Global Sources
EE Times-Asia
Stay in touch with EE Times Asia
EE Times-Asia > Controls/MCUs

NI: Path leads to parallel for test and measurement

Posted: 14 Mar 2008 ?? ?Print Version ?Bookmark and Share

Keywords:NIDays 2008? Chandran Nair? parallel shift? parallel computing? algorithm engineering?

Nair: The Parallel Shift brings a huge challenge: the marriage of parallelism with computing, and in related applications of measurement and control.

Today is an interesting time for the industry as it transitions to adopting a fresher approach in design through the use of parallel technologies. The use of parallel technologies opens more opportunities to step-up product development and design, which complements the constant need for innovation in the industry. That was the view stressed by Chandran Nair, managing director of National Instruments Southeast Asia during his keynote at NIDays 2008 in Manila last month.

Actions, computing and instructions occurring simultaneously are increasingly common requirements in today's world, as evidenced by the unstoppable migration from single-core to multicore processors in PCs. According to Nair, this new paradigm!the Parallel Shift!brings a huge challenge to test engineers: the marriage of parallelism with computing, and in related applications of measurement and control.

Parallel shift
The move to adopt the parallel paradigm is nothing new. However, the advancement of adopting parallel was eclipsed by the rise of the sequential methodology that remains widely used in computing today, given its easier implementation both in hardware and software. However, today's high performance and demanding applications requires a rethinking of the predominant methodology.

"Long ago, engineers have realized the big gap between the computational process and the physical world. Most physical activities like traffic flow are in parallel while our programming style is sequential," explained Nair. "In the early days, many have tried to bridge the gap between parallel reality vs. the dominant sequential programming model but the paradigm never took off the ground because the software available then was not ready to synthesize and develop parallel algorithm."

He said that it is only today that the tools and software capabilities necessary to benefit from parallelism!like multicore processors and advanced programming solutions!are made available.

Some of the emerging parallel technologies include:

  • Single-core to multicore processors

  • Single thread to multithreaded applications

  • ASICs to FPGAs

  • Shared bus (e.g. PCI) to point-to-point buses (e.g. PCIe)

  • Software drivers available are multithread-safe

Multicore, multithreading
According to Nair, there has always been a need for a computing platform that can meet the high performance demand of consumer and industrial products. Previously, CPU clock rates were increased to achieve higher processing power but this also increased the amount of heat dissipated in the system. To address these needs, processor chipmakers such as Intel and AMD switched to multicore technology, which proved beneficial in providing higher performance computing and lower system power.

However, there are still some issues with the effective use of multicore technology. Nair noted that the biggest challenge on the shift from single-core to multicore is the ability to develop parallel competence, exchange sequences and increase performance. Integral to the solution for this challenge is the software. "Most technology thinkers in computing science recognize that the issue does not lie in hardware. It is the software that can help designers optimize the use of these new parallel technologies," said Nair.

In addressing this need, Nair said the National Instruments LabVIEW graphical programming solution, which has been available for over 20 years, is inherently parallel and allows designers to use multithreaded and multicore applications extremely easily and analytically.

"The traditional way of sequential programming is writing huge blocks of textual code to implement your desired application. When using graphical programming solutions such as LabVIEW, it allows you to iconically represent your code!instead of programming textual code which takes more time and effort!and then you just 'cut and paste' your graphical code in the design plate for debug and deployment." Noted Nair, "Programming using the textual method is possible. It continues to be done but the problem is that it is complex and requires a lot of relearning and writing long codes."

Nair added that since LabVIEW added multithread capability since 1998, engineers can assign each part of the application for a specific component to run, say a CPU in a system.

Nair cited an example in terms of the optimal use of hardware capability. "When you migrate from a 1GHz single-core computer to a 2GHz machine, you can immediately see the performance improvement of the box in terms of running applications such as word processing. Now, when you shift from a dual-core to a quad-core machine, you'll see no difference and then you'll wonder: 'Why have I spent more money to get a higher-performance processor but I am not getting any inherent benefit?'"

