Increase Performance of CPU Workload

GPUs are great for certain workloads, but how can a person increase performance for the workload that remains on the CPU or if no GPU is available? I have recently been looking into various solutions to this problem. Intel's compiler and MKL libraries have shown very promising results on some of my projects. I would like to share what I have learned so far, as well as hear from other group members about any experiences they have had with these tools.

Intel has started a series of webinars on this topic. I posted a link to them in the discussion section of our site.

I would also like to allocate some time for group discussions on CUDA 5.5, or anything else people would like to talk about.

FYI - There may be a women's basketball game that night and I have not been on campus yet this year, so it may be worth looking at the parking options (or light rail if that is also available).

Parking details at U of M can be found at http://www1.umn.edu/pts/park/eventparking.html

"During events, there are a few things to note:

Some parking facilities change rates for special events. Event rates go into effect between two and three hours before the eventDuring events, only cash and checks are accepted (no credit cards)Check the PTS Event Calendar (http://www1.umn.edu/pts/park/eventparkingcalendar.html) for a consolidated listing of events. To bypass event rates, try these tips: Park at a facility that doesn't charge event rates: East River Road Garage, Weisman Art Museum Garage or 19th Avenue Ramp. With a free hangtag permit, evening students can pay the daily rate ($4 a day) at Church Street Garage when space is available, Monday - Thursday after 4:30 p.m. This is only available during special events. Permits are obtained at One Stop Student Services, 333 Science Teaching & Student Services, 222 Pleasant Street SE. Call 612-624-1111 for more information. "