On Wed, 18 Jan 2012, Sven Schreiber wrote:
just wondering: is gretl (automatically) using the
SSE[2/3/whatever]-optimized versions of Lapack and/or Blas and so on? (I
know it can be configured to use SSE2 for RNG.)
For MS Windows gretl supplies the standard lapack and netlib
"reference blas", cross-compiled using gcc on Linux. I think it's
lapack/blas version 3.1.1 but I'll have to check. Anyway, no sse,
since that's not supported by the vanilla netlib code.
For OS X gretl relies on the native vecLib framework, which
presumably is optimized for the given processor.
On Linux gretl uses whatever the system offers for lapack + blas,
which may or may not be aggressively optimized. See
http://ricardo.ecn.wfu.edu/~cottrell/tmp/gretl_speed.html
Allin