--- ai05s/ai05-0047-1.txt 2008/05/10 05:14:33 1.7 +++ ai05s/ai05-0047-1.txt 2008/05/29 04:36:46 1.8 @@ -1,4 +1,4 @@ -!standard G.3.1(78/2) 08-04-21 AI05-0047-1/06 +!standard G.3.1(78/2) 08-05-27 AI05-0047-1/07 !standard G.3.1(90/2) !standard G.3.2(16/2) !standard G.3.2(75/2) @@ -140,10 +140,10 @@ Note moreover that the componentwise accuracy requirements are not met by subcubic methods for matrix multiplication such as that devised by Strassen. These methods which are typically used for the fast -multiplication of very large matrices (e.g. order > 1000) have normwise -accuracy properties. If it desired to use such methods then distinct -subprograms should be provided perhaps in a child package. See Section -22.2.2 in the above reference. +multiplication of very large matrices (e.g. order more than a few thousands) +have normwise accuracy properties. If it desired to use such methods +then distinct subprograms should be provided perhaps in a child package. +See Section 22.2.2 in the above reference. Insert after G.3.1(90/2)

Questions? Ask the ACAA Technical Agent