Coffeehouse Post

Single Post Permalink

View Thread: Microsoft Research: Accelerator
  • User profile image

    Maybe it was a bit sneaky, but really only a bit.  Smiley

    Two things to consider:

    1) PCIe is much better than AGP and transferring data back.  In effect, the asymmetry has been erased.

    2) On the smaller benchmarks, the communication time is kind of irrelevant.  If convolving with a 5x5 filter is your entire computation, then it’s probably not worth parallelizing anyway.  For the larger benchmarks, take StereoMatch as an example.  The communication time adds < 25%.  That’s pretty small compared to a 4x speedup over  C++.