2011-01-08 13:45:56     regarding fract32 format

Document created by Aaronwu Employee on Aug 25, 2013
Version 1Show Document
  • View in full screen mode

2011-01-08 13:45:56     regarding fract32 format


Message: 97274   


I am trying to do SURF feature detection and matching. The code was taking more than one minute to execute on BF537.


I used fast floating point operation which reduced the execution time to half. Now I am using fixed point operation.


With fract32 1.31 format, data between +1to -1 can be represented. In my code, I have many operations in which some integers are multiplied with floating point data. I need to find a way in which data outside +1 to -1 range can be represent in fixed point format. Any suggestion?




2011-01-08 18:47:35     Re: regarding fract32 format

Mike Frysinger (UNITED STATES)

Message: 97275   


i'm not sure how many people you'll find here that are familiar with using the fract code.  you might want to try posting your question here:



this forum tends to be focused on the Linux side of things rather than pure Blackfin optimization.