Yeah? Still no benchmarks in the article, just generalizations like 10% (which is how much they cut the number of operations). It is good that another guy here referenced the original report, where there are actual benchmarks. The so called Strassen algorithm is actually slower on CPU than the standard algorithm for reasonably sized matrices, no matter that it performs significantly smaller number of operations. Mind blowing, huh? Seems computers are no so simple after all.

So, my question was pretty reasonable. Stop being lazy and learn something instead of insulting people who do.