poulson

Joined 13 years ago

Karma

218

Submissions

Member since

Jul 26

View on Hacker News

Recent submissions

Comment · 12 years ago

Academic research is funded through taxpayer money, and, as an academic, I can assure you that much of it has a much lower impact than many open source projects.

Comment · 12 years ago

I have no idea who those three people were. I do agree that "unprofessional" is a good way to describe their behavior. I emailed a few colleagues about it but it was all too strange to speak publicly about. There was also a bit of a religious vibe to their questions, so there is a (small) chance that they were evangelicals claiming to be interrogators.

Comment · 12 years ago

I had a somewhat similar experience the day shellshock was announced on a layover in Atlanta flying between two different universities. I had been browsing the news and coding for maybe an hour after dinner at the battery-charging station in my completely empty terminal, and three clean-cut men in their late thirties sat down on all sides of me and started joking about how I was in trouble.

Long story short, after a few minutes of awkward chit chat where they repeatedly try to determine whether I "believed in absolutes", they let me know that, yes, they were "professional interrogators". They then all immediately got up and left for what they claimed was their flight to Minneapolis.

Comment · 12 years ago

Good point. So it would seem to be hard to find solutions for 6, 10, 12, etc. colors. Have you given any thought to how to solve those cases?

Comment · 12 years ago

Thanks for the link! There is even a nice discussion of the explicit solutions for Inverter up to 7x7's on the page: http://mathworld.wolfram.com/LightsOutPuzzle.html

Almost nothing about the solver would need to change in order to support computation over GF(p), which suggests that someone should create a version of Lights Out / Inverter which cycles through a prime number of colors rather than just off/on.

Comment · 12 years ago

I've manually verified that the binary matrix printed out by the following Octave/MATLAB routine provides solutions to Inverter. (NOTE: You may want to use the gflineq.m implementation from http://lost-contact.mit.edu/afs/cs.stanford.edu/package/matl...)

  function M = Inverter(nx,ny)
  
  A=eye(nx*ny,nx*ny);
  for x=1:nx, for y=1:ny,
    i=x+(y-1)*nx;
    if( x ~= 1 ), A(i,i-1)=1; end;
    if( x ~= nx ), A(i,i+1)=1; end;
    if( y ~= 1 ), A(i,i-nx)=1; end;
    if( y ~= ny ), A(i,i+nx)=1; end;
  end; end;

  b=ones(nx*ny,1);
  
  s=gflineq(A,b);

  M=zeros(ny,nx);
  for x=1:nx, for y=1:ny, M(x,y)=s(x+(y-1)*nx); end; end;
  
  return

Comment · 12 years ago

A more effective approach might be to maintain a peer-reviewed ranking of various large tech companies based upon their patent history and to semi-regularly publicly base your employment decisions upon said list.

Comment · 12 years ago

Or, "Brian builds distributed-memory computers". Despite the misleading title, I think such a project would be a great tool for a parallel computing class. Such a machine would be roughly the same cost as a textbook and would be much more rewarding than running MPI on a multicore laptop/desktop.

Comment · 12 years ago

Was there a typo in your message? Why would you be unable to afford a home that you could immediately put down a 33% down payment in cash for?

Comment · 12 years ago

Even if one buys the (demonstrably false) claim that all real-world problems are sparse, most sparse techniques (especially sparse-direct, and, to a limited degree, Krylov subspace methods) boil down to dense linear algebra on smaller matrices. When executing dense linear algebra on accelerators, it can be surprising just how carefully one must organize the computation in order to make the best use of the memory hierarchy. When I was writing these types of routines years ago, it was often beneficial to pre-/post-process certain operations, such as A^T B^T = C, by explicitly transposing either the input or output matrix (in my case, to ensure that reads from global memory could be coalesced in one of the inner loops).

With that said, an example of efficiently transposing dense matrices was one of the CUDA examples five years ago...