When I enable clock gating in my synthesis flow (using Genus 18.15), my simulation (using Xcelium) on the post-synthesis netlist fails. The simulation succeeds pre-synthesis and also if I remove clock-gating in the design. I use set_db  lp_insert_clock_gating true to enable clock gating during synthesis. I printed out some of the signals from the netlist and can see where it fails (it incorrectly writes a register). However, I am not sure how to solve this issue or what I should be looking for. Any help would be appreciated. Thanks.

Hi,

Is it possible to use a SystemVerilog package which is compiled into a library and then use it in a VHDL-2008 design file? Is such mixed-language flow supported?

I'm considering the latest versions of Incisive / Xcelium available today (Oct 2019).

Thank you,

Michal

For a netlist vs. netlist LEC flow we have to solve the following problem:

- in the RTL code we replicate a large array of N x M all-identical hard-macros, let call them MACRO_A

- MACRO_A is pre-assembled in Innovus and contains digital parts and analog parts (bottom-up hierarchical flow)

- at top-level (full-chip) we instantiate this array of all-identical macros

- in the top-level place-and-route flow we perform ecoChangeCell to remaster the top row of this array with MACRO_B

- MACRO_B is just a copy of the original MACRO_A cell containing same pins position, same internal digital functionality and also same digital layout, only slight differences in one analog block inside the macro

- MACRO_A and MACRO_B have the same .lib file generated with the do_extract_model command at the end of the Innovus flow, they only differ in the name of the macro

- when performing post-synthesis netlist vs post-place-and-route we load .lib files of both macros in Conformal LEC

- the LEC flow fails because Conformal LEC sees only MACRO_A instantiated in the post-synthesis netlist and both MACRO_A and MACRO_B in the post-palce-and-route netlist

Since both digital functionality and STD cells layout are the same between MACRO_A and MACRO_B we don't want to keep track of this difference already at RTL stage, we just want to perform this ECO change in place-and-route and force Conformal to assume equivalence between MACRO_A and MACRO_B .

Basically what I'm searching for is something similar to the add_instance_equivalences Conformal command but that works between Golden and Revised designs on cell primitives/black-boxes .

Is this flow supported ?

Thanks in advance

Luca

hello sir this is Ganesh  from NIT Hamirpur pursuing MTech in VLSI. I have doubt regarding SDF i'm using genus tool for synthesis & after synthesis when i'm generating SDF it is giving delays by default for maximum values but i want all the delays like minimum:Typical:Maximum how can i do this. Is there any provision to set PVT values manually for SDF generation so that i can get all the delay values.

I have been working on a an ASA 17.2 project for the last 6 months.

When I go to Project --> Settings, the settings window does not open. 

The tool indicates that a window is open, as I cannot click on anything else in the project. But it does not show the Settings window.

This has been happening only for the last 2 months. Before that it was working fine.

If I send the project to my colleague, the settings window shows up for him.

Happy New Year! Have you tried the new Transmission Line Library (rfTlineLib) yet? In case you missed it, rfTlineLib was introduced in IC 6.1.6 ISR1 plus MMSIM 12.1.1 -or- MMSIM13.1. You may wonder....Why should I use the new rfTlineLib ? Well...(read more)

Hi Folks, A question that I've often received from designers, "Is there a method to determine the amount of memory required before I submit a job? I use distributed processing and need to provide an estimate before submitting jobs." The answer...(read more)

RF Enthusiasts, Come connect with Cadence RF experts and discover the latest advances in Cadence RF technologies, including Spectre RF at the IEEE International Microwave Symposium (IMS) 2014. This year, IMS will be held in Tampa, Florida. Cadence...(read more)

Last year, I wrote a blog post entitled Modeling Oscillators with Arbitrary Phase Noise Profiles . We now have an easier way to do this. Starting in MMSIM 13.1 , you can specify the phase noise as an instance parameter in Spectre sources, including...(read more)

Hi Folks, Check out this great new video on YouTube: CDNLive SV 2014: PMC Improves Visibility and Performance with Spectre APS In this video from CDNLive Silicon Valley 2014, Jurgen Hissen, principal engineer, MSCAD, at PMC, discusses an aggressive...(read more)

Hello Spectre RF Users, Next week is my all time favorite technical conference - the International Microwave Symposium IMS2016 , May 22-27 in San Francisco, CA at the Moscone Center. If you're at the conference, please stop by the Cadence booth and...(read more)

