Microsemi to showcase third-party IP offerings for machine vision applications using its cost-optimised, low power mid-range PolarFire FPGAs at embedded world

Microsemi Corporation, a leading provider of semiconductor solutions differentiated by power, security, reliability and performance, has announced it has expanded the third-party intellectual property (IP) offerings for its cost-optimised, low power, mid-range PolarFire field programmable gate arrays (FPGAs). With new support of artificial intelligence (AI)/machine learning IP and HDMI 2.0b interfaces, the company’s award-winning PolarFire device can now be used in industrial artificial intelligence applications which leverage the rich resources in the FPGA, particularly the large quantities of digital signal processor (DSP) math blocks and embedded RAMs.

“As the PolarFire FPGA product family has rolled out over the past year, we are seeing a broadening customer base adopt these devices for neural network based designs,” said Shakeel Peera, vice president FPGA marketing for Microsemi. “Our newly announced IP cores address these additional design requirements by leveraging the unique dot-product (DOTP) mode of PolarFire’s DSP block to perform the eight-bit integer convolutions used heavily in machine learning inferencing applications. In this mode, the DSP block can perform four nine-bit operations per clock cycle, making PolarFire an extremely resource and energy-efficient FPGA for machine learning applications.”

The newly expanded IP offerings make Microsemi’s PolarFire FPGAs ideal for a wide variety of machine vision applications within the industrial, medical imaging, retail, defence and automotive markets. The HDMI 2.0b IP, available through Microsemi’s collaboration with Bitec, enables displays up to 4K (ultrahigh definition) resolution, which can be used for AI applications such as retail advertising, smart mirror displays as well as traditional display designs like media servers and display walls. Machine learning or AI IP, offered via Microsemi’s collaboration with ASIC Design Services, is a key component for machine vision applications such as sensing objects, enabling applications such as surveillance cameras to detect faces or in retail for targeted advertising, as well as advanced driver assist systems (ADAS) applications allowing automobiles to detect cars, pedestrians or other objects.

“Our Core Deep Learning IP provides compelling machine learning solutions for numerous applications,” said Tony Dal Maso, chief executive officer of ASIC Design Services. “We blended our specialised FPGA design skills and deep learning expertise to produce an optimal solution, taking full advantage of the dot-product architectural feature of the new PolarFire FPGAs. Our solution will interest designers looking to reduce cost and form factor and optimise the performance/power ratio, an important metric in many applications.”

The low power characteristics of flash-based FPGAs, coupled with Microsemi’s FlashFreeze mode of operation and “instant on” capability, makes PolarFire a compelling platform for AI designs implemented on FPGAs. ASIC Design Services’ solution allows FPGA developers and machine learning experts to bridge the semantic gap between high-level model specification and FPGA design. For customers with limited prior experience, ASIC Design Services provides a full turnkey service including design and deep learning training, accelerating time to market.

Key features of the new machine learning/AI IP core include:

  • Full pipeline from convolutional neural network description to FPGA implementation
  • Network retraining for memory footprint minimization
  • Support for different network layers
    • Convolutional layer
    • Fully connected layer
    • Pooling layer
    • Activation layers
  • Convolutional layers can implement filters of any size and stride
  • Pooling layers supporting arbitrary kernel size
  • Support for padding
  • AXI interface for external memory

According to Tom Hackenberg, principal analyst at IHS Markit, AI, or neural network (NN), processing such as machine learning and deep learning is quickly replacing IoT in technology headline news. It is not a one-size-fits-all technology, but one for many applications in which FPGAs are used. Highly parallel, reconfigurable and often more reliable, this class of processors can provide a broad set of solutions enabling innovative and diverse AI solutions. Automotive safety, building automation, security and surveillance are examples of early adopters of NN techniques where FPGAs are expected to exhibit double digit compound growth rates from 2017 to 2021.

“We are excited to support Microsemi’s PolarFire FPGAs with our HDMI 2.0b IP core, as well as the option for HDCP 2.2,” said Andy Robertson, director, Bitec. “We believe the smaller size form factors of PolarFire will be of interest for many video and display applications.”

Key features of the new HDMI 2.0b IP core include:

  • Support for source and sink HDMI 2.0b
  • Video support for 8, 10, 12 and 16-bpc Deep Color operation
  • Supports ultrahigh definition
  • Compatible with digital video interface (DVI) and dual link DVI
  • Supports audio to 32-channel, 3D and multi-stream transport (MST)
  • Optional High-bandwidth Digital Content Protection (HDCP) 1.4 and 2.2
  • Optional Consumer Electronics Control (CEC) 1.4 and 2.0

The new IP cores for PolarFire FPGAs were brought to market by Bitec and ASIC Design Services via Microsemi’s Accelerate Ecosystem.


Check Also

Spirent and MultiLane join forces to enable 800G ecosystem

Spirent Communications, the provider of test, assurance, and analytics solutions for next-generation devices and networks, …