infinite performance at zero power consumption. Statistically speaking he is convinced it is just a matter of time.
Allen started in the ‘semiconductor IP industry’ before it was called the ‘semiconductor IP industry’. Back then, it was about ‘megafunctions’, ‘megablocks’ or MegaMacros™ (as trademarked by the pioneering UK IP company Allen was with… no, not that UK company). The biggest of these ‘mega’ things was an 8051! Today, of course, IP blocks are much larger and much more complex. And, it’s about the software, as well as the hardware. It’s also about working with a set of partners, sometimes called an ecosystem or community. Allen has been doing that for many years and is enjoying working with old and new friends on the ARC processor ecosystem.
And, they will be everywhere! The world is going to sensors in a big way. You currently interact with 50-100 daily. By 2020 this will be something like 1000 or so. Impossible you say. No, actually this is happening. Semico Research is estimating that 30 billion will be shipped in 2017. Welcome to your Brave New World.
Sensors are not new of course, they have been around and used in various applications for years. What is new is the level of processing that is taking place in sensors. The sensors of old were controlled by state-machines or 8-bit CPUs. In the past couple of years this has been changing rapidly because of the increasing complexity and merging of sensors to extract more meaningful information from the environment. This increased level of information has to be processed and this is being done with 32-bit processors that offer several orders of magnitude more performance than the 8-bit controllers of old. At the same time this new generation of 32-bit processor is comparable in size to 8-bit machines, so they offer the very low power consumptions and small areas that are required in sensors.
But, just having a small 32-bit processor isn’t enough. Because of their extreme low power and size limitations sensors are becoming highly optimized systems. Busing is eliminated. Hardware is merged into the processor pipeline. Peripherals are mapped into the processor with single cycle access. Specialized software and hardware functions are required. Efficiency at the system level is paramount. All of this takes longer to develop and as sensor systems get more complex, designers need an “SoC-Ready” sensor IP solution that helps them reduce their design and integration effort, lower their design risk and speed time-to-market.
Synopsys is bringing to market the industry’s first integrated, pre-verified hardware and software IP subsystem solution for a range of sensor control applications including smart sensors, sensor fusion and sensor hubs. Optimized to process data from digital and analog sensors for standalone applications, or for offloading a host processor, the subsystem enables more power efficient processing of sensor data. This complete IP subsystem solution consists of an ARC EM4 32-bit processor, the digital interfaces, analog-to-digital data converter (ADC) interfaces, hardware accelerators, a comprehensive software library of DSP functions and software I/O drivers. With Synopsys Sensor Subsystem designers can achieve a 40-60% savings on area and power while minimizing system latency because of the highly efficient system-level implementation that it offers. Even better the Sensor Subsystem is a full solution that is easily configured, so sensor systems can be implemented in hours and days, rather than weeks or months.
Applications in automobiles, mobile devices and the “internet of things” increasingly rely on sensors that give the ability to read and interpret environmental conditions such as pressure, temperature, motion, and proximity. The use of sensors will continue to increase as our connections into the digital realm become more seamless enabling us to routinely do things that seemed like science fiction only a few years ago.
There comes a time in every SoC designer’s life that the marketing guys start complaining that they are losing sockets to the competition, because the chip that you designed no longer has enough performance. Nothing lasts forever and this is especially true with electronics. What is state-of-the-art today is destined to be your Momma’s electronics, and sooner than you think. The performance demands for electronic applications increase at a constant rate. This is due to the combination of more stuff being added to applications over time, the convergence of functions from multiple devices into one device and the natural tendency that we all have to be less tolerant of slow functionality the longer we use a product. Oh yes, and the constant demand by marketing that engineering increase performance because they can’t think of anything else to do to beat the competition.
For 30 years 8-bit processors have been the workhorses of embedded applications. You no doubt interact with dozens of 8-bitters daily and probably don’t even know it. They are all around us but as the electronics in our world becomes more sophisticated 8-bit processors are becoming less and less able to keep up. To be sure, there will be 8-bitters used for many years – there are after all a few 4-bit processors that are still being used in some applications. But the time for 8-bit processors is passing as new 32-bit processors shrink to sizes that are comparable while offering an order of magnitude more performance, and advanced features that 8-bitters can only dream of.
Synopsys’ ARC EM family has been developed in part to hasten the exit of 8-bit processors. Starting at fewer than 10K gates the EM family is tiny, but at 1.52 DMIPS/MHz and clock speeds above 900MHz (28nm) it offers performance that can only be obtained from a modern 32-bit architecture. Even better while increasing performance the EM family processors can reducing power consumption, because they can do a lot more work per clock so they can be put to sleep for longer periods or slowed way down. They also have the flexibility to be configured to fit into the slot that the 8-bit processor is vacating. The memory configuration of the EM family is highly malleable – you can mix different types of memories and memories that are running at different speeds. The EM processors are also highly configurable so you can tailor them specifically for the instance on your SOC.