Hello Spectre Users, Simulating S-parameters in a time domain (transient, periodic steady state) simulator has been and continues to be a challenge for many analog and RF designers. I'm often asked: What is required in order to achieve accurate...(read more)

Hi All, If you were unable to attend IMS 2017 in June 2017, the IMS MicroApp “7 Habits of Highly Successful S-Parameters” is on our Cadence website. On Cadence Online Support , the in-depth AppNote is here: 20466646 . Best regards, Tawna...(read more)

The Triple Beat analysis is similar to Rapid IP2/IP3 analysis except that it uses three tones instead of two. It is used in cases where two closely-spaced small-signal inputs from a transmitter leak in to the receiver along with an intended small-signal RF input signal. (read more)

I have a technology library (given by foundry) with leLsw layer section defined.
I do not want to touch it

I added few layers with an ITDB approach. Now I'm unable to see the added layers, as it is not present in the leLsw layer section of the main techlib.

I want the user of the new techlib to see all the layers by default.(I don't want the users to go to the properties of palette and switch the display option to techfile layers instead of leLsw)

Hello,

I am in a need of a skill program to find all instances of a specific cell (Including Mosaics), throughout the hierarchy. The program should print the instance's name, xy coordinates at the top level, and extract a label name that is dropped on top of it. In case there is no label on top of the found instance, the program should print "No Label Found" in the report text file. This program aims to map PADs cells within top level.

I am using the below Cadence's solution to find instances and it works well. The missing feature is to identify LABELs that are on top of the found instances. 

I tried to use dbGetOverlap() function, within the below code, in few setups but it seems to fail to identify the existence of labels on top of the found instances.

For example: 

overlapLabel=dbGetTrueOverlaps(cv cadr(instBox) list("M1" "text"))

I am interested to add to the Cadence's solution below some code in order to identify labels on top of the found instances.

Any tip would be greatly appreciated.

Thanks,

Danny

--------------------------------------------------------

I can get the bound stdcells using bndGetBoundObjects, but not get what each individual stdcell corresponds in layout.

Is there a way to get the layout bound stdcells of an array schematic symbol if the layout stdcell name do or do not match the symbol naming?

Once the schematic array stdcells are bound to the layout stdcells, how to get the correct terminal term~>name and net~>name?

Example of a schematic symbol and layout stdcell:

Schematic

INV<0:2>    instTerms~>terms~>name = ("vss" "vdd" "A" "Y")

                   instTerms~>net~>name = ("<*3>vss" "<*3>vdd" "in<0:2>" "nand2A,nand3B,nor2B")

Layout ( I know it is bad practice, but it happens )

stdcell1 instTerms~>terms~>name = ("vss" "vdd" "A" "Y")

             instTerms~>net~>name = ("vss" "vdd" "in<0>" "nand2A")

I23        instTerms~>terms~>name = ("vss" "vdd" "A" "Y")

             instTerms~>net~>name = ("vss" "vdd" "in<1>" "nand3B")

INV(2) instTerms~>terms~>name = ("vss" "vdd" "A" "Y")

             instTerms~>net~>name = ("vss" "vdd" "in<2>" "nor2B")

Paul

Hi Team,

Problem Statement:In hierarchical layout, I want to get BBOX of particular layer without actually flattening the layout.

Description:The layer can be at any hierarchical depth i.e both from PCELL or shapes but at top level if they are overlapping then I want the merged BBOX.

Now, I am able to get BBOX of all the shapes present at different hierarchy.But i finding issue in merging BBOX.

Please can help me on the same issue as I require efficient way to merge the BBOX because list containing the BBOX is huge.

Thanks in advance.

Regrads,

Prasanna

Hi,

I am running simulation in ADEXL and need a custom function for VIVA to apply same procedure to all signals saved in output. For instance, I have clock nets and I want to get all of them and look at the duty-cycle, edge rate etc.

It is a little more involved than about part since I have some regex and setof to filter before processing but if I can get all signals for current history, I can postprocess them later.

