Open Programming Model to Create Middle Ground for Embedded, Enterprise & IoT Developers
November 04, 2021
Working with real-time operating systems has historically been an esoteric endeavor – especially if the RTOS controls safety- and security-critical equipment like industrial machinery or advanced driver assistance systems (ADAS).
The embedded devices that these OSs run on are numerous, heterogeneous, and distributed, which means embedded engineers must have intimate knowledge of the target hardware, OS, firmware, and applications to ensure reliable, deterministic performance.
But the IoT is changing all of that. The modern developer, by contrast, is accustomed to programming in scripting languages and deploying their applications on homogeneous enterprise hardware infrastructure. All of this enables new development paradigms built on agility and speed such as “shift left” and write once, run anywhere (WORA).
Meanwhile, Data Bridge Market Research estimates a 6% CAGR in embedded systems through 2027 with AI-based technologies like advanced driver assistance systems (ADAS) and machine vision expected to grow at 21% and 34%, respectively. If the two disciplines hadn’t converged enough by now, they certainly will in the coming years.
Now, a partnership between eSOL and Codeplay is bridging those two realities. The two companies are combining Codeplay Acoran – an open software platform that integrates standards and open-source components like OpenCL, oneAPI, SYCL, SPIR-V, TensorFlow – and the eSOL eMCOS RTOS.
One Model to Program Them All
Essentially, the eSOL/Codeplay offering allows developers to port programs written in C++ directly into the POSIX-compliant eSOL eMCOS framework. This allows, for example, ADAS engineers to integrate AI and HPC workloads native to the data center into an embedded environment managed by the multi-kernel eMCOS RTOS.
Acoran makes this integration possible through heavy use of the Khronos Group’s SYCL standard. SYCL is a programming model based on ISO C++ that promotes heterogeneous programming across targets that include multicore processors supplied by, but not limited to, Intel, AMD, Nvidia, Renesas, Arm, and Imagination.
In addition, the Acoran programming model layers in optimized, domain-specific libraries that allow developers to build and deploy applications with confidence on top of device-specific optimized routines.
In short, the integration provides safe rendering of AI and HPC-accelerated applications in a real-time industrial and automotive embedded operational context. Today, the partnership will enable greater support for multicore CPU environments, but in the future will also support integrated AI and computer vision accelerators as well.
"The momentum has grown with SYCL, with organizations embracing SYCL for their next-generation processing systems, and developers favoring the open standards approach." saysMichael Wong, Chairperson of the SYCL Working Group within The Khronos Group and Distinguished Engineer at Codeplay Software.