So, with all of the things that you have to worry about why are you still trying to make do with that old 8-bitter? Make your life easy and drop a 32-bit processor in that socket. You will be happy with the result and you will love not having to listen to the marketing guys complain about losing sockets because of the performance of your design.
Well of course that depends on what you are talking about. In the world of microprocessors size does matter. Not only does size equate to cost, but also to power consumption. Depending on the design that you are doing both can be major considerations. There can also be size considerations for the amount of memory that will be used with the processor making code density a critical parameter. If the only thing that you care about in a design is performance then size maybe doesn’t matter, but then again maybe it does. It used to be that I would have customers tell me that performance is all that mattered. I don’t hear that so much anymore – in fact it is rare.
For processors that are designed for embedded and deeply embedded applications size and power more often than not are the main requirements. That a processor can run at 500 MHz doesn’t matter so much if you are only planning on clocking it at 10 MHz, but size gets really important if the application is a sensor and you are planning to put it in a 2×2 mm package. As small as that is power consumption can be even more challenging. I have seen specifications for sensors where 100uW is the total power available for the whole sensor. When you consider that this level of power could easily be consumed by leakage it becomes apparent that the total sensor implementation has to be very small. This takes a special breed of microprocessors that are designed from the ground up to be implemented to minimum size and power, but also with excellent instruction performance to minimize the memory requirements.
With this in mind Synopsys developed the EM family of 32-bit processors. Designed for sensor and embedded control applications the cores start at fewer than 10K gates and can be quickly configured to a specific application, so that no gates are wasted. This is especially important at the older process nodes where many of the sensor and control applications are being implemented and each additional gate takes a lot more area and power.
The EM processor family has also been designed to reduce the memory footprint offering 20% better code density, which reduces power consumption and leakage. It is not unusual for the memory that is attached to a processor to take more area than the processor. A 20% reduction in the amount of memory needed for an application can be very significant, especially when area and power are limited.
Performance is still important in many applications, but more and more power and size are dominating processor selection criteria. Whether it is to increase battery life or reduce the operating cost of an application the size of a processor and the memory that it requires matters, and this will be increasingly true as we move forward.
Last March (yeah, time flies), I attended Sensorscon in San Jose. The event was insightful, with a wide variety of participants and speakers. The consistent theme of the speakers was that sensors will proliferate and broadly penetrate our lives. I saw some pretty big numbers there, such as a $19.5B market by 2016, and sensors in each smartphone approaching 20 by 2015. There are many driving factors behind these numbers, such as ubiquitous data collection through sensor networks, better health through lower cost consumer medical devices, and improved user experiences on our consumer electronics devices.
The recent explosion in the sensor market has not been driven by wide proliferation, but really, just a few device categories ramping to very high volumes with increasing sensor penetration per device. Mainly driven by, smartphones, gaming devices, and tablets. Sensor networks, medical devices, and really everything else, is peanuts.
What was interesting to me was the opportunity behind the opportunity. How many sensors does a phone need? Could a few really intelligent sensors replace the dozen plus used today? Consumers value an elegant user experience and advertisers value knowing where you are and what you’re doing, but does that mean more sensors? Or, could it be just smarter sensors? What was evident is that to date, very much a “brute force” approach has been taken by the industry. Brute force can get the job done fast, but it can also be unnecessarily expensive.
There were some early public indications of a smarter approach at Sensorscon. Start-ups dedicated to developing the software to get more from the hardware, and companies working together to establish software related standards. Not just sensor algorithms, but “smarts” through analysis, and in some cases, crunching through large amounts of data remotely, to get more out of each sensor. I also know first-hand that similar effort and investment is being made within device manufacturers. How do we do more (collect more data, improve user experience) at lower cost (e.g. fewer smarter, not more, sensors).
Reducing the number of unique sensors and finding ways to still add user functionality will be the job of software engineers and ultra-efficient processors. Processing the potentially massive amounts of data, in real time for some applications, has become the domain of integrated 32-bit processors, like the Synopsys ARC EM processor. The processor needs to be small and consume almost no power and generate almost no heat. The EM family does this while fitting into spaces as small as 0.01 mm2 and consuming as little as 2 uW/MHz.
But that’s not enough. This will be the domain of software engineers. Pervasive and easy to use development tools with highly efficient compilers are also important to embedded developers. Synopsys ARC EM processors are supported by the popular Eclipse-based IDE, MetaWare, as well as GNU tool chain.
