From 8e2ce973b5ed35c1c50c8ba24d56c4a4639fa81d Mon Sep 17 00:00:00 2001
From: wolf <wolf@0785d39b-7218-0410-832d-ea1e28bc413d>
Date: Mon, 24 Sep 2001 14:36:43 +0000
Subject: [PATCH] More info on Fortran interfaces.

git-svn-id: https://svn.dealii.org/trunk@5043 0785d39b-7218-0410-832d-ea1e28bc413d
---
 deal.II/doc/future.html | 24 ++++++++++++++++--------
 1 file changed, 16 insertions(+), 8 deletions(-)

diff --git a/deal.II/doc/future.html b/deal.II/doc/future.html
index ecaf8770ab..91a6f37f9a 100644
--- a/deal.II/doc/future.html
+++ b/deal.II/doc/future.html
@@ -48,14 +48,22 @@
          complicated applications.
 	 </p>
 
-    <li> <p>
-         <strong>Faster linear algebra:</strong>
-	 The linear algebra implementation in
-         <acronym>deal.II</acronym> is quite fast, but only compared
-         to other C/C++ implementations. We know that Fortran
-         compilers can do a much better job in optimizing the tight
-         inner loops of matrix-vector operations. We plan to try
-         interfaces from C to Fortran for that purpose.
+    <li> <p> 
+         <strong>Faster linear algebra:</strong> The linear
+         algebra implementation in <acronym>deal.II</acronym> is quite
+         fast, at least compared to other C/C++ implementations. We
+         know that Fortran compilers can sometimes do a better job in
+         optimizing some algorithms. We therefore would like to
+         interface from C to Fortran for that purpose.
+	 </p>
+	 
+	 <p>
+	 On the other hand, we have made some initial experiments with
+	 interfacing to BLAS routines on SUN, but it turned out that
+	 our own implementation of the routines we tried was faster
+	 than the ones in the vendor BLAS. Nevertheless, interfacing
+	 to LAPACK or other libraries implementing high level
+	 algorithms would be interesting.
 	 </p>
 
     <li> <p>
-- 
2.39.5