The BenchPress
The Original Embedded Microprocessor Benchmark Newsletter, from EEMBC
Issue #65 - Q3'2021

Topics


IoTMark™-Wi-Fi Released

This summer, our IoT Connectivity working group released IoTMark-Wi-Fi, rounding out our RF benchmark capabilities for IoT by adding an 802.11 suite. This software represents a significant achievement in the realm of benchmark analysis, both from the technical construction of the testbed framework, and the collaborative decision-making process that established a meaningful behavioral model. We've moved far beyond the self-hosted compile-and-run C code of yesteryear, to an environment that now assumes a baseline of electrical and RF isolation, as well as a highly modularized porting API. In return, we now offer tools for deeper insight into the energy cost of Wi-Fi IoT devices' connected standby and low-bandwidth secure communication.

"For customers to make informed decisions, the industry needs a way to make objective measurements of battery life for IoT products,” says Omer Cheema, the Senior Director of IoT Wi-Fi Business Unit at Renesas (formerly Dialog Semiconductor). “That’s why [we've] been so excited to collaborate with EEMBC in developing a standardized low power Wi-Fi benchmark."

CoreMark-PRO Bare-metal Tutorial Video and Application Note

We've listened to requests from our customers, and released an application note and a video tutorial explaining how to port CoreMark-PRO to bare-metal on a remote target. By default, CoreMark-PRO ships with a make-based build system using GCC tools. This video shows how to correctly run CoreMark-PRO on a Linux system (Ubuntu) in single- and multicore-modes, as well as detailed walkthrough explaining how to port the multi-instance test harness (MITH) and one workload to a remote single-thread embedded target, using an STMicroelectronics STM32H747 and the IAR EWARM compiler system for reference.

Cockroach Labs Selects CoreMark for Their In-Depth Cloud Report

Cockroach Labs, the company behind CockroachDB, compiled a comprehensive report comparing AWS vs Azure vs GCP: "The 2021 Cloud Report runs over 1,000 microbenchmark tests to evaluate CPU, network, storage, and TPC-C performance. We want to help others understand the performance and cost tradeoffs of each cloud and its machines." CoreMark was selected as the performance benchmark because it is "open-source, cloud-agnostic, and more representative of general CPU performance ... CoreMark tests against various real-world workloads."

The report is available on their website.

Academics and Non-profits Around the World using EEMBC Benchmarks

EEMBC welcomes the following institutions using our benchmarks for their research:

Score Certifications

EEMBC offers a certification program that verifies the results of a benchmark. While most of our benchmarks self-check, some Run Rules cannot be explicitly enforced remotely. The certification process is performed at the EEMBC lab, and recreates the scores on the actual platforms. Here we scrutinize the implementation with logic analyzers and power probes to verify correctness, even going so far as to hand-analyze assembly code if a custom compiler is used. EEMBC certification guarantees that a score is valid.

Since the last newsletter, several ULPMark and CoreMark scores have been published:

During the first half of the year, Renesas certified almost two dozen CoreMark scores for the RA and RX class processors. The scores can be found on the CoreMark score page (filter for certified).

Before merging with Renesas, Dialog Semiconductor submitted the first-ever certified IoTMark-Wi-Fi score of 815 on their DA16200 device.

STMicroelectronics followed up their earlier ULPMark-CoreProfile score with two new ULPMark-PeripheralProfile scores on the STM32WLEx/5x, scoring 73.9 at 3.3V and 138 at 1.8V.