feature article
Subscribe Now

Starting Over

Magma Develops Tekton From a Clean Sheet

How many times have you said, “If I had it all to do over again…”? It’s a natural consequence of the human learning process. Since we’re not privy to the future, we take a direction – some direction typically being preferable to going nowhere at all – and make the best of it. And as we go, we learn what was good and what was bad about that particular direction.

And along the way we’re typically presented with opportunities to make further decisions as forks in the road appear, and, at each such juncture, we can make best use of the knowledge accumulated since the start so that, hopefully, each decision is better than the prior one. And with a little luck, we keep moving forward.

But sometimes, after much valiant thrashing against the elements and with nothing to show for it, we decide that perhaps we shouldn’t have embarked in that direction at all, in which case we have to come all the way back and start over. Which, in general terms, suggests failure, something none of us wants to be associated with.

Fortunately, most directions provide some value, even if they aren’t perfect. Up to a point.

Typically, after a history of success, there comes a time of diminishing returns, when we seem to have gotten as far as we can go along a given path. We’ve gotten nearer our goal (a hazy concept since, typically, we can’t even clearly articulate the goal when we start out; we simply know that we’ll know it when we see it), but it seems we can’t quite reach it, no matter how much tweaking and juggling we do.

At this point it’s as if we’ve gone as far as we can along a promising path in a maze – but it turns out to be a path that won’t exit, so we’ll have to backtrack and try a different path. But there’s a problem: backtracking feels like failure. So we push on.

Backtracking is also a lot of work. We plan on the forward march. We don’t plan on revisiting the past. We don’t plan on rethinking our decisions. And so we stubbornly resist the call to fall back.

It’s like the bravado of the cult of “no regrets.” Always look forward, never look back. Beat the chest. Feel the testosterone surge. But what are regrets, really? If we ignore the maudlin mopings of the scotch-and-water sad sack at the end of the bar, regrets are, to a large extent, a manifestation of what we’ve learned. If we had it all to do again, we’d do it differently. “No regrets” means, “If I had it all to do again, I’d do it the same dang way: I’ve learned nothing useful.”

So you wonder what you can do if you relax and toss aside everything you’ve done, but keep everything you’ve learned, and start over. And there are a couple of recent examples of what’s possible when you rethink things. We saw one with Oasys’s upside-down approach to synthesis and with the resultant dramatic improvement in performance and memory footprint.

Magma appears to have pulled a similar trick with static timing analysis. With their new Tekton tool, they’ve more or less laid out a challenge saying that they can do any design in less than an hour on one machine. There’s a lot of wiggle-room in that statement, but, given the rest of what they say, the assumption is that any modern design, including all scenarios, and including cross-talk and on-chip variation (OCV), is fair game for this challenge.

The problem they’re trying to solve has gotten more complex in at least two dimensions. The first is simply the obvious fact that designs are bigger. Process and environment corners have always been an issue, but the modes are the new dimension that’s killing everyone now. These dimensions are orthogonal, meaning that, for full coverage, you have to generate a complete cross-product of modes and corners.

One design that Magma uses to illustrate their point comes with three modes and seven corners for a total of 21 scenarios. This is a 1.1-million-cell 65-nm design; anyone designing on tighter processes knows that three modes is nothing these days. There can be dozens of scenarios in more complex designs. And Magma suggests, for 28-nm designs, over 400 scenarios.

And yet Magma appears willing to take them on. And here’s why: the 21-scenario design on the “existing STA tool” (you can decide which that one might be) takes about 2 hours per scenario given one machine for each scenario. So you can get the results in 2 hours if you use 21 machines (and pay for 21 licenses), or you can get the results in 42 hours on one machine.

They claim to handle this same design in less than a half hour on one machine. They have a table of other designs, ranging up to 11.4 million cells; none takes longer than 41 minutes to run.

Of the examples they showed, none is on technology more aggressive than 40 nm. So if, in fact, you throw a 400-scenario 28-nm design at Tekton, linear arithmetic would suggest that it would take longer than an hour. 400 is roughly 20 times the size of their half-hour design, suggesting a 10-hour run for all 400 scenarios. But if everything scales, then this design would take 840 hours on the current incumbent. That’s five weeks. (In fact, it’s exactly five weeks… that’s almost suspicious in its convenientness…)

So, whether or not they can do every design in less than an hour, the numbers they can do appear to be a dramatic departure from what’s been possible up to now. As far as I can tell, there are no obvious shenanigans in defining their results. They’re all with cross-talk and OCV on, and they’re all completely correlatable to Primetime.

If you ask how they were able to do it, they don’t give much in the way of detailed specifics; they boil it all down to this: they threw everything away and started with a clean slate.

Link: Tekton

Leave a Reply

featured blogs
Nov 25, 2020
It constantly amazes me how there are always multiple ways of doing things. The problem is that sometimes it'€™s hard to decide which option is best....
Nov 25, 2020
[From the last episode: We looked at what it takes to generate data that can be used to train machine-learning .] We take a break from learning how IoT technology works for one of our occasional posts on how IoT technology is used. In this case, we look at trucking fleet mana...
Nov 25, 2020
It might seem simple, but database units and accuracy directly relate to the artwork generated, and it is possible to misunderstand the artwork format as it relates to the board setup. Thirty years... [[ Click on the title to access the full blog on the Cadence Community sit...
Nov 23, 2020
Readers of the Samtec blog know we are always talking about next-gen speed. Current channels rates are running at 56 Gbps PAM4. However, system designers are starting to look at 112 Gbps PAM4 data rates. Intuition would say that bleeding edge data rates like 112 Gbps PAM4 onl...

featured video

Available DesignWare MIPI D-PHY IP for 22-nm Process

Sponsored by Synopsys

This video describes the advantages of Synopsys' MIPI D-PHY IP for 22-nm process, available in RX, TX, bidirectional mode, 2 and 4 lanes, operating at 10 Gbps. The IP is ideal for IoT, automotive, and AI Edge applications.

Click here for more information about DesignWare MIPI IP Solutions

featured paper

Keys to quick success using high-speed data converters

Sponsored by Texas Instruments

Whether you’re designing an aerospace system, test and measurement equipment or automotive lidar AFE, hardware designers using high-speed data converters face tough challenges with high-frequency inputs, outputs, clock rates and digital interface. Issues might include connecting with your field-programmable gate array, being confident that your first design pass will work or determining how to best model the system before building it. In this article, we take a look at each of these challenges.

Click here to download the whitepaper

Featured Chalk Talk

Maxim's Himalaya uSLIC Portfolio

Sponsored by Mouser Electronics and Maxim Integrated

With form factors continuing to shrink, most engineers are working hard to reduce the number of discrete components in their designs. Power supplies, in particular, are problematic - often requiring a number of large components. In this episode of Chalk Talk, Amelia Dalton chats with John Woodward of Maxim Integrated about how power modules can save board space, improve performance, and help reliability.

Click here for more information about Maxim Integrated Himalaya uSLIC™ MAXM1546x Step-Down Power Modules