e i g h t - c u b e d . c o m 

AEST and performance, part 2

Following on from a previous post, while I was reverse engineering the module, the business was hunting down the original source code, and (not too surprisingly) they got it from the ex-directors of the company that originally wrote the code.

The graph is a measure of utilization of four CPUs with approximately the same workload, with three diferent versions of the module. Reading left to right, we have the binary translated version produced by AEST, followed by the natively compiled version of the original source code, and lastly, a version of the module with alignment faults corrected.

A picture sometimes speaks volumes.