Fusion RTOS Compare

Compare Performance

In the July 2000 issue of Embedded Systems Programming, an article titled "Commercial RTOSes for Automotive Applications" evaluated 10 operating systems running on a 16 MHz Motorola 68332.

The test was duplicated with the Fusion RTOS running on a Motorola 56805 DSP running at 16 MHz. Even though this is a different processor used in the tests referenced in the article, this was a controller type application running on a DSP. While DSPs are not known for their strength in microcontroller applications, The Fusion RTOS takes advantage of the performance difference Motorola has created in the design of the 56800 core to allow it to be used in these applications. The "OS PERFORMANCE" chart shows how well the Fusion RTOS performed compared to the other 10 Operating Systems. However the 56800 family is not limited to 16 MHz. Most members of the 56800 family run at 80 MHz.

The 10 Operating systems tested in the evaluation are listed below. The order of the companies do not correlate to the names in the chart. The intent is not to put down a company or kernel, but to demonstrate the performance capabilities when combining the Fusion RTOS with the 56800 family of DSPs.

*Company*	*Kernel*
Accelerated Technology	Nucleus
CMX	Tiny+
Enea Data	OSE Classic
JMI	JMI
Microtec Research	VRTXmc
U.S. Software	MTASK
Integrated Systems	pSOSelect
Embedded Power	RTXC
Embedded Power/Motorola	RTEK
University of Michigan	Emeralds

Below is a chart from testing that describes the performance of the different operating systems. The test has an interrupt that triggers a task to execute. The interrupt rate is then increased and the CPU usage of the OS is measured. This is the percentage of processor bandwidth used by the OS to switch into a task that was started from an interrupt. The chart shows an increasing interrupt rate and the percentage of CPU usage to switch into the task.

The kernel ram usage section of the evaluation is where the Fusion RTOS excels. This section of the evaluation adds an ISR signaling a unique task. The evaluation then compares the ram usage as each task is added. With the capability of the LRTASK only two stacks are needed plus 28 bytes for each added LRTASK. 10 LRTASKs plus one standard task, two stacks and 32 core kernel bytes adds up to a total of 852 bytes. The next best OS used nearly 3000 bytes.