readme

Author: AustinPrivett

README.md

@@ -7,7 +7,7 @@
 The cost of multiplying a chain of matrices can vary significantly depending on the order in which the multiplication steps are applied. The [Matrix chain multiplication](https://www.wikiwand.com/en/Matrix_chain_multiplication) algorithm applied here (described in *Introduction to Algorithms, 3rd Edition* by Cormen et al.) finds the optimal multiplication sequence.
 
 With Julia, it is easy to allow for specialization on the matrix types.
-For example, the types given in the ArrayFire.jl package should work here, too.
+For example, the types given in the ArrayFire.jl package or matrices of matrices (block matrices) should work, too.
 
 ## Installing this Package
 
@@ -56,10 +56,12 @@ use of the optimal operation order generated here.
 If all matrices are the same size, the overhead of the optimization is
 not worth the (non-existing) benefit.
 
-## Further improvements to be made
+## Future improvements
 
 * Recognize situations that don't need the analysis to eliminate the
-overhead (allowing for a more ubiquitous use of `matrixchainmultiply(...)`
-instead of `*(...)`).
+  overhead (allowing for a more ubiquitous use of `matrixchainmultiply(...)`
+  instead of `*(...)`).
 * For repeated uses, given some matrix size sequence, produce the
-function a single time so that the analysis isn't done unnecessarily.
+  function a single time so that the analysis isn't done unnecessarily.
+* Faster or more general algorithms (i.e., those that apply to the general
+  tensor problem) are certainly desirable if you would like to add them.