<div dir="ltr"><br><div class="gmail_extra"><br><div class="gmail_quote">On Fri, Feb 27, 2015 at 9:02 AM, Mads Ipsen <span dir="ltr"><<a href="mailto:mads.ipsen@gmail.com" target="_blank">mads.ipsen@gmail.com</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">Hi,<br>
<br>
If I build Python 2.7.2 and numpy-1.9.1 and run the following script<br>
<br>
import numpy<br>
c = numpy.ones((500000, 4))<br>
mat = numpy.identity(4)<br>
r = numpy.dot(c, mat)<br>
<br>
the evaluation of the 'dot' increases the memory by app. 35 MB.<br>
<br>
If, in addition, I build numpy-1.9.1 with MKL support, and run the<br>
script, the evaluation of the 'dot' increases the memory by app. 450 MB.<br>
<br>
Is the expected?<br>
<br>
Best regards,<br>
<br>
Mads<br>
<br>
specs:<br>
<br>
Ubuntu 12.04<br>
ifort (IFORT) 14.0.1 2013100<br>
gcc version 4.6.3 (Ubuntu/Linaro 4.6.3-1ubuntu5)<br>
<span class="HOEnZb"><font color="#888888"><br></font></span></blockquote><div><br></div><div>No, but I don't know why that is happening with MKL. Can anyone else reproduce this?<br><br></div><div>Chuck <br></div><br></div></div></div>