complex*32 and real*16

[paolodenti]

SUBROUTINE LUCINQP(AM,NDDMST,N,IER)

DIMENSION AM(40,40),V(40)

COMPLEX*32 AM

REAL*16 V

REAL*16 SUM

DO 1309 I=1,N

SUM=0.0Q0

DO 1319 J=1,N

1319 SUM=SUM+REAL(AM(I,J))**2+IMAG(AM(I,J))**2

1309 V(I)=1.0Q0/SUM

return

end

gives the message

t.f(1): warning #18000: The specified alignment (32) cannot be honored. Instead an alignment of 16 is used

Instead,

dimension a(1000,1000),b(1000)

complex*32 a

real*16 b

do 1 i=1,1000

b(i)=0.0q0

do 1 j=1,1000

1 a(i,j)=(0.0q0,0.0q0)

b(611)=qacos(0.0q0)*2

a(27,311)=(1.0q0,-2.0q0)*.5q0

write(11,*)sqrt(a(27,311))

write(11,*)b(611)

stop

end

returns no message and works fairly in q-precision.

??????

[TimP]

Quoting - paolodenti

SUBROUTINE LUCINQP(AM,NDDMST,N,IER)

DIMENSION AM(40,40),V(40)

COMPLEX*32 AM

REAL*16 V

REAL*16 SUM

DO 1309 I=1,N

SUM=0.0Q0

DO 1319 J=1,N

1319 SUM=SUM+REAL(AM(I,J))**2+IMAG(AM(I,J))**2

1309 V(I)=1.0Q0/SUM

return

end

gives the message

t.f(1): warning #18000: The specified alignment (32) cannot be honored. Instead an alignment of 16 is used

I don't get any message with ifort 11 beta Intel64. You must have used a different compiler; how is anyone to guess? Are you implying that it didn't work? By the way, you are lucky if non-standard quad precision in the style of 20 years ago is working for you. In particular, use of REAL this way could be ambiguous.

[Steven_L_Intel1 · Accepted solution]

This is a bug that is fixed in version 11.