130,000 frames, 90M core hours: Exascale’s Potential for Animation
The following contribution from a writer at Argonne National Labs is another compelling example of advanced scale computing’s impact on industry – in this case, the movie industry. Animated film production has a voracious appetitie for compute power, and in this article John Spizzirri looks at the potential impact of exascale (a billion billion calculations per second) on animation.
There is no questioning the power of a full head of shiny, buoyant hair. Not in real life, not in commercials, and, it turns out, not in computer-generated (CG) animation. Just as more expensive brands of shampoos provide volume, luster, and flow to a human head of hair, so too does more expensive computational power provide the waggle of a prince’s mane or raise the hackles of an evil yak.
Hair proves to be one of the most complex assets in animation, as each strand is comprised of near-infinite individual particles, affecting the way every other strand behaves. With the 2016 release of their feature Trolls, DreamWorks Animation had an entire ensemble of characters with hair as a primary feature. The studio will raise the bar again with the film sequel slated for 2020.
The history of DreamWorks Animation is, in many ways, the history of technical advances in computing over the last three decades. Those milestones are made evident by that flow of hair — or lack thereof — the ripple in a dragon’s leathery wing, or the texture and number of environments in any given film.
Exascale and Media/Entertainment
As the development and accessibility of high-performance computers explode beyond current limits, so too will the creative possibilities for the future of CG animation ignite.
Jeff Wike, Chief Technology Officer (CTO) of DreamWorks Animation, has seen many of the company’s innovations come and go, and fully appreciates both the obstacles and the potential of technological advances on his industry.
“Even today, technology limits what our artists can create,” says Wike. “They always want to up the game, and with the massive amount of technology that we throw at these films, the stakes are enormous.”
Along with his duties as CTO, Wike is a member of the U.S. Department of Energy’s Exascale Computing Project (ECP) Industry Council, comprised of industry leaders reliant on and looking to the future of high-performance computing now hurtling toward the exascale frontier.
The ability to perform a billion billion operations per second changes the manufacturing and services landscape for many types of industries and, as Wike will tell you, strip away the creative process and those in the animation industry are manufacturers of digital products.
“This is bigger than any one company or any one industry,” he says. “As a member of the ECP’s Industry Council, we share a common interest and goal with companies representing a diverse group of U.S. industries anxiously anticipating the era of exascale computing.”
Such capability could open a speed-of-light gap between DreamWorks’ current 3D animation and the studio’s origins, 23 years ago, as a 2D animation company producing computer-aided hand-drawn images
Growing CG animation
Wike’s role has evolved since he joined DreamWorks in 1997, with the distinctive job title of “technical gunslinger,” in which he served as part inventor, part MacGyver and part tech support.
When Chris deFaria joined DreamWorks Animation as president in March 2017, he instantly identified an untapped opportunity that only could be pursued at a studio where storytellers and technology innovators work in close proximity. He created a collaboration between these two areas in which the artists’ imaginations drive cutting-edge technology innovations which, in turn, drive the engineers to imagine even bigger. In essence, a perpetual motion machine of innovation and efficiency.
Under this new reign, Wike distills his broader role into three simple goals: make sure employees have what they need, reduce the cost and production time of films and continue to innovate in those areas that are transformational.
HPC and Innovation
For DreamWorks — and other large industry players like Disney and Pixar — the transformation of the animated landscape is driven by innovations in computer software and hardware.
Much of the CG animation industry was built on the backs of what were, in the late 1990s, fairly high-performance graphics-enabled processors. But computer technology advanced so quickly that DreamWorks was challenged to keep up with the latest and greatest.
“Some of the animators had home computers that were faster than what we had at work,” Wike recalls.
By the time Shrek appeared in 2001, after the early successes of DreamWorks’ first fully CG animated feature, Antz, and Pixar’s Toy Story, it was clear to the fledgling industry, and the movie industry as a whole, that CG animation was the next big wave. Audiences, too, already were expecting higher quality, more complexity and greater diversification with each succeeding film.
To meet mounting expectations, the industry needed a computational overhaul to afford them more power and greater consistency. As the early graphics processors faced more competition, the industry banded together to agree on common requirements, such as commodity hardware, open source libraries and codes. This developed into an approved list that makes it easier for vendors to support.
Today, DreamWorks’ artists use high-end dual processor, 32-core workstations with network-attached storage and HPE Gen9 servers utilizing 22,000 cores in the company’s data center. That number is expected to nearly double soon, as the company has now ramped up for production of How to Train Your Dragon 3.
It's still a long way from exascale. For that matter, it’s still a long way from petascale computers that can comprise upwards of 750,000 cores. But the industry continues to push the envelope of what’s possible and what is available. Continuous upgrades in hardware, along with retooling and development of software, create ever-more astounding visuals and further prepare the industry for the next massive leap in computing power.
“I’d be naïve to say that we’re ready for exascale, but we’re certainly mindful of it,” says Wike. “That’s one reason we are so interested in what the ECP is doing. The interaction with the technology stakeholders from a wide variety of industries is invaluable as we try to understand the full implications and benefits of exascale as an innovation driver for our own industry.”