With the right processors, the right tools, and the right engineers, I look forward to seeing us do a lot more with less.
Sensors are becoming more prolific and changing the way that you interact with your world. That smart phone in your pocket is a good example. A lot of what you can do with it is the result of sophisticated sensors that are built into it. For instance, the accelerometer inside determines the orientation in which you are holding the phone so the screen can switch making it easier to read. It also makes games that you control by moving the phone and applications like a bubble level possible. The accelerometer is actually a very sophisticated piece of technology, which is true of many of the sensors that are being developed today.
Controlling this new generation of sensors and interpreting the information that they generate requires a lot of compute power. This is increasingly being done with a 32-bit processor. The challenge is that the processor has to use almost no power and has to be infinitesimally small. The power consumption of a cell phone accelerometer is typically less than 100uA for the whole sensor. The processor has to use less than this but has to be able to process the necessary information in real-time. This requires a new class of 32-bit processor, and is one of the reasons that Synopsys developed the ARC EM family.
Targeted at sensors and other deeply embedded applications the EM family offers almost twice the performance of the processors used only 5 years ago to run a cellphone. The EM family does this while fitting into spaces as small as 0.01 mm2 and consuming as little as 2 uW/MHz.
As the sophistication of sensors continues to evolve the need for floating point calculations will grow to handle increasing levels of precision. Adding a floating point coprocessor is out of the question in sensors because of the extreme power and size limitations. While floating point calculations can be done on a 32-bit processor the performance will not be enough for many applications. Recognizing this Synopsys recently released a floating point unit that can be integrated with the EM processor cores. This floating point option (starting at 10K gates) requires only about 10% of the area and power of a coprocessor, but offers high performance for single and double precision math and complies with the IEEE-754 standard.
Because of the capabilities that they bring, sensor usage is growing geometrically and they are starting to show up everywhere and in everything. You interact with many more sensors every day than you do people. This is happening because of the new generation of 32-bit processors that are making devices possible that we could only dream about a few years ago. Can you imagine what we will be able to do five years from now!
We like to blog about how technology changes our lives and will continue to do so. And, it is exciting to see Synopsys leading the semiconductor industry and pioneering new solutions.
For many years System on Chip (SoC) integrators have faced the challenge of packing more and more features into their designs, of course using fewer resources and within a shorter time frame. Synopsys has lead the way offering a broad range of SoC IP cores. We are now taking the next step with the introduction of the Industry’s First Complete Audio IP Subsystem that enables designers to cope with the continued increase in SoC design and audio complexity.
Talking with customers we found that what really drives system complexity is software. There is a lot of software in SoCs today, and the amount is growing. This is increasing the design task and in many ways making the hardware IP implementation almost easy by comparison. Of course, all of this software has to be integrated into the applications software that is running on the host-processor, and then everything has to be verified. From the chart below you can see that the number of IP blocks per SoC is growing rapidly, and will continue to do so. What the chart doesn’t tell you is that the amount of software required to support all of these IP blocks is growing even faster. Certainly, the number of IP blocks in SoCs has lead to a lot of discussion in the press about subsystems. What you don’t often see is the real motivator behind the trend to subsystems, and that is software and the integration of the software to make an SoC. A subsystem is more than IP cores and has to include the full software solution, or it will be of limited value.
We have developed the SoundWave Audio Subsystem in response to customer requests and the changing conditions that we see in the IP market. The SoundWave Subsystem is a complete audio subsystem solution. It includes all of the hardware IP blocks and all of the software that is needed to implement a broad range of audio solutions for SoCs.
The hardware IP includes: 32-bit ARC audio processors, digital I2S and S/PDIF interfaces for off-chip audio connections as well as high-bandwidth on-chip connections to interfaces like HDMI. ARM® AMBA® 3 AXI™/AHB protocol system interfaces ease integration into the SoC infrastructure. Analog audio codecs provide high-quality audio connections for line inputs and outputs, microphones, loud speakers and headphones. An easy-to-use configuration tool allows designers to quickly select options such as number of channels and number of audio interfaces, enabling a complete audio subsystem to be configured in hours instead of weeks if done manually.
The software included with the SoundWave Audio Subsystem is a complete, ready-to-use environment including an integrated media streaming framework, an RTOS, a broad portfolio of codecs and post-processing components, and a plug-in to the host application software. This is a complete software stack that includes everything that is needed to create great audio solutions for SoCs. The software is fully integrated with the hardware, it is configurable, and it is verified. We also have support for virtual and FPGA prototyping to make the software integration easier.
