Prof Mark D. Hill had a paper published in ISCA this year titled Performance Pathologies in Hardware Transactional Memory. It's available here:
http://www.cs.wisc.edu/multifacet/papers/isca07_pathologies.pdf It's a worthwhile read. -j On Thu, Oct 25, 2007 at 04:41:48PM -0300, Rafael Vanoni wrote: > Hi everyone > > I'm here at SBAC-PAD (that's an International Symposium on Computer > Architecture and High Performance Computing). I've been attending > various interesting talks and though I'd share a couple with you. > > Earlier today I attended a talk by prof.Kunle Olukotun about > transactional memory, and got curious as to what this group might have > to say about it. He mentioned that his group is testing it with FPGAs > and getting some good results. > > Another curious thing, at the end of his talk, someone asked him what > would be the implications to operating systems in this world of > multi-core processors. He said that OS'es are picking up very slowly, > and that in Niagara's case, the scheduler was using a lot of cores just > to schedule things around when it first came out. He didn't mention > which OS. > > I'll try to talk to him personally tomorrow, couldn't get to him today. > > Yesterday, a guy from Google (and Luis Barroso, remotely) talked about > power consumption and efficiency for their warehouse-servers (or > land-held computers). He spent most of the time talking about how their > load average is at 30-45% most of the time and that the machines are > wasting electricity by not adapting to that. His numbers pointed that > the CPU is the most efficient component though, and that HDs and network > cards were just wasting power away. > > But as always, they showed the problem and where they would like to be, > but nothing about what they are doing to address those issues. Except > that they are spec'ing their own PSU's. According to their numbers, > regular PSU's waste at least half of the input current just converting > it to a usable level. > > If you have any thoughts on these topics, I'd love to hear it. > > > Rafael > > > > _______________________________________________ > perf-discuss mailing list > perf-discuss@opensolaris.org _______________________________________________ perf-discuss mailing list perf-discuss@opensolaris.org