In ocean, I am just doing outputs() and getting all saved signals but I was able to do this in VIVA calculator due to the difficulties in getting current history, test name and opening result directory

thanks

yayla

Version Info:

ICADV12.3 64b 500.21

spectre -W =>

Tool 'cadenceMMSIM' Current project version '16.10.479'
sub-version  16.1.0.479.isr9

I have a modeless informational dialog box defined at the beginning of a SKILL script, but its contents don't display until the script finishes.

How do you get a modeless dialog box contents to display while a SKILL script is running?

procedure(myproc()

   prog((myvars)

     hiDisplayAppDBox()    ; opens blank dialog box - no dboxText contents show until script completes!

     ....rest of SKILL code in script...launches child processes

   );prog

);proc

When I place a pcell and do not change the W parameter (default is used) the value is not saved in the OA cell property.

When I change the default value of the super master now, the old pcell will get the new default value automatically because there is nothing saved inside the OA cell for this parameter.

Do you have any Idea, that how we can save the default values in the OA cell properties so that this value doesn't get updated if the default values are updated in the new PDKs

HI All,

I am new to skill. My requirement is

open layout 

get m0 layer cordinates in a layout

dump info into a text file

For example 2 input Nand, A,B output , vcc , vssx and internal net (n2) will be the m0 layers. I need info like in a text file.

n2 co ordinate

vssx (co ordinate)

a (co ordinate)

b (co ordinate ) .

I found similar code in cadence form . Can you help me on this

 procedure(printPts()
let(    (type
    (cnt 0)
    (objList geGetSelSet()))

foreach(obj objList
    ++cnt
    type = obj~>objType
    case(type
        ("inst"
            printf("%s %L at %L " type obj~>xy))
        ("rect"
            printf("%s on layer %L at %L " type obj~>lpp obj~>bBox))
        ("polygon"
            printf("%s on layer %L at %L " type obj~>lpp obj~>points))
        ("path"
            printf("%s on layer %L at %L " type obj~>lpp obj~>points))
        ("pathSeg"
            printf("%s on layer %L at %L " type obj~>lpp list(obj~>beginPt obj~>endPt)))
        ("label"
            printf("%s on layer %L at %L " type obj~>lpp obj~>xy))
        (t    printf("%s not defined " type))
    )
)
printf("%n objects selected " cnt)
); end of let
); end of printPts

Hi,

I am trying add multiple choices to my radio field in cdf parameters. when i see the select the instance and try editing the Instance properties I can not view them in a single window. Instead i get a vertical sliding bar. Is there a way to display them in multiple rows?

-Haareeth

Hi there,

I'm running a transient simulation, and I want to get all instances with model implementation hisim_hv because after that I want to process the data and to adjust some parameters for this kind of devices before dumping the values.

What is the easiest/fastest way to get those instances in skill/ocean?

What I did until now: 

- save the final OP of the simulation and then in skill

openResults()
selectResults('tranOp)
report(?type "hisim_hv" ?param "vgs")

Output seems to be promising, and looks like I can redirect it to a file and after that I have to parse the file.

Is there other simple way? I mean to not save data to file and to parse it.

Eventually having an instance name, is it possible to get the model implementation (hsim_hv, bsim4, etc..)? 

Best Regards,

Marcel

It’s been quite a long 5-year journey building and deploying Performance Analysis, Verification, and Debug capabilities for Arm-based SoCs. We worked with some of the smartest engineers on the planet. First with the engineers at Arm, with whom we...(read more)

Here we conclude the blog series and highlight the results of Mediatek 's use of Cadence Perspec™ System Verifier for their SoC level verification. In case you missed it, Part 1 of the blog is here , and Part 2 of the blog is here . One of their key...(read more)

It’s now official: Perspec System Verifier is rated the #1 product in the #1 category of Portable Stimulus, according to the 2017 EDA User Survey published on Deepchip.com. There were 33 user responses in favor of Perspec as the #1 tool, and dr...(read more)

It’s always good to hear what real users think of products. Here is a very detailed review (~4000 words) by an Anonymous user, nick named Ant-Man (from the movie). Overall it’s a very strong endorsement of Perspec, and summarize...(read more)

CDNLive is a user conference, and verification is one of the largest categories of content with multiple tracks covering multiple days. Portable stimulus is one of the hottest new areas in verification, and continues to be popular in all venues. At l...(read more)

There’s nothing like the heat of a DAC demo to stress new technology and the engineers behind it! Such was the case at DAC 2018 at the new locale of Moscone Center West, San Francisco. Cadence and AMIQ were two of several vendors who announced ...(read more)

