Difference between BLAS implementation

[Pinczakko]

Hi guys, maybe some of you are already experienced with Basic Linear Algebra Subprogram (BLAS) libraries from several vendors. Right now I have been looking (skimming) at Intel and AMD implementation but I noticed the interface functions are not exactly the same albeit the parameters might be (having only 24 hours experience on this don't justify the precision of my observation . I'm specifically looking at the C/C++ language "bindings"/"interface" to the aforementioned vendors BLAS libraries. What about other BLAS libraries? Would they be a little different too?

 

[srizbf]

i have not used the reference implementations fro intel or amd.

but i can guess that the interface can be different.
(this is only my view)

the reasoning is:
(quoting from lapack discussion...)

" LAPACK is designed as a two-tiered Fortran library, comprising higher level subroutines and "lower-level Basic Linear Algebra Subprograms (BLAS) in order to effectively exploit the caches on modern cache-based architectures...."

the BLas implementaion of LAPACK is only refence one.

the two tiered architecture of LAPACK leaves the BLAS implementation to library developers for optimisation.

so , intel and amd having their own processor architecture cache implentations , would have a differnt BLAS libs.

Since LAPACK communicate to BLAS at lower level , i think it may not be an issue.