compiler -r16 -r8 lapack95

[diedro]

hi everyone,

I do not why but I have the following problem:

I wrote a program as:

[bash]SUBROUTINE quadraturepoints

REAL    (R16P),  DIMENSION(nGP1D)        :: sGP1D
REAL    (R16P),  DIMENSION(nGP1D)        :: wGP1D

...
...

 Inv=DBLE(Coeff1D)
 CALL GETRF(Inv,IPIV,info)  !lapack95

....
.....


[/bash]

then I compile as:

[bash]ifort *.f90 -L$MKLPATH/lib/em64t  -lmkl_blas95_ilp64 -lmkl_lapack95_ilp64 -lmkl_intel_ilp64 -lmkl_sequential -lmkl_core -lpthread[/bash]

and I have not problem

However I write again the program as

[bash]SUBROUTINE quadraturepoints

REAL    ,  DIMENSION(nGP1D)        :: sGP1D
REAL    ,  DIMENSION(nGP1D)        :: wGP1D
REAL    ,  DIMENSION(nGP1D)        :: xiCel
REAL    ,  DIMENSION(di,nGP)       :: xCel


.....

  CALL GETRF(Inv,IPIV,info)  !lapack95

....

[/bash]

and this time I compile as:

[bash]ifort -r16 *.f90 -L$MKLPATH/lib/em64t  -lmkl_blas95_ilp64 -lmkl_lapack95_ilp64 -lmkl_intel_ilp64 -lmkl_sequential -lmkl_core -lpthread[/bash]

and I get the following error:

[bash] There is no matching specific subroutine for this generic subroutine call.   [GETRF]
  CALL GETRF(Inv,IPIV,info)  !lapack95
[/bash]

I try also with

[bash]-r8[/bash]

but I get the same error

What is the problem?

What's wrong?

Thank you very much

[mecej4]

There is no support in the standard BLAS and LAPACK libraries for performing linear algebra operations on REAL*16 arrays.

For example, to do the y <= a.x + y operation, one calles saxpy for REAL arguments and daxpy for double precision arguments, but nothing is provided for higher precision.

If you look at the LAPACK sources, you will find build instructions for

make single
make single double
make single double complex
make single double complex dcomplex

but no

make quadruple

[diedro]

hi,

thank you

so I will compile i double precision with r8