Thanks Matthias for your prompt response.
Rana
On Thu, Jun 18, 2020 at 9:00 PM Rana Alotaibi wrote:
> Hi SystemML developers,
>
> I have been experimenting with SystemML lately. I found some interesting
> numbers when evaluating these two expressions: *t(M%*%N)* and *t(N)%*%
> t(M)* (The size
Thanks for the question and the detailed inputs - this is an effect of
simplification rewrites that only apply in one of the cases.
Specifically, (t(N)%*%t(M))[1,1] is rewritten to t(N)[1,] %*% t(M)[,1],
which is a form of selection pushdown. You can do the following, for
benchmarking*:
Hi SystemML developers,
I have been experimenting with SystemML lately. I found some interesting
numbers when evaluating these two expressions: *t(M%*%N)* and *t(N)%*% t(M)*
(The size of matrix M is 20Kx100 and N is 100x20K). The pipeline *t(N)%*%
t(M) *drastically outperforms* t(M%*%N)*.
-