Cadence continues to be a leader in SoC verification and has expanded our industry investment in Accellera portable stimulus language standardization. Some customers have expressed reservations that portable stimulus requires the effort of learn...(read more)

Vegas, here we come. All of us fun EDA engineers at once. Be prepared, next week’s Design Automation Conference will be busy! The trends I had outlined after last DAC in 2018—system design, cloud, and machine learning—have...(read more)

Hi, I want to add values to components in my SiP design such as 1nF or 15nH. There is already in existence a COMP_VALUE property reserved for this as shown during BOM generation. This property is not visible under the Edit/Properties dialog for component or symbol find filters. We have already created user properties called COMP_MFG and COMP_MFG_PN that it editable at a component level. When we try to add COMP_VALUE it is reported as a reserved name in Cadence but this name is not listed in the properties dialog. Is there a way to turn on the visibility and editablility of this or other hidden reserved Cadence property names? How can I assign a string value to the COMP_VALUE property?

Thanks

Have you ever manufactured a printed circuit board (PCB) without analyzing all the routed signal traces? Most designers will say “yes, all the time.” Trace widths and spacing are set by constraints,...

[[ Click on the title to access the full blog on the Cadence Community site. ]]

Those who work in the IC Packaging design space have some unique challenges. We bridge between the IC design world (90/45-degree traces with rectangular and octagonal pins) and the PCB domain...

[[ Click on the title to access the full blog on the Cadence Community site. ]]

Today, we are living in the era where whatever we think of as an idea is not far from being implemented…thanks to machine learning (ML) and artificial intelligence (AI) entering into the...

[[ Click on the title to access the full blog on the Cadence Community site. ]]

In this week’s Whiteboard Wednesdays video, Dave Apte discusses how to create the lowest power design possible by using architectural exploration and Cadence’s Stratus HLS solution....

[[ Click on the title to access the full blog on the Cadence Community site. ]]

Recently, I read an article about how failure is the stepping stone to success in life. It instantly struck a chord and a thought came zinging from nowhere about what happens to the failed arcs of a...

[[ Click on the title to access the full blog on the Cadence Community site. ]]

Cadence ® Spectre ® AMS Designer is a high-performance mixed-signal simulation system. The ability to use multiple engines, and drive from a variety of platforms enables you to "rev...

[[ Click on the title to access the full blog on the Cadence Community site. ]]

The Linley Processor Conference always opens with a keynote by Linley Gwenapp giving an overview of processors in whatever is the hottest area. Most of the other presentations during the conference...

[[ Click on the title to access the full blog on the Cadence Community site. ]]

The 2019 HF1 production release for Clarity, Celsius, and Sigrity Tools is now available for download at Cadence Downloads . SIGRITY2019 HF1 For information about supported platforms, compatibility...

[[ Click on the title to access the full blog on the Cadence Community site. ]]

www.youtube.com/watch Made on my balcony (camera Carey Guo) Monday: EDA101 Video Tuesday: Weekend Update Wednesday: RAMAC Park and the Origin of the Disk Drive Thursday: 1G Mobile: AMPS, TOPS, C-450,...

[[ Click on the title to access the full blog on the Cadence Community site. ]]

I recently attended a webinar presented by Wally Rhines about his new book, Predicting Semiconductor Business Trends After Moore's Law . Wally was the CEO of Mentor, as you probably know. Now he...

[[ Click on the title to access the full blog on the Cadence Community site. ]]

We have talked about aspects of the in-design symbol edit application mode in the past. This is the environment specific to the Allegro® Package Designer Plus layout tools allowing you to work...

[[ Click on the title to access the full blog on the Cadence Community site. ]]

If I talk about my life, it was much simpler when I used to live with my parents. They took good care of whatever I wanted - in fact, they still do. But now, I am living alone, and sometimes I buy...

[[ Click on the title to access the full blog on the Cadence Community site. ]]

Complexity is driving the urgency for advanced artificial intelligence systems more than ever—and that means someone has to supply the tools to create those systems. Cadence is up to the task: we’ve been expanding our AI offerings. If you haven’t already seen what Cadence can do for your AI needs, or if you’re not quite up-to-date on this whole AI boom, let this presentation given by K.T. Moore at the Cadence Theater at DAC bring you up to speed.

