Bringing VxWorks support to Intel Xeon D-based EW systems

Story

August 07, 2018

Denis Smetana

Curtiss-Wright

The advantages that the Intel Xeon processor D-1500 product family brings to compute-intensive embedded electronic warfare (EW) system designs is clear. These 8-/12-/16-core devices deliver enhanced performance at low power, making them suitable for use on rugged open-architecture modules designed for deployment in harsh environment applications including electronic warfare (EW) and command, control, communications, computers, intelligence, surveillance, and reconnaissance (C4ISR). These system-on-chip (SoC) devices make large numbers of x86 processing cores readily available for embedded defense applications.

The advantages that the Intel Xeon processor D-1500 product family brings to compute-intensive embedded electronic warfare (EW) system designs is clear. These 8-/12-/16-core devices deliver enhanced performance at low power, making them suitable for use on rugged open-architecture modules designed for deployment in harsh environment applications including electronic warfare (EW) and command, control, communications, computers, intelligence, surveillance, and reconnaissance (C4ISR). These system-on-chip (SoC) devices make large numbers of x86 processing cores readily available for embedded defense applications.

Embedded EW system designers haven’t optimized the Xeon D processor’s many performance advantages in a Wind River VxWorks operating environment. No board support packages (BSP) provided software drivers to access the Xeon D’s QuickData Technology DMA engine, which frees the multicore processor from devoting critical resources to communications handling.

In the same way, no software drivers supported 40 Gigabit Ethernet (GbE) using Mellanox Ethernet controllers. High throughput, low latency, and determinism are all associated with the real-time performance required by sensor processors. The unavailability of software drivers necessarily drove system designers to turn to Linux, which, unlike VxWorks, is not a true real-time operating system (RTOS). For EW applications, many of which require highly accurate task handling down to the nanosecond, the robust determinism of VxWorks delivers precise timing and tight controls. In contrast, Linux supports less accurate probabilistic timing.

With an architecture designed to support math-intensive processing and very-high-bandwidth data transfers, Xeon D enables advanced cognitive EW applications to operate in small size, weight, and power (SWaP)-constrained platforms.

During the last five to ten years, while field-programmable gate arrays (FPGAs) have dominated EW system development, Intel processors were limited to a maximum of four cores. Until recently, x86-based general-purpose processors were used in EW systems only to provide system management or to handle the man-machine interface. The Xeon D, with its expanded multicore architecture – delivering four, 12, or 16 cores – enables the use of x86 devices as active participants in the prosecution of RF emitter stacks. Even better, because these devices are available in ball-grid-array (BGA) packages, their entire bottom surface can be used for interconnection pins, which is necessary for high-bandwidth operations for EW applications. Also, each of the Xeon D cores is supplemented by a powerful AVX2 SIMD [single instruction, multiple data] engine, which delivers enough processing power to execute the complex decision-making and high-bandwidth DSP math needed to run sophisticated EW algorithms.

The Xeon D’s QuickData Technology DMA engine is used to push data between multiple Xeon Ds – or to talk to GPUs or FPGAs – without having the Xeon D handle the communications burden. This arrangement enables data to be moved around with PCI Express or Ethernet without taxing the Xeon D, which instead is able to keep working undisturbed on the radar or other EW application while the DMA handles all of the low-latency, high-speed throughput transmissions.

Typically, without software driver support for the DMA engine, these communications chores would require the use of entire processor core, as well as some of the available memory bandwidth. By making it possible to use the VxWorks operating environment on solutions architected with the supercomputer-class Xeon D devices, EW system designers are able to leverage millions of previously developed lines of VxWorks software code, protecting their investment and eliminating the need to migrate to a different operating environment.

An example of software drivers that enable Xeon D-based high performance embedded computing (HPEC) systems for ISR applications to use VxWorks are some recently introduced new software drivers included in the Curtiss-Wright BSPs for use with Xeon D-based CHAMP-XD1 3U OpenVPX and CHAMP-XD2 6U OpenVPX DSP modules. When running VxWorks, these cards can deliver 40 GbE rated at ~37 Gbps, near line rate. (Figure 1.)

Figure 1: Curtiss-Wright’s Intel Xeon D processor-based CHAMP-XD1 and CHAMP-XD2 DSP modules are designed for use in VxWorks-based EW applications.