Update documentation of MEX interface

Author: mfasi

mex/gemmi.m

@@ -1,14 +1,25 @@
 %  GEMMI    Compute matrix product using integer Ozaki scheme.
-%   C = GEMMI(A,B,ASPLITS,BSPLITS) computes the product A*B using
-%   the Ozaki scheme with ASPLITS slices for the matrix A and BSPLITS
-%   slices for the matrix B.
+%   C = GEMMI(A,B,ASPLITS,BSPLITS,ALGORITHM) computes the product A*B
+%   using the Ozaki scheme with ASPLITS and BSPLITS slices for the
+%   matrices A and B, respectively. The ALGORITHM parameter must be
+%   a struct, with the following fields currently supported.
+%   'split' - selects the stragegy to be used to split A and B into
+%             slices. Possible values are 'b' for bitmasking and 'n
+%             for round-to-nearest (default).
+%   'acc'   - selects how the exact integer matrix products are
+%             accumulated. Possible values are 'f' for floating-point
+%             arithmetic and 'i' for integer accumulation (default).
+%    
+%   C = GEMMI(A,B,ASPLITS,BSPLITS) uses the last ALGORITHM parameter
+%   in the most recent call to GEMMI, or the default values if no
+%   previous call was made.
 %
 %   C = GEMMI(A,B,SPLITS) uses SPLITS slices for both A and B.
 % 
 %   The splits are stored as 8-bit signed integer, the dot products are
 %   performed using 32-bit signed arithmetic, and the final accumulation
-%   uses either binary32 or binary64 arithmetic, depending on the type
-%   of A and B.
+%   uses either the same format as the matrices A and B (if 'acc' is 'f')
+%   or 32-bit arithmetic (if 'acc' is 'i').
 %
 %   The matrices A and B must be conformable, and multiplication by a
 %   scalar is not supported.