Once upon a time there was a emerging technology called CUDA, which offered all sorts of really intriguing new possibilities in scientific and parallel computation. And once upon a time, Stack Overflow was full of interesting questions about CUDA, and how to use it. So I started answering them. Eventually I answered almost 700 questions, became Stack Overflow's highest reputation participant on the CUDA tag, and had a lot of fun doing it.
Alas, CUDA is now very mature and most of the good questions about CUDA have already been asked and answered. What appears on Stack Overflow today is mostly dross, and I spend most of my time editing, down-voting and closing rather than answering questions. Those answers I add are community wiki entries (over
200 300 400 500 600 700 at the time of writing). A lot of toil has gotten and kept the unanswered question queue down to about 10% 7% 4% 3% of the total number of CUDA questions for a good part of my tenure here.
"People like Coldplay and voted for the Nazis, you can’t trust people Jeremy."
Member for 7 years, 1 month
1 profile view
Last seen Mar 19 '19 at 17:53
Top network posts
- 304 What is the canonical way to check for errors using the CUDA runtime API?
- 258 What is the canonical way to check for errors using the CUDA runtime API?
- 148 How do I choose grid and block dimensions for CUDA kernels?
- 138 Relationship between SciPy and NumPy
- 99 How to get the nvidia driver version from the command line?
- 97 Should I unify two similar kernels with an 'if' statement, risking performance loss?
- 86 allocating shared memory
- View more network posts →
Keeping a low profile.
This user hasn't posted yet.
Sep 24 '14