Intel has compiled a set of MKL (Math Kernel Libraries) for Linux. Does
anyone know of any plans to have these compiled under QNX? I am currently
installing CBLAS (Basic Linear Algebra Subroutines) and LAPACK (Linear
Algebra Package) for evaluation, but I expect this to be significantly
slower than a machine- and OS-optimized solution.
David MacKinnon
National Research Council of Canada
If MKL does not call linushish stuff from within I think you should be able
to link QNX6 binary against Linux MKL lib. I don’t quite understand what is
‘math kernel’ though…
- igor
“David MacKinnon” <David.MacKinnon@nrc.ca> wrote in message
news:9jmlla$k63$1@moonstone.imsb.nrc.ca…
Intel has compiled a set of MKL (Math Kernel Libraries) for Linux. Does
anyone know of any plans to have these compiled under QNX? I am currently
installing CBLAS (Basic Linear Algebra Subroutines) and LAPACK (Linear
Algebra Package) for evaluation, but I expect this to be significantly
slower than a machine- and OS-optimized solution.
David MacKinnon
National Research Council of Canada
Yes, this should work. We have done the same with similar Linux libraries
from Sandia National Laboratories which are unofficial UNIX ports of the Intel MKL.
http://www.cs.utk.edu/~ghenry/distrib/
The “math kernel” is processor optimized libraries for linear algebra, FFT and similar
which have a standardized API (e.g. BLAS and LAPACK).
I think they are using MMX/SSE and similar since they e.g. can get 3GFLOPS
throughput for FFT on a 1500MHz P4 processor.
http://developer.intel.com/software/products/mkl/mkl51w/mklspecs.htm
Tom
Igor Kovalenko wrote:
If MKL does not call linushish stuff from within I think you should be able
to link QNX6 binary against Linux MKL lib. I don’t quite understand what is
‘math kernel’ though…
- igor
“David MacKinnon” <> David.MacKinnon@nrc.ca> > wrote in message
news:9jmlla$k63$> 1@moonstone.imsb.nrc.ca> …
Intel has compiled a set of MKL (Math Kernel Libraries) for Linux. Does
anyone know of any plans to have these compiled under QNX? I am currently
installing CBLAS (Basic Linear Algebra Subroutines) and LAPACK (Linear
Algebra Package) for evaluation, but I expect this to be significantly
slower than a machine- and OS-optimized solution.
David MacKinnon
National Research Council of Canada