Michael Trott
The Science and Art of Mathematica
Theoretical physicist and senior Mathematica developer Michael Trott.
Dual G5 Processors and Intel Core Duo
Dual processors and high-end graphics configurations in the Power Mac provide an ideal platform for creating visualizations with Mathematica. “I love to use the new Macs with dual processors, because Mathematica is divided into two parts,” Trott explains. “The kernel does the real computations, and the front end does the interactive things like typesetting and display of graphics. So if you create large, complex graphics like I do, this is quite a serious computational effort. In the dual processor machines you can continue to work in your Mathematica notebook because nothing slows down. Some of my other computers get really slow when the system is working extremely hard.”
“On the new Intel Macs, the PNG export is just blindingly fast. It exceeded all my expectations.”
There are also simple software utilities in Mac OS X that make everyday life simpler for scientists like Trott, who deal with huge volumes of visuals for both static figures and animated presentations of results. Before handing off his results to colleagues who use Final Cut Pro and QuickTime to create and compress high-quality movies of his images, Trott uses Preview to quickly analyze each frame.
He explains: “When I make animations I use Mathematica to export PNGs, so to make big animations with hundreds of frames, I need to check them to see that they are all okay. On other platforms, I have to open them individually to inspect each one, but on the Mac — with Preview — I see thumbnails on the right side which is real nice and I can page through them quickly.”
Trott also uses Safari and Preview to examine 20 to 25 preprints and journal articles per day. In the true form of a mathematician, Trott calculated out the time he saves by opening each article so quickly. “I save a few seconds compared to other platforms,” he says. “This comes to a minute or two per day. Over the course of a year with 250 working days, this ‘little’ feature saves me nearly a whole day’s time.”
The speed of the new Intel Core Duo systems has also offered a significant benefit to Trott. While he plans on upgrading his notebook to an Intel Core Duo system, for now he shares an iMac with a colleague to use along with his Power Mac G5. The performance of the Intel Core Duo is so good that they commonly offload back-end computations to it when rendering large animated visualizations. “On the new Intel Macs, the PNG export is just blindingly fastTrott states,” he says “It exceeded all my expectations.”
A Multi-Platform Cluster
While in many cases running quick computations on an Intel-based Mac works well for Trott, his group also has an Xserve cluster for large jobs. The head node is an Intel Pentium server running Linux. The compute nodes consist of eight Apple Xserves running Mac OS X. The unique setup is a testament to the cross-platform compatibility of both Mathematica and the UNIX-based Mac OS X.
“Mathematica has a couple of tools which allow you to do this easily,” Trott explains. “You can connect to a cluster and have functions run that are easily parallelized. You send it out from Mathematica on your desktop machine, which collects all the results and displays them nicely together. This is a quite convenient tool if you have a large calculation at hand.”
A cross-platform cluster is unusual, but in many scientific computing environments cross-platform interaction is a common component in the work day. Mathematica runs on most variants of UNIX, Linux, and Windows, in addition to Mac OS X. Wolfram Research’s network consists of a litany of architectures and operating systems so it is important that Mac OS X can share with each platform.
“Our network has all kinds of machines,” says Trott. “Because everybody needs files from everybody else, they are all networked. It’s really nice how Mac OS X lets me have icons that are shortcuts for directories on other machines in Finder. It’s absolutely flawless — couldn’t be better.”
