Discover science successes with XSEDE. Read feature stories where XSEDE has helped researchers in their work.
Supercomputers are making DISSCO different
Supercomputers are making DISSCO different
Sever Tipei, a Professor of Composition-Theory at the University of Illinois at Urbana-Champaign, composes with a keyboard -- a QWERTY keyboard, that is, and a healthy appetite for experimentation.
By Katherine Kendig, NCSA Public Affairs
Sever Tipei, a Professor of Composition-Theory at the University of Illinois at Urbana-Champaign, composes with a keyboard -- a QWERTY keyboard, that is, and a healthy appetite for experimentation. In 2005, he and mathematician Hans Kaper (of the DOE's Argonne National Laboratory) created DISSCO, a Digital Instrument for Sound Synthesis and Composition. As its name suggests, DISSCO doesn't just allow users to compose music; it also enables them to digitally create the sounds that will comprise each piece. Unlike an electronic instrument with a variety of pre-programmed sound options, DISSCO users build sounds algorithmically – and completely from scratch.
Now in its second decade, Tipei says DISSCO is "an ongoing project." The aesthetics and the functionality of the program have come a long way since the beginning, with user-friendly updates like the addition of a custom DISSCO interface to replace the original command-line operation. Now, Tipei is implementing new features and marshalling outside resources to keep driving DISSCO forward: to improve DISSCO's computational speed, he's enlisted the computing and consulting resources of XSEDE.
Tipei says Alan Craig, a member of XSEDE's Extended Collaborative Support Services (ECSS) team, was extremely helpful in pushing the collaboration forward: "Alan has been instrumental. I didn't know about XSEDE – he introduced me to it." Craig specializes in helping researchers from the arts, social sciences, and humanities advance their projects using XSEDE resources. To enhance DISSCO, ECSS consultants Paul Rodriguez and Bob Sinkovits worked on the Comet supercomputer at the San Diego Supercomputer Center to implement parallel processing and reduce the compute time for sound synthesis. Heading into the collaboration with XSEDE, a twelve-minute piece of music could take DISSCO fifty minutes to compute because for each second of sound, Tipei says, 44,100 samples representing frequency, volume, depth, location, and other "shades of gray" must be processed. Rodriguez and Sinkovits achieved a 70% reduction in computing time by implementing the Message Passing Interface (MPI) for parallel architectures on top of existing code -- meaning that same twelve-minute piece could now be computed in as little as fifteen minutes. Tipei says Craig provided consistent communication, cooperation, and support as Rodriguez and Sinkovits optimized DISSCO's code.
DISSCO forces users to engage with sound in multidimensional ways. Whereas pitch is often considered the most important aspect of a sound, DISSCO gives other features equal importance. For instance, spatialization – the perceived location of a sound – "brings the textures [of a piece] alive." The possibilities for each created sound are nearly unlimited, and Tipei says that the process of composing and synthesizing sounds is typically recursive. It can take a month, he says, to fully conceive and create a three-minute piece.
The music that results is inherently experimental. Just like in the sciences, Tipei says, "you set up an experiment, but you might be surprised by what comes out." Tipei himself actively employs uncertainty in his music, using DISSCO as a tool to break away from narrativity. "No matter what you do, there is going to be some cohesion – each piece has its own coherence. But," he adds, "not necessarily a story." By working against expectation, Tipei forces listeners to actively synthesize what they're hearing, instead of accepting something that feels ready-made. "It's a step farther than jazz," he says.
Sound synthesis can be useful for more than musical composition, and DISSCO isn't just for musicians: Tipei notes the program can be a powerful tool for sonification, or the representation of scientific data through sound. "DISSCO can handle more than a dozen degrees of freedom in the definition of a single sound," Tipei says, so data -- particularly complex computations -- can be distinctly rendered, giving researchers an auditory supplement or alternative to traditional visualization. Collaboration with scientists studying Graph Theory and Information Theory could also create a pathway for DISSCO to contribute to the study of artificial life: "Determining the evolution of a musical object in continuous transformation, an Evolving Entity… could trigger new ideas" in the field, Tipei suggests.
As DISSCO continues to develop in a changing technological landscape, Tipei is grateful for the advances working with XSEDE has enabled. While commercial sound synthesis software is more readily available in 2017 than it was when DISSCO was brought to life in 2005, Tipei says DISSCO is a step ahead, fusing composition and sound synthesis in one environment and providing a measure of continuity other programs don't offer. "I don't mean to brag," he adds, but "DISSCO is different."
"Quilt", Sever Tipei