Figure 1: To benefit from parallel technology, engineers must understand how to use multicore threads.

"The reason for this," according to Nair, "is that most applications usually have one CPU and one thread addressing those CPUs. To benefit from parallel technology, we must understand how to use multicore threads."

Engineering and scientific applications are typically on dedicated systems, which brings little multitasking. To increase performance, engineers and scientists must use threads to benefit from multicore processors (Figure 1)."Being able to divide and conquer your application effectively will help differentiate you from other solution providers and enable you to ride the wave of technology," said Nair.

"To fully take advantage of parallel control systems, it's not enough to have just one component say just multicore processors, point-to-point buses or software that can upgrade multithread apps. What is required is an entire system, such as the next-generation PXI Express, where you can make fully use of parallel processing technologies without any bottlenecks," said Nair.

Flexible FPGAs, bus
The other part of the parallel toolbox is the move from ASICs to FPGAs.

Noted Nair, "ASICs are widely used in design but as technology changes, the need for easily, highly reconfigurable silicon is extremely important. FPGAs are reconfigurable yet, are truly parallel, very highly deterministic solutions.

NI engineers YH Goh and Cindy Ong demonstrate the latest PXI solutions from NI for RF, TiVO and helmet crash test apps.

All the algorithm or logic runs on the FPGA silicon, it can conduct independent operations, and it can import IP cores or proprietary cores, allowing a variety of IP to be processed in silicon and in parallel. In brief, FPGAs offer efficiency in concurrent applications.

However, this flexibility comes with a cost. Hiring an FPGA programmer is not easy because out of 100 engineers/technology graduates only one or two are proficient in VHDL!the traditional language for programming FPGAs.

According to Nair, NI's LabVIEW adopts the textual VHDL paradigm in programming FPGAs but you implement it graphically: you draw the logic, add IP blocks via the analysis functions available and download them into the FPGA. This allows engineers to inherently learn how to use these parallel blocks in FPGA very easily ad employ the latest, next-generation hardware in their design projects. Proof to this is a LabVIEW-controlled desktop computer numerical control (CNC) mill built by a Singapore non-engineer for $12,000. (The case was documented and received a special mention award in NI's annual technical paper contest.)

Meanwhile, why is there a need to move from shared buses to point-to-point buses?

"Let's take the PCI bus, with 132Mbit/s data rate, as example," said Nair. "Normally, you have multiple devices connected to the computer PCI bus. Consider the attached devices have 132Mbit/s data rates. If all the devices will transfer data to the CPU, they have to share the 132Mbit/s bus!each device cannot work optimally because there is a bottleneck. When you use point-to-point buses, you have dedicated pathways each at very high-bandwidth connected directly to the processor."

Algorithm engineering
Apart from the parallel technologies and trends that you see, Nair said that the converging technologies in devices drive the need for a modular test approach. "Given the evolution and rapid change of standards like in the wireless cellphone industry, it is highly demanding for test engineers to change boxes as they to move from GSM, 3G to maybe, WiMAX tomorrow. In the modular approach, engineers can use the same box to test these different standards!they program the software to suit the standard they want to test," he said. He cited the NI PXI solution which can be programmed as an RF, GSM, GPS or TiVO tester.

Back in the 1970s, we had a very measurement-centric world!it was all about box instruments and what the vendor defined, said Nair. He explained that the user then has a very limited say in what the box can do. When virtual instrumentation though LabVIEW DAQ and PC-based control systems was introduced in the 1980s, this changed. As the world became more application-centric, instruments have become more user-defined, in which users in a very modular way put the software and hardware together to implement their desired measurement, control and test systems.

"Today, as we move toward a design-centric world, there is an increasing need among companies and the academia for a common platform that can be used throughout the design phase, prototype and test phase," Nair said. With a common platform, engineers can increase their performance and reduce design time while achieving greater flexibility in the organization and allowing teams to refocus on different platforms, he said.

