[FFTW-announce] [ANNOUNCE] FFTW-3.1.3 released

[athena at fftw.org]

Dear FFTW users,

we are pleased to announce the availability of FFTW 3.1.3, which
you can download from the web page:

       http://fftw.org/download.html

This release fixes a bug that causes incorrect results to be computed
in programs that use both REDFT11 (a.k.a. DCT-IV) and RODFT11
(a.k.a. DST-IV) at the same time.  Because of the potential for
incorrect results, users of FFTW 3.1.2 are encouraged to upgrade to
3.1.3, or alternatively apply the following one-line patch to FFTW
3.1.2:

--- kernel/twiddle.c.old	2008-10-03 10:50:57.791097391 -0400
+++ kernel/twiddle.c	2008-10-03 10:51:11.875771979 -0400
@@ -50,7 +50,7 @@
 
 	  switch (p->op) {
 	      case TW_NEXT:
-		   return 1;
+		   return (p->v == q->v); /* p->i is ignored */
 
 	      case TW_FULL:
 	      case TW_HALF:

Alpha releases of FFTW 3.2 are not affected by this bug.

We apologize for the inconvenience.  Please report any problems to
fftw at fftw.org.

Cordially,
Matteo Frigo
Steven G. Johnson

Changes since FFTW 3.1.2:

* Bug fix: FFTW computes incorrect results when the user plans both
  REDFT11 and RODFT11 transforms of certain sizes.  The bug is caused
  by incorrect sharing of twiddle-factor tables between the two
  transforms, and only occurs when both are used.  Thanks to Paul
  A. Valiant for the bug report.