qcd.txt

A C++ class library for scientific computing

The most expensive routines in terms of CPU time were a family of routines
that multiplied a 2-spinors by SU(3) gauge elements: The core of these
routines is basically the same as the following code:

COMPLEX M(V,3,3) res(V,3,2), src(V,3,2)
    DO spin=1,2
        DO col=1,3
            DO site=1,V
                res(site,col,spin)=
                         M(site,col,1) * src(site,1,spin)
                       + M(site,col,2) * src(site,2,spin)
                       + M(site,col,3) * src(site,3,spin)
            END DO
        END DO
    END DO

http://www.epcc.ed.ac.uk/t3d/documents/techreports/EPCC-TR96-03/EPCC-TR96-03.book_1.html