#15 Matrix multiply yields segfault

[David Grangier]

require 'torch'
require 'lab'

x=torch.Tensor(10,15,3)
for i=1,10 do for j=1,15 do for k=1,3 do x[i][j][k] = k end end end
y=lab.reshape(x,150,3)
m=lab.mean(y)

z=y*m:t()

[Raoul d'Andresy]

Logged In: YES
user_id=1651867
Originator: NO

Blas is sensitive to strides in a matrix, even if the number of columns is 1. THBlas has been changed to handle this particular case. Problem fixed.