AEMB Benchmarked

AEMB Benchmarked

A recent Masters of Science project from TU Delft did some comparison and benchmarking of various open source microprocessors including the AEMB. They compared it against several architectures including the LEON3 (SPARC V8), Plasma (MIPS) and others. They found that in terms of the speed and resources, the AEMB emerged as the optimised option.

Although they finally dropped the use of the AEMB, they presented some very useful results in Chapter 2 of the thesis. Among some of their more interesting results are the the post-synthesis speeds and size (MHz/LUT). The results put the AEMB well ahead of the rest of the competition – 279 MHz/926 LUT. They also found some performance advantages in the Dhrystone and Fibonacci benchmarks.

AEMB-TUDELFT

This is in-line with our claim that the AEMB is the world’s smallest and fastest 32-bit multi-threaded RISC processor.

Unfortunately, the author of the thesis faced several problems during implementation of their design of the AEMB. The biggest problem that they had was that it was difficult for them to modify the AEMB to fit their requirements. The author needed to target it for an ASIC target while the AEMB was designed for an FPGA implementation, with design trade-offs made to optimise it for the FPGA.

Hopefully, others will have more success with the AEMB in the future. We are currently working on the next generation, which will introduce many architectural enhancements designed for increased parallelism.