“Every single film we’ve done has progressed because of technology, and each has used more computing resources than the previous one,” says Dave Walvoord, visual effects supervisor for DreamWorks Animation. “Sometimes they’re small steps, other times bigger. But it’s always a combination of hardware coupled with algorithms.”
While this can be traced back to the hairiness factor for each succeeding movie, it is perhaps most evident in the evolution of the company’s dragons.
DreamWorks’ initial dragon creation appears in the first Shrek. Donkey’s love interest, Dragon, as she is aptly named, is a less-than-dynamic character per the technical limitations at that time. By the time we meet the multitude of dragons in How to Train Your Dragon 2, in 2014, they are completely different beasts, fluid and graceful, the leathery skin of their wings subtly fluttering as they soar.
The visuals derived from the complex calculations, or algorithms, required to make dragons fly or water crash on icy shores, can only be as good as the hardware that is running them. In this case, the single-core-produced lumbering Dragon of Shrek was replaced by parallelized, multi-core hardware that brought to life the lithe swarm of creatures for How to Train Your Dragon 2.
Fast forward to the upcoming How to Train Your Dragon 3, and suddenly there has been an even more dramatic leap in animation quality since the introduction of the studio’s proprietary Moonray ray tracer for the rendering process, which determines the quality and the visual complexity of each image.
Rendering, which is the most computationally intensive part of animation, accounts for 90 percent of a film’s overall compute time, as each frame must be rendered individually, and some, like water simulations, are more computationally expensive than others. The 130,000 frames that comprise How to Train Your Dragon 2 took more than 90-million core hours to render.
Ray tracers use inherently highly scalable algorithms and model physics much more accurately, notes Walvoord, which should make production for How to Train Your Dragon 3 faster and its images more resonant.
“We’re scaling linearly at 32 cores and the software algorithms are truly parallel in a way we’ve never seen before,” he adds. “And because we’re calculating light correctly, it’s completely changing the richness of the films. In fact, if you watch clips from How to Train Your Dragon 2, they look very dated. What we’re doing right now just doesn’t compare. It’s really exciting.”
But an eventual transition to exascale won’t necessarily make the process any less demanding, say both Wike and Walvoord. In fact, the massive processing, parallelism and memory of exascale machines would further drive the “bigger, faster, better” quotient that has always been the expectation of both animator and audience member alike.
“If, let’s say, exascale is 10 times faster than what I have now, just to use a number, I have two choices,” Walvoord suggests. “There’s the business choice, which is I buy fewer computers, or there’s the artistic choice, which is I make an image 10 times more complicated. The answer may lie somewhere in between, but in general, creativity wins.”
Behind the Scenes — Literally
For Wike, technology, especially digital film, is the new oral tradition, and exascale can only open new avenues by which to tell richer, more complex stories.
“Technology is what enables people to tell those stories and distribute them to more people. If you think about what impact compute technology has had on our lives, it’s really been in our ability to communicate in so many different ways,” he says.
With a toehold on new high-performance computing platforms, Wike sees virtual reality as a key contributor to change in animation media, just as CG was 30 years ago. Perhaps in its adolescence right now, virtual reality is gaining ground among gamers, smartphone users, and the scientific simulation community. Propelled by exascale, virtual reality could quickly reach maturity.
Currently, cameras are fixed and we are only aware of the action happening in front of us. Environments beyond a camera’s scope are minimalized or nonexistent—a panda may be bare-bottomed if it only faces forward in a scene, or the bulk of a house may lie threadbare if all the action is in the kitchen.
But virtual reality changes that by allowing a director or the audience to go anywhere in a scene. Now the panda must be fully furred, the house fully constructed, which requires huge amounts of compute time and power.
“That could increase the amount of content in a shot by an order of magnitude or more,” says Wike. “If virtual reality explodes, you’re either going to make things that aren’t quite film quality, or you’re going to need far more compute. And that’s when you want that exascale capability.”
But Wike understands that achieving exascale is not reliant solely on his hardware arsenal nor does the company’s success ride on the generation of a new, more magnificent mane. He and DreamWorks are betting on a paradigm shift to make the transition to massively parallel computing less daunting.
The company has spent a substantial amount of money over the last five years in re-architecting its primary tool set to take advantage of hyper-multicore processing. It created, for instance, a rendering service for its ray tracer that can reconfigure, on the fly, how much compute power to allocate to a particular job or artist, and to distribute the rendering process across any number of machines.
And rather than try to port old software to meet high-performance requirements, they opted for new software designed for greater scalability on multiple cores. Not only does it add complexity to the visuals, but it transforms work culture, as well.
“We’re training our artists to adjust their workflow to take advantage of the number of cores that are going to be available to them. We all have to remain flexible and always think about how we can better position ourselves,” says Wike. “If you don’t do that, if you wait for the hardware to start this process, then you’re going to miss out on the competitive advantage.”
While costs and other factors could prove hurdles on the road to exascale, Wike is optimistic about the outcomes the ECP partnership could tender, and confident in the approach his company has taken to start the journey.
“By the time exascale becomes a commodity, which it will, we’ll be able to produce unbelievable experiences for audiences all over the world, and do it in a way and at a cost that is very different than it is today.”
John Spizzirri is a contributing writer at Argonne National Laboratory.