11

My machine's RAM is 3G on Windows XP and 'float32' data's precision is enough for my current application (based on Pandas 0.10 + NumPy 1.6.2). So I want to reset the default floating datatype to 'float32' instead of 'float64' to reduce the memory usage. How to do that ? (I know I can explictly put dtype = 'float32' in functions, but i just want a global setting)

And is there any bad side-effect if I do so ? (calculation speed, compatible of interacting with other modules, future upgrading etc..)

bigbug
  • 55,954
  • 42
  • 77
  • 96
  • I don't think there will be, at least according to these questions. Interesting question though. – Andy Hayden Mar 28 '13 at 10:06
  • Follow [this link](http://stackoverflow.com/a/5351486/832621) to a related question that also solves your problem. Other related questions: -[Python: Making numpy default to float32](http://stackoverflow.com/questions/5721831/python-making-numpy-default-to-float32) -[Floating point precision in Python array](http://stackoverflow.com/questions/5160339/floating-point-precision-in-python-array) – Saullo G. P. Castro May 02 '13 at 10:59

0 Answers0