The technology behind AI isn’t as new as you’d think—the principles that govern how AI learns have been in development since 1959, when Arthur Samuel defined the concept of “machine learning.” At the time, there was nothing even resembling the necessary compute power to put Samuel’s concepts into practice—but now we can. AI designs are huge, and they’re massively parallel—simulating them on older computers and simulators would have taken ages; never mind how long it would take to do some by-hand measure like they had to do in the '60s.

But with advancements in server technology and the parallelization technology in products like Xcelium Parallel Logic Simulator and JasperGold smart technology, plus hardware-based engines like the Palladium and Protium platforms, verifying AI designs is not only possible—it’s easy.  But, read on, its not just about simulation technology.

AI tech is flooding the industry. It’s applicable to almost every vertical—cloud computing can use AI to intelligently manage a user’s required resources, consumer electronics are using it to tailor a user experience based on a whole host of collected data, automotive companies want to use AI to drive cars, healthcare to assist in diagnoses given a set of symptoms and a database of other, similar patients—and that’s saying nothing of the multitude of industrial applications. AI is also useful in the creation of developers’ tools themselves. Part of what’s causing the semiconductor industry boom is just this—an exploding interest in AI chips. And with 5G technology imminent, and with the looming billion-gate plus sizes of the SoCs that implement 5G, AI-assisted developers' tools might need to become the norm, not an outlier.

So: in all of this, where is Cadence?

Cadence is focusing its efforts on two areas, dubbed “machine learning inside” and “machine learning outside.” ML inside in the digital design flow refers to improving PPA, faster engines, and better testing and diagnostics. None of this physically affects how you use a tool, but it makes using that tool a much better experience. ML outside talks about the design flow in general, working toward an automated design flow, as well as productivity improvements across the flow. These things do change how you use a tool, but don’t worry, it’s all for the better.

Additionally, Cadence is working to improve design enablement; that is, hardware and software co-design. Smart Genus and Innovus solutions make designing your SoC easier than ever—using the full flow can result in up to a 21% PPA gain.

If you’re looking specifically for IP to enable AI on your SoC, the Tensilica DNA 100 processor has you covered, too. It’s great for companies designing edge or AI chips, offers great compression rates and efficient power usage, and has 4.7X the performance of other AI SoC IP on similar array sizes.

Cadence has you covered no matter where you’re going in this new world of AI systems—with our AI-enabled tools, IP,  and our strong partner ecosystem, you can be at ease knowing you’ll be supported no matter how complex your needs are.

Everyone keeps talking about “the cloud” this and “the cloud” that these days—but you’re a semiconductor designer. Everyone keeps saying “the cloud” is revolutionizing all aspects of electronics design—but what does it mean for you? Cadence's own Tom Hackett discussed this in a presentation at the Cadence Theater during DAC 2019.

What people refer to as “the cloud” is commonly divided into three categories: Infrastructure as a Service (IaaS), Platform as a Service (PaaS), and software as a Service (SaaS). With IaaS, you bring your own software—i.e. loading your owned or appropriately licensed tools onto cloud hardware that you rent by the minute. This service is available from providers like Google Cloud Platform, Amazon Web Service, and Microsoft Azure. In PaaS (also available from the major cloud providers), you create your own offering using capabilities and a software design environment provided by the cloud vendor that makes subsequent scaling and distribution really easy because the service was “born in the cloud”.  Lastly, there’s SaaS, where the cloud is used to access and manage functionality and data without requiring users to set up or manage any of the underlying infrastructure used to provide it.  SaaS companies like Workday and Salesforce deliver their value in this manner.  The Cadence Cloud portfolio makes use of both IaaS and SaaS, depending on the customers’ interest.  Cadence doesn’t have PaaS offerings because our customers don’t create their own EDA software from building blocks that Cadence provides.

All of these designations are great, but you’re a semiconductor designer. Presumably you use Workday or some similar software, or have in the past when you were an intern, but what about all of your tools? Those aren’t on the cloud.

Wait—actually, they are.

Using EDA tools in the cloud allows you to address complexity and data explosion issues you would have to simply struggle through before. Since you don’t have to worry about having the compute-power on-site, you can use way more power than you could before. You may be wary about this new generation of cloud-based tools, but don’t worry: the old rules of cloud computing no longer apply. Cloud capacity is far larger than it used to be, and it’s more secure. Updates to scheduling software means that resource competition isn’t as big of a deal anymore. Clouds today have nearly unlimited capacity—they’re so large that you don’t ever need to worry about running out of space.

