Doubt the average computer user would even know what SSE2 means ;>) There's nothing to correct though, maybe a footnote to add and ideally an automated test on devices CPUs if SSE2 is absent to reduce the fail rate, plus an event log message to curtail frustration. But maybe that's utopian.

BTW, posting result links still does not make them viewable at WCG.

KLiK,

We had made a decision back when we launched UGM to compile the code with SSE2 enabled. This sped up the results being completed to a factor of 5 if I remember correctly (could be more or less but it was significant). When I did a scan on the number of active devices by processor type, there was less than 150 pentium 3 or older devices. Thus by adding in the verbage to the minimum requirements page did not seem needed. We did attempt to communicate that SSE2 was used in our initial postings in the forums and maybe news articles. I will see what others on the team think about adding it into the minimum requirements page.

Thanks,
-Uplinger

Just out of curiosity, was any improvement noticed (if it was even tested) for any newer x86 extensions, and if so was the reason for not enabling them too many active CPUs that don't support them :P?

Uplinger,

Have you tried compiling it with AVX or AVX2? Presumably that could cut the runtime in half as the compiler you are using appears to vectorize critical sections of the UGM code.

We do try other compile options during early alpha. But the science code had a specific path for using SSE2 which allowed it to optimize the runtime greatly.

Sek, I had thought it was 20x increase, but couldn't remember 100%. Thanks for pulling that data.

Thanks,
-Uplinger