readme

基本的C语言线性代数函数库,在linux下可直接编译；在windows下要显示地申明包含

INSTALLATION

   Please execute the following first:
      
      prompt> ln -s Makefile.ARCH Makefile.in

   where ARCH is one of ALPHA, HPPA, LINUX, SGI64, SUN4, SUN4SOL2, or
   your own version (which should be trivial to do for other architectures).
   Make sure to set these variables appropriately in your Makefile.ARCH:

      CBDIR  is the directory where you unpacked the tar file
      BLLIB  is your Legacy BLAS library

   Then type:
      
      prompt> make help
      
   which will give you a detailed listing of targets to make.

EXECUTING THE TESTERS

   Type: 

./testing/xscblat1
./testing/xdcblat1
./testing/xccblat1
./testing/xzcblat1
./testing/xscblat2 < testing/sin2
./testing/xdcblat2 < testing/din2
./testing/xccblat2 < testing/cin2
./testing/xzcblat2 < testing/zin2
./testing/xscblat3 < testing/sin3
./testing/xdcblat3 < testing/din3
./testing/xccblat3 < testing/cin3
./testing/xzcblat3 < testing/zin3
_______________________________________________________________________________

   This package contains C interface to Legacy BLAS.
   If you want to know how to use makefile, type 'make help.'

Written by Keita Teranishi (5/20/98)
_______________________________________________________________________________

   This release updates an inconsistency between the BLAST document and
   the interface. According to the document, the enumerated types for
   the C interface to the BLAS are not typedef'ed. 

   It also updates the Level 2 and 3 testers which check for correct
   exiting of routines when called with bad arguments. This is done by
   overriding the Legacy BLAS library's implementation of xerbla().  If
   this cannot be done ( for instance one cannot override some calls
   to xerbla() in Sun's Performance library), then correct error
   exiting cannot be checked.

Updated by Jeff Horner (3/15/99)
_______________________________________________________________________________

Updated by R. Clint Whaley (2/23/03):

Fixed the i?amax error that I reported three years ago: standard dictates
IAMAX return vals in range 0 <= iamax < N, but reference was mistakenly
returning like F77: 0 < iamax <= N.