[PyOpenCL] Wrong results for matrix-multiply.py

[Tobias Lindstroem Jensen]

When I run matrix-multiply.py from the examples directory and choose a Tesla
C1060 from my machine I get

GPU==CPU: False

print out also shows that the result is completely garbage, except for the first
row. If I change dimensions to

a_width = 4*block_size
a_height = block_size
b_width = block_size

I get correct results. 

Is it a known issue with this file?

/Tobias