The SoundWave Audio Subsystem is the first complete subsystem (hardware, software, prototyping) to be offered by any company. All of the hardware and all of the software needed to implement a broad range of audio solutions for SoCs is included and ready to use. With the release of the SoundWave Subsystem Synopsys is leading the way and defining what a subsystem is for the market.
I few months ago, I posted a blog (“Bringing Order to Chaos”) about the inherent power of our hyper-connectivity to bring order to potentially chaotic situations. This hyper-connectivity is the result of the dramatic increase in processing power that we are realizing with microprocessor technology and is giving us the tools we need (we know where you are, how many are in an area, and can get you information at the push of a button) to instantly modify behavior through the proliferation of information.
On a recent trip to visit Synopsys customers in Israel, I saw this all come to life in real-time, maybe not completely bringing order to chaos, but certainly reducing chaos. What is more chaotic than your daily commute? It’s likely many of you reading this already know about and are using this crowd sourcing tool, Waze? “When you download Waze, you not only get free navigation, but also become part of the local driving community in your area, joining forces with other drivers nearby to outsmart traffic, save time and improve everyone’s daily commute.”
Waze leverages the powerful (and expensive) computing, navigation, connectivity, display, and battery power of your smart phone. But what about other opportunities for data collection and information sharing where a phone won’t fit or can’t be plugged in daily to recharge batteries? This is where tiny, power efficient microprocessors, like the ARC EM family come in. Wireless sensor products based on the ARC EM processor cores could go without the need to ever change (or charge via plug-in) batteries, or without batteries all together. Maybe this is in asset tracking, under a tag, or simply, built into a product. Maybe this is in clothing, measuring and reporting body temperatures and heart rates. They could be embedded in the rubber in your car’s tires, monitoring pressure and temperature. The applications, and the benefits, are endless.
The ARC EM 32-bit processor core was introduced to the market last year, and at less than 10,000 gates this processor consumes 0.01mm2 of silicon area in 28nm and draws just 2uW/MHz of current. Oh, and although it is 8-bit in size, the performance is flat out 32-bit. The ARC EM delivers a whopping 1.52 DMIPS/MHz and 2.29 CoreMark/MHz. That’s best in class performance for low gate count 32-bit processors.
The low gate count and efficient sleep modes means your leakage current is kept to a minimum. The real-time performance efficiency means you can wake up, process fast, and get back to sleep, to minimize power on time. With microprocessor like this you can create a long lasting, low-maintenance sensing node that can go just about anywhere.
In a few years sensor nodes will be everywhere and will be possible everywhere because of the advances that are being realized in 32-bit microprocessor architectures. The extreme performance efficiency of processors like the ARC EM cores is changing our world and bringing to life capabilities that up until now only seemed possible in science fiction. The increased access to data and connection to the cloud that results will bring us all closer together and with the data available analyzed, shared, and distributed in intelligent ways, like Waze has done with the smart phone, will change the world we live in.
When we have a choice to pick a name for something, it usually reflects some meaning for us. For example, I chose my daughter’s name as Abigail because I once saw a Mike Leigh play called ‘Abigail’s Party’ (the video below is a clip from the play). But, I also chose it because the name means ‘Father’s Joy’. Similarly, my name is Allen. It is of Scottish and Irish origin and it means ’handsome’. So, now you know why my parents chose to call me Allen.
Recently, we were battling with what to call our new partner program for the DesignWare ARC processor cores. We could have just called it the ARC Partner Program or the ARC Alliance Program or the ARC Community Program. But, I really wanted the name to have meaning.
I started by thinking about what the objectives of the program are:
• To broaden embedded industry support for the ARC Architecture and cores
• To partner with leading embedded software and hardware vendors to provide compelling ARC-based solutions
• Finally, to increase the awareness of ARC-based solutions
While these are important & noble objectives, nothing jumped out at me for choosing a name. So, I then thought about what the program means for our customers.
• Develop your ARC-based embedded processor solutions faster by leveraging compatible products from leading embedded industry vendors
• Reduce your project risk by taking advantage of design solutions pre-ported and tested for the DesignWare ARC architecture
• Save on development costs and resources by using products optimized for ARC-based designs
These are all good things. Develop faster. Reduce Risk. Save costs. I could call it the ARC Faster Reduce Risk Save Cost Program or ARC FRRSC Program. But, I thought better of that.
What to do? Thinking about what the program’s objectives and meaning for customers was not getting me to a name. That’s when I thought about how we are going to do this. We will work with partners by providing them access to ARC software & hardware development tools and provide support so that that they can optimize their product for the ARC Architecture. This, in turn, will give our customers access to a broad array of industry solutions. Eureka! Did you spot the operative word? Access. We’re providing access for our partners and, in turn, providing access for our customers.