The vast increase in raw compute available to designers through the cloud makes something like automotive functional safety verification, previously an extremely long verification task, doable in a reasonable time frame. With the cloud, it’s easy to scale the amount of compute you’re using to fit your task—whether it’s an automotive functional safety-related design or a small one.

Nowadays, the Cadence Cloud Portfolio brings you the best and brightest in cloud technology. No matter what your use case is, the Cadence Cloud Portfolio has a solution that works for you. You can even access the Palladium Cloud, allowing you to try out the benefits of an accelerator without having to buy one.

Cloud computing is the future of EDA. See the future here.

Emulators are great. They vastly speed up verification to the point where it’s hard to imagine life without them; as designs grow in complexity, simple simulation can’t keep up for the biggest designs. The extra oomph from emulation is almost a necessity for the top percentages of design sizes. However, many users of Palladium aren’t efficiently using their unit’s processing power, and as a result they’re missing out on the full speed-up potential that Palladium can provide.

Altair’s HERO is here for you. With its help, your Palladium unit can be even more amazing for your productivity than before.

HERO (that’s Hardware Emulator Resource Optimizer) adds emulator support to Altair’s Accelerator. You already know and love Altair’s scheduling tools; so why not make them do more for you, so you can be one of those people who are making the most out of their Palladium system?

Emulators are kind of like big computers, but it’s a lot harder to manage leftover resources on an emulator than it is on, say, a CPU. A scheduler like HERO neatly sidesteps this problem by more intelligently using the resources available to ensure that there’s a minimal patchwork of leftover resources to begin with.

HERO supports past generations of Palladium as well, so if you’re still using an older version, you can still take advantage of the upgrades HERO provides. There’s a wide variety of features HERO has that make your emulator easier to use. HERO separates a job into a “select” section and a “run” section: the “select” part makes a last-minute decision on which domains or boards to use, while the “run” part is the actual job. This makes it easier to ensure that your Palladium emulator is being used as efficiently as possible. Jobs are placed using “shapes”, which are a set of job types; these can be selected from a list of pre-defined ones by the user. Shapes can have special constraints if those are needed.

A new reservation system also helps HERO organize Palladium’s processing power better. HERO offers both “hard” reservations and “soft” reservations. A hard reservation locks other users out of reserving any part of the emulator at all, while a soft reservation allows a user to reserve a part of the emulator for a later use. Think of it like this: a soft reservation is like grabbing a ticket from the deli counter, while a hard reservation stops you from ever entering the market.

When using HERO, you can manage your entire verification workload. You’ll find that your utilization of your emulator vastly increases—it’s been reported that some users using only 30% of the capabilities of their Palladium unit(s) saw a massive increase to over 90% once they made the switch to HERO.

If you’re ready to take your Palladium productivity to the next level, Altair has a HERO for you.

To see the full presentation given by Andrea Casotto in the Cadence Theater at DAC 2019, check here.

Autonomous vehicles are coming. In a statistic from the U.S. Department of Transportation, about 37,000 people died in car accidents in the United States in 2018. Having safe, fully automatic vehicles could drastically reduce that number—but the trick is figuring out how to make an autonomous vehicle safe. Internet-enabled systems in cars are more common than ever, and it’s unlikely that the use of them will slow or stop—and while they provide many conveniences to a driver, they also represent another attack surface that a potential criminal could use to disable a vehicle while driving.

So—what’s being done to combat this? Green Hills Software is on the case, and they explained the landscape of security in automotive systems in a presentation given by Max Hinson in the Cadence Theater at DAC 2019. They have software embedded [FS1] in most parts of a car, and all the major OEMs use their tech. The challenge they’ve taken on is far from a simple one—between the sheer complexity of modern automotive computer systems, safety requirements like the ISO 26262 standard, and the cost to develop and deploy software, they’ve got their work cut out for them. It’s the complexity of the systems that represents the biggest challenge, though. The autonomous cars of the future have dynamic behaviors, cognitive networks, require security certification to at least ASIL-D, require cyber security like you’d have on an important regular computer system to cover for the internet-enabled systems—and all of this comes with a caveat: under current verification abilities, it’s not possible to test every test case for the autonomous system. You’d be looking at trillions of test cases to reach full coverage—not even the strongest emulation units can cover that today.

