 Posted Thursday, June 26, 2008 10:50 PM
|
|
|
|
Firstly, the bluebit matrix library is fantastic. I've been porting all my matlab code, and its been relatively painless.
Now to the question. Can I use floats for matricies instead of doubles? I'm in the position where I need to sacrifice precision for speed, and I'm looking for any way I can achieve that.
Thanks!
-redc
|
|
|
Posted Friday, June 27, 2008 2:09 AM
|
|
|
|
I am afraid this is not possible. Perhaps in the future we may implement a float (Single) matrix library but even so the expected performance improvement wouldn't be significant. It would mostly affect the maximum matrix size which could be used.
Trifon Triantafillidis | Lead Developer | 
|
|
|
|
|