[plt-scheme] Re: A segmentation fault on OS X Intel

From: Will Farr (farr at mit.edu)
Date: Sat Jan 12 11:01:33 EST 2008


I've added your code to the test suite, and it runs just fine
(repeatedly) on my machine (Mac OS X 10.4, PPC).  If
matrix-f64vector-mul works fine on your machine, then I have a guess
as to the problem: check the values of the enum for _cblas-transpose
(defined in blas-lapack.ss) against their corresponding values in the
BLAS headers.  Unlike matrix-f64vector-mul, f64vector-matrix-mul uses
the transpose flag to BLAS when doing the multiplication.  Otherwise
the routines are the same.

Let me know if this doesn't work---and I'll have another look at the code.


On Jan 12, 2008 6:59 AM, Noel Welsh <noelwelsh at gmail.com> wrote:
> I've tracked this down to some (but not all) uses of
> f64vector-matrix-mul.  The following reliably causes a segmentation
> fault in 372 and 3.99.09
> > (require (planet "all.ss" ("wmfarr" "plt-linalg.plt" 1 5)))
> > (f64vector-matrix-mul (f64vector 1. .0 .0 .0 .0 .0 .0 .0 .0 .0) (make-matrix 10 10 1.))
> Segmentation fault
> So reading about foreign.ss and the BLAS interface is next, I suppose.
> N.

Posted on the users mailing list.