EEMBC Benchmark Targets Heterogeneous Processor Architectures for Automotive Vision, Compute, and Mobile Imaging

EEMBC Members to Collaborate on System Benchmark Addressing Performance Requirements for Applications Such as Autonomous Driving and Mobile Imaging

EL DORADO HILLS, CA-- Jun 16, 2016 - The Embedded Microprocessor Benchmark Consortium (EEMBC, pronounced "embassy") today announced its focus on a benchmark targeting the compute-intensive applications commonly implemented on embedded heterogeneous compute architectures. The benchmark will utilize real-world workloads that stress highly parallel applications such as automotive surround view and image recognition and mobile augmented reality. This benchmark will help system designers from the automotive, industrial, and mobile sectors select the optimal processing solution for their applications, and will provide silicon and intellectual property vendors with an equitable method in which to compare and compete with their product offerings.

Identifying the potential compute performance of a heterogeneous architecture is a daunting task with the currently available benchmarks, as they focus either on monolithic application use cases or on isolated compute operations. Real-world scenarios on these architectures require an optimal utilization of the available compute resources, such as the CPU, GPU, and hardware accelerators.

"Optimal use of heterogeneous architectures implies load balancing of the compute tasks and distribution of data across multiple compute resources and separate fine-tuning for their individual performance profiles. This requires intimate knowledge of the architecture of the individual compute elements and of the heterogeneous architecture as a whole," said Rafal Malewski, chair of EEMBC's Compute working group and senior graphics engineering manager at NXP Semiconductor. "EEMBC has set out to create a benchmark that assists in identifying the performance criteria of the heterogeneous compute architecture and in determining the true potential of the architectures for real-world application use cases."

"As one of the pioneers of the graphics industry, it's been very interesting to see GPUs evolve into a main component of heterogeneous compute architectures, often referred to as GPU-Compute. This evolution is being driven by highly-parallel, computer vision, deep-learning, and neural-net applications in market segments such as automotive, mobile, scientific, and industrial," said Dr. Jon Peddie, president of Jon Peddie Research. "The competition will be fierce amongst the many software and silicon providers targeting the heterogeneous computing industry, and there will be a great need to help sort this out with real-world benchmarks, such as the one being developed by EEMBC."

"All EEMBC's benchmarks are about being repeatable, verifiable, and certifiable, and this compute benchmark will be no different," said Markus Levy, EEMBC president. "To ensure consistency between compute implementations, EEMBC's compute benchmark's framework will utilize the popular Khronos™ OpenCL™ 1.2 Embedded Profile API, which is supported by most vendors providing a heterogeneous architecture. Once the OpenCL reference implementation is validated, the benchmark will be open for vendors to submit platform specific optimizations."

The Compute working group, led by Rafal Malewski, is comprised of EEMBC members with strong interest in the computer vision, autonomous driving, and mobile imaging markets. These include EEMBC members ARM, CodePlay, Imagination Technologies, Intel, Marvell, NXP, STMicroelectronics, Synopsys, Texas Instruments, Verisilicon, and others not mentioned here. Technical advisors to this working group include Jon Peddie of Jon Peddie Research and Professor David Kaeli of Northeastern University.

EEMBC encourages vendors and manufacturers to join the consortium's working groups to contribute to the definition and development of its next-generation benchmark suites. To join the Compute benchmark working group, or other working groups, contact Markus Levy. Visit http://www.eembc.org/compute/about.php for more details about this benchmark.

About EEMBC

EEMBC was formed in 1997 to develop performance benchmarks for the hardware and software used in embedded systems. EEMBC benchmarks help predict the performance and energy consumption of embedded processors and systems in a range of applications (e.g. autonomous driving, mobile imaging, Internet of Things, scale-out servers, and mobile devices) and disciplines (processor core functionality, floating-point, multicore, and energy consumption).

EEMBC members include Ambiq Micro, AMD, Analog Devices, Andes Technology, ARM, C-Sky Microsystems, Cavium, Codeplay Software, Cypress Semiconductor, Dell, Flextronics, Green Hills Software, Huawei Technologies, IAR Systems, Imagination Technologies-MIPS, Infineon Technologies, Intel, Marvell Semiconductor, Microchip Technology, Nokia, Nordic Semiconductor, NVIDIA, NXP Semiconductors, Realtek Semiconductor, Renesas Electronics, Samsung Electronics, Silicon Labs, Somnium Technologies, Sony Interactive Entertainment, STMicroelectronics, Synopsys, Texas Instruments, and Wind River Systems.