Parallel Processing Speeds Toolpath Calculations

Multitasking—the word conjures images of frenzied cubicle dwellers juggling (or, more likely, attempting to juggle) e-mail, phone conversations, Web surfing, word processing and more all at once. Although a study recently published in the Proceedings of the National Academy of Sciences suggests humans aren’t wired for such activity, computers are far better equipped to balance multiple tasks simultaneously. Advances in parallel computing techniques and the relatively recent availability of affordable multi-core processors have dramatically improved this capability.

What does this mean for the manufacturing industry? Significant productivity increases, according to one of the CAD/CAM developers moving to take advantage of these recent hardware developments. Delcam (Windsor, Ontario) says version 10 of its PowerMill CAM software leverages the power of parallel computing to reduce toolpath calculation times and increase output.

In essence, parallel computing refers to the division of program tasks among multiple processors, or cores. This allows a properly outfitted computer to perform faster by simultaneously executing multiple calculations or separate portions of a single calculation. According to Delcam, incorporating this technology into PowerMill provides three important benefits:

• The ability to calculate or edit one set of tool paths in the foreground while the program calculates another set in the background, with minimal degradation in processing speed. Known as background processing, this capability eliminates the need to wait for each calculation to be completed before preparing for the next operation. The user simply adds tool paths to a queue, and the program will calculate each in sequence.

• The ability to split calculation of a single complex tool path among multiple processing units to reduce overall calculation time. Known as parallel processing, this happens automatically—the user doesn’t need to do anything to activate it.

• The application of parallel processing to both foreground and background calculations to provide even greater performance gains.

Of course, in order to take full advantage of these capabilities, users need the proper hardware—a computer equipped with one or more multi-core processors. Performance also varies significantly depending on the processor configuration, the company says. For example, although a quad-core processor will provide greater benefit than a dual-core, a single quad-core processor is actually faster than two quad-cores. (A white paper from Delcam provides more detail about this and other topics).

Regardless, most computers sold today are equipped with at least dual-core processors, and productivity improvements typically enable all but the most casual users to quickly recoup any investment required for an upgrade, says Colin Jones, PowerMill software development manager. Delcam is willing to work with customers to determine what benefits can be expected and what hardware configurations would best suit their particular needs, he adds.

Additionally, performance gains depend heavily on the specific tool paths and machining strategies employed by the user. "Everyone wants to know exactly how much faster it will go, but that’s a difficult question to answer," Mr. Jones explains. "It really depends on what you’re doing. We could make a blanket statement—‘it will go four times faster,’ for example—but that would be misleading. Tests may show that one algorithm will go four times faster, but that algorithm could be used in different ways across different machining strategies."

In PowerMill 10, raster machining calculations benefit most from parallel processing, Mr. Jones says. When determining which algorithms were most suitable for rewriting to take advantage of parallelization—a complicated process—the company decided that raster machining tool paths would provide the most "bang for the buck," he explains. "That’s a core algorithm that’s used across a lot of the strategies in PowerMill, so if we can make that faster, we can make everything faster."

Point distribution also makes heavy use of parallel processing. Other improved strategies include constant Z, 3D offset, area clearance, interleaved constant Z, optimized constant Z and boundary calculations. However, Mr. Jones is quick to point out that the current version of PowerMill provides only a fraction of the performance gains expected to be available in future upgrades. He notes that Delcam develops all its own code, as opposed to relying on libraries from external suppliers, so it has the freedom to rewrite virtually any part of PowerMill that it deems suitable for parallelization. "This is by no means the end of it," he says. "There are lots of other algorithms to address, and we’ll be working across the whole program to see where we can apply this technology to speed things up."