1.6.4 docs update

Former-commit-id: bfa0d260b1bf050c9cacb85c1d75fc997718b419

CHANGE_LOG.TXT

@@ -1,3 +1,16 @@
+1.6.4    12/06/2016
+    - Updated sm_5x, sm_6x tuning policies for radix sorting (3.5B and 3.4B 
+      32b keys/s on TitanX and GTX 1080, respectively)
+    - Remove out-dated VC project files
+    - Bug fixes: 
+        - Issue 65: DeviceSegmentedRadixSort should allow inputs to have 
+          pointer-to-const type 
+        - Mollify Clang device-side warnings
+        - Restore fence work-around for scan (reduce-by-key, etc.) hangs 
+          in CUDA 8.5
+
+//-----------------------------------------------------------------------------
+
 1.6.3    11/20/2016
     - API change: BlockLoad and BlockStore are now templated by the local
       data type, instead of the Iterator type.  This allows for output iterators

README.md

@@ -1,7 +1,7 @@
 <hr>
 <h3>About CUB</h3>
 
-Current release: v1.6.0 (11/20/2016)
+Current release: v1.6.4 (12/06/2016)
 
 We recommend the [CUB Project Website](http://nvlabs.github.com/cub) and the [cub-users discussion forum](http://groups.google.com/group/cub-users) for further information and examples.