minor doc twits

README.md

@@ -8,7 +8,7 @@ The goal of gmresls is to solve a least squares problem Ax\~=b for which the mat
 
 ## Installation
 
-You can install the development version of gmresls like so:
+You can install the current version of gmresls like so:
 
 ``` r
 intall.package("gmres") # from CRAN

man/gmresls.Rd

@@ -21,7 +21,7 @@ gmresls(f_Ax, f_Br, b, x0 = NULL, k = 0, maxit = 0, tol = 1e-07, ...)
 
 \item{tol}{A tolerance for solution, estimated as ||B(Ax-b)||/||Bb||, default 1.e-7}
 
-\item{...}{Parameters passed through to f_ax and f_br}
+\item{...}{Parameters passed through to f_Ax and f_Br}
 }
 \value{
 The solution x, having the structure of Bb.
@@ -32,11 +32,12 @@ preconditioner B: BAx=Bb (dim(B)=(n,m)).
 Implemented method uses GMRES(k) with callback functions, i.e. no explicit A or B are required. GMRES can be restarted after k iterations.
 }
 \details{
-Implemented method is equivalent to a classical GMRES(k) method with restart after constructing k basis vectors and applied to a system BAx=Bb.
+Implemented method is equivalent to a classical GMRES(k) method with restart after constructing k basis vectors and applied to a square system BAx=Bb.
 Dense matrices constructed and stored by this method are of size (length(x), k) and (k+1, k) where k is GMRES current basis vector number. If maxit > k, GMRES will be restarted after each k iterations
 Particularity of this implementation that matrices A and B have no to be stored explicitly.
 User provides just callback function mimicking their multiplication by adequate vectors.
 In case of non convergence after maxit iterations, attr(x) will contain a field 'warning' with the message which will be also issued with warning()
+If the operator BA is not of full rank, iterations will be stopped before reaching convergence or maxit. A warning will be emitted in this case.
 }
 \examples{
 # prepare a 4x3 toy problem Ax=b