How to force python float operation on float32 rather than float64?

Question

I want to do some math operations (+, -, *, /) on float32 rather than on float64 type. I need do these operations on number or numpy.array, and also some numpy math functions, such as sqrt mean. How do I do this?

http://docs.scipy.org/doc/numpy/user/basics.types.html

atupal
– atupal

2013-12-25 03:01:36 +00:00
Commented Dec 25, 2013 at 3:01 — atupal
– atupal, Commented Dec 25, 2013 at 3:01

WKPlus · Accepted Answer · 2013-12-25 03:05:23Z

16

Will numpy.float32 help?

>>>PI=3.1415926535897 >>> print PI*PI 9.86960440109 >>> PI32=numpy.float32(PI) >>> print PI32*PI32 9.86961

If you want to do math operation on float32, convert the operands to float32 may help you.

answered Dec 25, 2013 at 3:05

WKPlus

7,3352 gold badges38 silver badges55 bronze badges

Sign up to request clarification or add additional context in comments.

2 Comments

Samuel Over a year ago

Yes. numpy.float32 can help me. Are there any other methods to do this

exogeographer Over a year ago

Yes, when using an array you can call it's astype method if it is not float32: docs.scipy.org/doc/numpy/reference/generated/…

Danijel · Accepted Answer · 2020-09-28 16:00:38Z

Use numpy.ndarray.astype:

import numpy as np arr_f64 = np.array([1.0000123456789, 2.0000123456789, 3.0000123456789], dtype=np.float64) arr_f32 = arr_f64.astype(np.float32)

Pay attention to precision:

np.set_printoptions(precision=16) print("arr_f64 = ", arr_f64) print("arr_f32 = ", arr_f32)

gives

arr_f64 = [1.0000123456789 2.0000123456789 3.0000123456789] arr_f32 = [1.0000124000000 2.0000124000000 3.0000124000000]

Collectives™ on Stack Overflow

How to force python float operation on float32 rather than float64?

2 Answers 2

2 Comments

Comments

Hot Network Questions

Collectives™ on Stack Overflow

2 Answers 2

2 Comments

Comments

Related