rewrite roadmap

Author: benvanwerkhoven

roadmap.md

@@ -4,28 +4,32 @@ This roadmap presents an overview of the features we are currently planning to
 implement. Please note that this is a living document that will evolve as
 priorities grow and shift.
 
-### version 0.4.0
+## version 0.3.1
 
-This is the list of features that we want to have implemented by the next version.
+ * Support for C++ templated kernels through NVRTC and template parameter tuning
+ * Document a vocabulary of reserved/special tunable parameter names
 
- * Extend Fortran support, no more warnings on data types or missing block size parameter etc.
- * Turn the C backend into a more general compiler backend
- * A get_parameterized_kernel_source function to return the parameterized kernel source for inspection
- * Function to instrument source files with parameter values after tuning
- * Function to generate wrapper kernels for directly calling device functions
- 
-### version 1.0.0
+## version 0.3.2
+
+ * Support for specifying metrics, using tunable parameters, results, and/or a function
+ * Allow strategies to tune for a metric other than time
+
+## version 0.4.0
+
+ * Multi-objective optimization
+
+## version 1.0.0
 
 These functions are to be implemented by version 1.0.0, but may already be
 implemented in earlier versions.
 
- * Functionality for including auto-tuned kernels in applications
  * Tuning kernels in parallel on a set of nodes in a GPU cluster
+ * Functionality for including auto-tuned kernels in applications
 
-### Low priority
+## Wish list
 
 These are the things that we would like to implement, but we currently have no
-demand for it. If you are interested in any of these, let us know!
+immediate demand for it. If you are interested in any of these, let us know!
 
  * Option to set dynamically allocated shared memory for CUDA backend
  * Option to set function that computes search space restriction, instead of a list of strings
@@ -35,5 +39,8 @@ demand for it. If you are interested in any of these, let us know!
  * Example that tunes a kernel using thread block re-indexing
  * Example CUDA host code that uses runtime compilation
  * A test_kernel function to perform parameterized testing without tuning
-
-
+ * Extend Fortran support, no more warnings on data types or missing block size parameter etc.
+ * Turn the C backend into a more general compiler backend
+ * A get_parameterized_kernel_source function to return the parameterized kernel source for inspection
+ * Function to generate wrapper kernels for directly calling/testing device functions
+ * Function to instrument source files with parameter values after tuning