btw, I just ran ftq on t420 on 9front. ftq has the option to wire to a core: I wired to core 2.
This is a full desktop, with rio, wifi stack, and so on. The results from ftq on 9front are mindboggling. One simple figure of merit we now develop is "if I had all the machine all the time, how much would I get?" ftq prints "how much did I get as a fraction of how much could I get?" It is never 1. It is rarely .98. On VMs it sits around .91. There's always interference On 9front, wired to core 2, it is .994. This shocked me. The amount of careful finagling you have to do on linux to get close to this number is ... a lot. And even on the almost-fastest-machine in the world, Frontier, they still have to work hard to quiet the OS environment down. Also, sadly, right now, on NIX, the AC run I just did only gets .976, so .. I need to look a bit more. It used to be much better, but this could be a flaw in the port. It will be interesting to see if grap can handle 1/2 million data points. So far, acme struggles with this file. ed just says ?TMP ------------------------------------------ 9fans: 9fans Permalink: https://9fans.topicbox.com/groups/9fans/Tb8c8d6c502962279-M293b0f2409ff296b6258d53b Delivery options: https://9fans.topicbox.com/groups/9fans/subscription
