M x S x T(M) multiplication with BLAS

Asked Nov 08 '12 at 15:40

Active Nov 17 '12 at 19:55

Viewed 89 times

Possible Duplicate:
Efficient way of computing matrix product AXA'?

I have two double precision matrices:

M, which has size l ⨯ m and no particular property;
S, which has size m ⨯ m and is symmetric.

I would like to compute:

R = M ⨯ S ⨯ t(M)

where t(M) is the transpose of M.

Looking at the list of (C) BLAS routines I didn't find anything that directly addresses this case. What is the most efficient combination of routines, then, that would allow me to compute R?

edited May 23 '17 at 10:26

Community

asked Nov 08 '12 at 15:40

akborder

2

Do you think, that there is something more effective then DSYMM and DGEMM (`P=MxS; R=P*t(M)`)? – osgx Nov 08 '12 at 15:56
*GEMM + *SYMM seems like the best you can hope to do. – Stephen Canon Nov 08 '12 at 15:59

M x S x T(M) multiplication with BLAS

0 Answers0

Linked