Re: [Numpy-discussion] dot and MKL memory

2015-02-27 Thread Mads Ipsen
On 27/02/15 17:50, Charles R Harris wrote: > > > On Fri, Feb 27, 2015 at 9:02 AM, Mads Ipsen > wrote: > > Hi, > > If I build Python 2.7.2 and numpy-1.9.1 and run the following script > > import numpy > c = numpy.ones((50, 4)) > mat = numpy.ide

Re: [Numpy-discussion] dot and MKL memory

2015-02-27 Thread Daπid
On 27 February 2015 at 17:50, Charles R Harris wrote: >> > > No, but I don't know why that is happening with MKL. Can anyone else > reproduce this? > > Chuck I can't reproduce. I have checked on my system python (ATLAS) and Conda with MKL running in parallel with vmstat. The difference between th

Re: [Numpy-discussion] dot and MKL memory

2015-02-27 Thread Charles R Harris
On Fri, Feb 27, 2015 at 9:02 AM, Mads Ipsen wrote: > Hi, > > If I build Python 2.7.2 and numpy-1.9.1 and run the following script > > import numpy > c = numpy.ones((50, 4)) > mat = numpy.identity(4) > r = numpy.dot(c, mat) > > the evaluation of the 'dot' increases the memory by app. 35 MB. >

[Numpy-discussion] dot and MKL memory

2015-02-27 Thread Mads Ipsen
Hi, If I build Python 2.7.2 and numpy-1.9.1 and run the following script import numpy c = numpy.ones((50, 4)) mat = numpy.identity(4) r = numpy.dot(c, mat) the evaluation of the 'dot' increases the memory by app. 35 MB. If, in addition, I build numpy-1.9.1 with MKL support, and run the scr