[CMake] problem with FindBLAS.cmake looking for cblas_dgemm

Michael Wild themiwi at gmail.com
Wed Jun 2 09:56:34 EDT 2010


On 2. Jun, 2010, at 15:09 , Crni Gorac wrote:

> On Wed, Jun 2, 2010 at 2:08 PM, Michael Wild <themiwi at gmail.com> wrote:
>> 
>> Just to correct a thing: cblas_dgemm is NOT an implementation detail, it is part of CBLAS which is the official C-interface to BLAS.
> 
> I think "implementation detail" mentioned in the OP message was not
> about CBLAS per se, but about linking C code from the Fortran compiler
> instead, which is indeed an implementation detail, and which is
> surfacing because of FindBLAS.cmake is currently working exactly by
> asking Fortran compiler to find a CBLAS function - so much more
> meaningful test would be to ask the compiler to find BLAS function,
> again exactly as suggested in the OP message.

Looks like I misunderstood the OP, sorry for that...

[...]

>  Obviously, no maintainer of these modules could afford to
> track all of the possible implementation, but basically users of
> particular implementation are usually reporting errors and suggesting
> improvements, and this worked very well so far.  It seems only in this
> particular case that the issue somehow got below of the radar of
> Kitware people, hope they would fix it now...

Well, for one it still doesn't find Intel MKL-11, and that one's appeared back in late 2008...

Another thing which stumped me so far is the search order. If somebody shells out the money to get his hands on Intel MKL he probably wants it to be preferred over most of the other implementations, right? So, why then, is it searched for last?

Michael


More information about the CMake mailing list