Re: Matrix Multiplication in BLAS or Lapack
beliavsky_at_aol.com
Date: 10/23/04
- Next message: Andreas Krebs: "Re: Matrix Multiplication in BLAS or Lapack"
- Previous message: pascalv_at_despammed.com: "Re: Q:where can I find c++ numarical anaysis codes?"
- In reply to: Andreas Krebs: "Re: Matrix Multiplication in BLAS or Lapack"
- Next in thread: Andreas Krebs: "Re: Matrix Multiplication in BLAS or Lapack"
- Reply: Andreas Krebs: "Re: Matrix Multiplication in BLAS or Lapack"
- Messages sorted by: [ date ] [ thread ]
Date: 23 Oct 2004 04:01:43 -0700
Andreas Krebs <krebs.andreas@gmx.de> wrote in message news:<41761636.5000703@gmx.de>...
> The xgemm x\in\{s,d,c,z\} routines are members of the
> BLAS-library and not of LAPACK;
> so you need not to link the lapack library. But maybe your
> LAPACK library includes the blas routines already due to
> library linking. Then there would be no need to link with libblas.a,
> but this seems uncommon to me.
>
> But linking the lapack library should not disturb anything:
> if no lapack routines are needed, they will not be included.
>
> Some words about the LAPACK philosophy:
> LAPACK offers the high-level and driver routines, e.g. the
> dgeevx routine to compute eigenvectors.
>
> BLAS offers the generic and simple routines, e.g. daxpy, dgemv, dgemm,
> drot needed by LAPACK or directly. These routines can be compiled by a
> FORTRAN 77 compiler. But it is far better to use a processor optimized
> BLAS-library. This optimization can be done using the ATLAS package or
> playing around with the optimization options of the compiler -
> something which is done by ATLAS automatically.
> For intel processors intel offers the optimized mkl-library which is
> the optimized BLAS. One gets a noncommercial linux version for
> registration. On SUN the BLAS routines are include in SUN's performance
> library.
>
> Regards, Andreas
>
>
> Peng Yu wrote:
> > Is matrix multiplication supported by lapack or is it supported by
> > BLAS?
> >
> > I saw level 3 blas is support matrix-matix multiply. Does it mean that
> > l don't have to use lapack to do matrix multiplication.
Fortran 90 and later standards have many array intrinsic functions
that compute the same things as the BLAS, including MATMUL for matrix
multiplication. I would expect these intrinsics to be fast, and at
least for Compaq Visual Fortran they are, in my experience.
- Next message: Andreas Krebs: "Re: Matrix Multiplication in BLAS or Lapack"
- Previous message: pascalv_at_despammed.com: "Re: Q:where can I find c++ numarical anaysis codes?"
- In reply to: Andreas Krebs: "Re: Matrix Multiplication in BLAS or Lapack"
- Next in thread: Andreas Krebs: "Re: Matrix Multiplication in BLAS or Lapack"
- Reply: Andreas Krebs: "Re: Matrix Multiplication in BLAS or Lapack"
- Messages sorted by: [ date ] [ thread ]
Relevant Pages
|