optimized daxpy kernel for increments != 1

This commit is contained in:
Werner Saar 2015-04-24 11:39:17 +02:00
parent 3a67daa954
commit 0273966abb
1 changed files with 21 additions and 0 deletions

View File

@ -101,6 +101,27 @@ int CNAME(BLASLONG n, BLASLONG dummy0, BLASLONG dummy1, FLOAT da, FLOAT *x, BLAS
}
BLASLONG n1 = n & -4;
while(i < n1)
{
FLOAT m1 = da * x[ix] ;
FLOAT m2 = da * x[ix+inc_x] ;
FLOAT m3 = da * x[ix+2*inc_x] ;
FLOAT m4 = da * x[ix+3*inc_x] ;
y[iy] += m1 ;
y[iy+inc_y] += m2 ;
y[iy+2*inc_y] += m3 ;
y[iy+3*inc_y] += m4 ;
ix += inc_x*4 ;
iy += inc_y*4 ;
i+=4 ;
}
while(i < n)
{