EDA Q2 figures; AI at the edge; advanced headlights; full-chip ILT; unmanned ships

The global EDA industry is in good shape: according to the report just released by the ESD Alliance, worldwide revenue increased 6.6 percent for Q2 2019 to $2472.1 million, compared to $2318.5 million in Q2 2018. Revenue breakdown by product category shows that the biggest growth was in semiconductor IP, with a 19.7 percent increase. Revenue breakdown by region indicates a 5.5 decrease in the Americas and a 23.7 percent increase in Asia/Pacific.

EDA expertise, a good match for machine learning

Among the many factors impacting EDA revenue, the booming AI/ML market may have a positive effect. As Cadence’s Anirudh Devgan said at the DAC’19 Troublemakers Panel, “The good thing about machine learning is it’s computational. So, if you look at the software industry in the last ten years, most of the activity is in social media, which is not really computational. (…) When it is actually computational, it’s lot of matrix multiplies and all kinds of numerical analysis — which is in the sweet spot of traditional EDA expertise. (…) The EDA companies are very well suited to do machine learning in a fundamental way. (…) Inherently our [Cadence’s] core strength matches very well with machine learning.” Transcripts from the panel have been recently posted on the moderator’s website, John Cooley.

Anirudh Devgan. Image credit: Cadence

AI at the edge: CEVA, GOWIN, Synopsys, Infineon

And speaking of machine learning, news concerning AI at the edge abound this week. At the recent AutoSens show in Brussels, Belgium, CEVA announced NeuPro-S, its second-generation AI processor architecture for deep neural network inferencing at the edge. The company also introduced the CDNN-Invite API, a neural network compiler technology that supports heterogeneous co-processing of NeuPro-S cores together with custom neural network engines, in a unified run-time firmware. The combined solution is aimed at vision-based devices, including autonomous cars, smartphones, surveillance cameras, consumer cameras etc. Optimization enables NeuPro-S to achieve, on average, 50% higher performance, 40% lower memory bandwidth and 30% lower power consumption than CEVA’s first-generation AI processor.

Image credit: CEVA

GOWIN has released a solution called GoAI, providing acceleration for artificial intelligence at the edge on GOWIN FPGAs. GoAI provides full stack support to test and deploy AI inference solutions at the edge by connecting into Caffe and Arm CMSIS-NN frameworks. This allows users to train and test a model, quantize and retest the trained model on a microcontroller within the GOWIN FPGA and then accelerate their model in FPGA fabric to enable real-time performance. Connecting into popular ML software frameworks eliminates the need for proprietary software.

More news regarding AI at the edge are coming these days from the ARC Processor Summit Silicon Valley (September 19 in Santa Clara, CA). Synopsys introduced its new DesignWare ARC EV7x Embedded Vision Processor family with Deep Neural Network accelerator. The new processors integrate up to four enhanced vector processing units (VPUs) and a DNN accelerator with up to 14,080 MACs to deliver up to 35 TOPS performance in 16-nanometer FinFET process technologies. Synopsys’ MetaWare EV Development Toolkit is based on common embedded vision standards, including OpenVX and OpenCL C. The ARC EV7x Embedded Vision Processors targets ADAS, video surveillance, smart home, and AR/VR applications.

And Synopsys ARC EV Processor IP is part of the Parallel Processing Unit (PPU) that Infineon will integrate in the next generation of its Aurix microcontrollers for automotive applications. Already today, Aurix supports certain types of neural networks; however, the PPU’s performance will be significantly higher than that of today’s accelerators, enabling Aurix to process data from advanced sensors in real-time.

More than just high beam and low beam

But of course, car electronics is not just about AI. The upcoming International Symposium on Automotive Lighting (September 23-25 in Darmstadt, Germany) will draw attention on the increased sophistication of car headlights. An example of the opportunity offered by advanced automotive lighting solutions comes from the collaboration between Nichia Corporation and Infineon Technologies: together, the two companies are developing a high-definition (HD) light engine with more than 16,000 micro-LEDs for front light applications. Unlike current HD solutions, the new device will provide high resolution light to the entire field of view of the driver. For example, HD light can be used to warn the driver of hazards by highlighting people or objects on the roadside, or to project markings on the road to help the driver navigate through a construction site. The new HD light engine will employ micro-LED technology from Nichia and a new driver IC from Infineon. Production launch is planned for 2023.

The micro-LED light engine developed by Nichia and Infineon. Image credit: Nichia

To address the design challenges that are emerging in this sector, Synopsys has just released the new version of its LucidShape software for the design, simulation, and analysis of automotive exterior lighting products. LucidShape version 2019.09 gives designers new tools to create sophisticated freeform reflectors and dynamic pixel light headlamps, as well as enhanced tools for visualization.

Stitchless, full-chip inverse lithography technology

Advanced chip manufacturing technologies are also making news this week. D2S has introduced TrueMask ILT, a GPU-accelerated hardware and software system that enables IC manufacturers to implement stitchless, full-chip inverse lithography technology (ILT) for advanced-node designs in a single day. As explained in a press release, ILT is a computational approach to determine the mask shapes that will produce the desired on-wafer results. In particular, unconstrained curvilinear ILT can produce the best results. Until recently, two major obstacles have kept ILT from being widely applied. One of these barriers – the ability to write curvilinear mask patterns – was eliminated recently through the introduction of multi-beam mask writers. The second barrier – long runtimes due to the sheer volume of computations – has remained unresolved until now. Previous attempts to overcome this final hurdle involved dividing the chip into partitions and then “stitching” the partitions back together; this method, however, introduces errors that must be corrected. The stitchless, full-chip solution developed by D2S now enables users to leverage the power of multi-beam mask writing and the benefits of ILT.

Autonomous maritime navigation

As the automotive industry struggles to advance autonomous driving, it’s no surprise that similar concepts start being applied to other means of transportation. According to Chinese media, China has built the first test base for unmanned ships, which will be operational at the end of this year. The test base – named Xiangshan Marine Scientific & Technological Port – is in Zhuhai, Guangdong Province. It will create an innovation platform for the design, research and testing of unmanned marine systems and intelligent equipment.

Upcoming events

Let’s finish with some information about two upcoming events. The IEEE Artificial Intelligence Symposium will take place in San Jose, CA, on September 23. The TSMC 2019 Open Innovation Platform Ecosystem Forum will be held in Santa Clara, Ca, on September 26 and in Beijing, China, on October 29.

This entry was posted on Thursday, September 19th, 2019 at 6:40 pm. You can follow any responses to this entry through the RSS 2.0 feed. You can leave a response, or trackback from your own site.