P.S. The actual reason I was called Allen is because it was my mother’s maiden name. I was not given a middle name, so that my full name is the combination of each of my parent’s last names. As I said, there’s always a reason why we name things the way we do!
P.P.S. Although the play was called ‘Abigail’s Party’, we never get to see Abigail or her party.
There was a time when you got your new computer home it was likely to be an HP and it ran Microsoft Windows on an Intel processor. You knew what to expect. Well, not anymore. At Microsoft’s BUILD conference for developers in Southern California, they unveiled their next operating system on a machine that didn’t have an Intel processor inside. At the same time, a few hundred miles north in San Francisco, Intel was unveiling a software partnership for an operating system – and it wasn’t with Microsoft! A few weeks ago, HP said it was spinning off or getting rid of its PC business, or not. What’s going on! Can’t we trust the status quo anymore? What happened to the Old Order?
Microsoft’s announcement of Windows 8 is showing a Windows that doesn’t look much like the old Windows we know and love. It has ‘tiles’ and ‘charms’ and ‘snapping’, but not much in the way of actual windows. Applications run full screen. It works on desktop PCs, laptops and tablets. Interestingly, Intel’s announcement with Google wasn’t for a PC operating system, but one for smart phones and tablets. It was Android, of course.
In the more strictly embedded world, we‘re used to seeing a multitude of processors and operating systems. There was a seismic shift to Linux a few years ago and one, more recently, to Android, in all kinds of devices. But, changes like this, in the PC world, are not common. Intel is working hard to re-invigorate the PC market with their Ultrabook™ category. These are sleek, lightweight, but powerful laptops that will be manufactured by companies such as Samsung and Acer. They are very cool.
So, the next few years will be very exciting for the new devices that we’ll all be using every day. PC’s, phones and tablets with new form factors and new (and new-looking) operating systems. And what has caused all this disruption to our lives? The one company I haven’t mentioned: Apple.
We are moving from the “Mobile Revolution” – the revolutionary advance that is allowing everyone on the planet to connect to everyone else, all the time – to the “Internet of Things.” This latest revolution will be profound because not only will we be connected to each other, but our stuff will be connected too. Our cars, refrigerators, light switches, cameras, and every other device we interact with will become intelligent and will connect seamlessly to other intelligent devices to make life easier without our even having to pay attention.
Just as the Mobile Revolution required a new type of processor, new types of processors are emerging to handle this new task. These processors put less emphasis on the ability to run a human-interactive OS and more emphasis on even lower power (even drawing power from their environment), lower cost, and smaller size. These processors, along with the devices they control, need to disappear into the background and do their job without bothering us.
These deeply embedded processors will be shipped in the billions. To meet that need, we’ve seen an explosion of new small processors. Synopsys is supporting many designs using innovative new processors from IP companies like ARM, MIPS, and Tensilica, plus in-house embedded processors from traditional IDMs like Renesas/NEC. We have also updated our own ARC family to bring the traditional strengths of the ARC architecture – configurability, very low cost, and excellent power/performance – to this new deeply embedded market. The 32-bit ARC ARC EM4 and ARC EM6 processors are small, starting at less than 10K gates, but they deliver over 1.5 DMIPS/MHz. At 28nm they can be clocked at more than 950MHz, delivering 1425 total DMIPS while consuming as little as 2.3uW/MHz.
What are the interesting places where these new 32-bit processors are likely to start showing up? Of course, there are the typical applications, such as sensors, actuators, 8- & 16-bit replacement, portable devices, power management, offload processing, and so on… However Merriam Webster defines “embed” as:
1a: to enclose closely in or as if in a matrix
b: to make something an integral part of
With these new processors “embedded processing” can take on a new meaning. For example, in Minneapolis the St. Anthony Falls Bridge is being built with a network of 323 sensors which will monitor the span for corrosion in the concrete, strained joints, or other structural weaknesses. An anti-icing system will track the roadway’s temperature and spray potassium acetate before ice begins to form. There’s also a traffic monitoring system, which detects the speed and volume of cars on the span. If there’s an accident that blocks the roadway, information can be relayed to central command so drivers can be alerted or rerouted. In other words, intelligence is being embedded right into the bridge’s concrete and steel.
Where else will these new deeply embedded processors go? Will they be in everything that we interact with? Will they be in our clothes? Will they even be in products that you use once and then throw away? Time will tell, but with the “Internet of Things we can expect to find deeply embedded microprocessors everywhere, in everything, and in places previously out of reach, literally.