With regular cars, you could do testing with crash-test dummies, and ramming the car into walls at high speeds in a lab and studying the results. Today, though, that won’t cut it. Testing like that doesn’t see if a car has side-channel vulnerabilities in its infotainment system, or if it can tell the difference between a stop sign and a yield sign. While driving might seem simple enough to those of us that have been doing it for a long time, to a computer, the sheer number of variables is astounding. A regular person can easily filter what’s important and what’s not, but a machine learning system would have to learn all of that from scratch. Green Hills Software posits that it would take nine billion miles of driving for a machine learning system of today’s caliber to reach an average driver’s level—and for an autonomous car, “average” isn’t good enough. It has to be perfect.

A certifier for autonomous vehicles has a herculean task, then. And if that doesn’t sound hard enough, consider this: in modern machine-vision systems, something called the “single pixel hack” can be exploited to mess them up. Let’s say you have a stop sign, and a system designed to recognize that object as a stop sign. Randomly, you change one pixel of the image to a different color, and then check to see if the system still recognizes the stop sign. To a human, who knows that a stop sign is octagonal, red, and has “STOP” written in white block letters, a stop sign that’s half blue and maybe bent a bit out of shape is still, obviously, a stop sign—plus, we can use context clues to ascertain that sign at an intersection where there’s a white line on the pavement in front of our vehicle probably means we should stop. We can do this because we can process the factors that identify a stop sign “softly”—it’s okay if it’s not quite right; we know what it’s supposed to be. Having a computer do the same is much more difficult. What if the stop sign has graffiti on it? Will the system still recognize it as a stop sign? How big of an aberration needs to be present before the system no longer acknowledges the mostly-red, mostly-octagonal object that might at one point have had “stop” written on it as a stop sign? To us, a stop sign is a stop sign, even with one pixel changed—but change it in the right spot, and the computer might disagree.

The National Institute of Security and Technology tracks vulnerabilities along those lines in all sorts of systems; by their database, a major vulnerability is found in Linux every three days. And despite all our efforts to promote security, this isn’t a battle we’re winning right now—the number of vulnerabilities is increasing all the time.

Check back next time to see the other side: what does Green Hills Software propose we do about these problems? Read part 2 now.

If you missed the first part of this series, you can find it here.

So: what does Green Hills Software propose we do?

The issue of “solving security” is, at its core, impossible—security can never be 100% assured. What we can do is make it as difficult as possible for security holes to develop. This can be done in a couple ways; one is to make small code in small packs executed by a “safing plan”—having each individual component be easier to verify goes a long way toward ensuring the security of the system. Don’t have sensors connect directly to objects—instead have them output to the safing plan first, which can establish control and ensure that nothing can be used incorrectly or in unintended ways. Make sure individual software components are sufficiently isolated to minimize the chances of a side-channel attack being viable.

What all of these practices mean, however, is that a system needs to be architected with security in mind from the very beginning. Managers need to emphasize and reward secure development right from the planning stages, or the comprehensive approach required to ensure that a system is as secure as it can be won’t come together. When something in someone else’s software breaks, pay attention—mistakes are costly, but only one person has to make it before others can learn from it and ensure it doesn’t happen again. Experts are experts for a reason—when an independent expert tells you something in your design is not secure, don’t brush them off because the fix is expensive. This is what Green Hills Software does, and it’s how they ensure that their software is secure.

Now, where does Cadence fit into all of this? Cadence has a number of certified secure offerings a user can take advantage of when planning their new designs. The Tensilica portfolio of IP is a great way to ensure basic components of your design are foolproof. As always, the Cadence Verification Suite is great for security verification in both simulation and emulation, and JasperGold platform’s formal apps are a part of that suite as well.

We are entering a new age of autonomous technology, and with that new age we have to update our security measures to match. It’s not good enough to “patch up” security at the end—security needs to beat the forefront of a verification engineer or hardware designer’s mind at all stages of development. For a lot of applications, quite literally, lives are at stake. It’s uncharted territory out there, but with Green Hills Software and Cadence’s tools and secure IP, we can ensure the safety of tomorrow.