LAPACK/ScaLAPACK Development

Hello, all.

I tested DGETC2 subroutine that is LU factorization with complete pivoting of general matrix,
and it is slower than my subroutine.
In dgetc2.f, the subroutine find maximum element by row major order.

Code: Select all

         DO 20 IP = I, N
            DO 10 JP = I, N
               IF( ABS( A( IP, JP ) ).GE.XMAX ) THEN
                  XMAX = ABS( A( IP, JP ) )
                  IPV = IP
                  JPV = JP
               END IF
   10       CONTINUE
   20    CONTINUE

I think DGETC2 will be faster if it find by column order and use IDAMAX in BLAS.
Are there some reasons for not using BLAS subroutine ?

Thanks in advance.