According to Nair, in the design centric world, almost every product requires some amount of algorithm, which is traditionally written in script-based language. The LabVIEW 8.5 now runs Mathscript, allowing engineers to write algorithm or import previously existing standard text-based Mathematics into LabVIEW, put it in a node and automatically run the algorithm. "This is algorithm engineering!the interfacing of the simulated world with the real world that allows the engineers to rapidly prototype and deploy systems. In today's world, whether you're in industry or academia, design-to-market is key," explained Nair.

NI engineer Victor Leong demos the Rubik cube solver (inset) that is a simple application of algorithm engineering.

In introducing the concept of algorithm engineering, NI has created a Rubik cube solver based on traditional cube algorithms. The solution was implemented using NI Vision Development Module, LabVIEW toolkit from Lego Mindstorms and a camera. The USB Web camera captures the image of the top cube face, in which the image is processed with the NI module and the instructions sent to the Lego Mindstorm hardware to execute the moves to complete the Rubik cube.

2008 roadmap
NI has made a handful of industry tie-ups in 2007, namely the membership in Power.Org and The Multicore Association. Nair said both partnerships are beneficial. For The Multicore Association, they will help mandate that multicore processors, from the hardware perspective, can be used effectively with the software technology NI offers. As for the Power.Org, NI will be involved in implementing the Power Architecture on instruction sets at the computational level (software).

In February, NI has completed its acquisition of long-time partner Microlex Systems, an early adopter of standard hardware modular platforms for video and audio. "With Microlex, we aim to bring the core IP blocks for advanced sound and video measurement into our analysis library and make them more easily available to our customers. On the research side, we now have some of the best engineers in the field of audio and video and we aim to innovate on emerging standards like HDTV, and bring those into the measurement capability provided through LabVIEW and other hardware platforms."

This year, the company aims to establish more footprint in Asia.

"The instrumentation industry, the industry in general is very interesting," said Nair. "The U.S. is seeing a slow-down and as for Europe, analysts forecast a slowdown. But we as a company see no slow-down in Asia!the region's growth accelerated more than what economists predict. And for us, now is a very good time to continue growing in Asia." In line with this, NI's new marketing arm National Instruments Philippines in Manila will start operations by midyear.

Nair added that NI sees potential in the Philippines for marketing and consulting. This February, NI has signed an agreement on the joint academe and industry test technology curriculum development with the Philippine's Ateneo de Manila University, Electronics, Computer and Communications Engineering Department (ECCE), School of Science and Engineering, Loyola Schools.

In an interview with EE Times-Asia, Jimmy Apolinar, professor and Celso Co, assistant professor, both from ECCE said that both parties are in talks on the use of NI LabVIEW and SW/HW tools for training their students in microelectronics. NI has also discussed the same possibility with San Carlos University in the Philippines. During the NIDays Manila event, Apolinar and Co presented a cost-effective workbench built with a single channel scope and two-channel digital I/O using NI tool USB 2008. Both academicians aim that other Philippine schools will replicate the cost-effective set-up.

For other countries, Nair said Vietnam is exciting, "a market that is very similar to what China is in the early days: providing the benefits of low-cost and highly trained people." He added that the government policies are also pro-business, coupled with the drive to invest on R&D. In Singapore, Nair said there is a marked shift away from manufacturing to academic research. In Malaysia, Nair said the market is growth oriented in consumer electronics, electronics manufacturing test and infrastructure acquisition which is basically using model tools of DAQ and analysis for bridge/earthquake monitoring, and oil and gas developments. For India, Nair noted that R&D remains extremely strong coupled with the influx of companies for manufacturing operations.

- EE Times-Asia

Article Comments - NI: Path leads to parallel for test ...
*? You can enter [0] more charecters.
*Verify code:


Visit Asia Webinars to learn about the latest in technology and get practical design tips.

Back to Top