bug fixes and enharcement to gemm reductionK fusion (#682)

* add two missing files

* fix bunch of bugs of gemm-reducek fusion and add a device interface

* small changes

Co-authored-by: Haicheng Wu <haichengw@nvidia.com>

include/cutlass/gemm/kernel/default_gemm_with_k_reduction.h

@@ -91,7 +91,7 @@ template <
     typename ElementAccumulator,
     /// Operator class tag
     typename OperatorClass,
-    ///
+    /// Reduce A or B along the K dimension
     bool ReduceKForA_,
     /// Tag indicating architecture to tune for
     typename ArchTag,

include/cutlass/gemm/kernel/gemm_with_k_reduction.h

@@ -41,6 +41,7 @@
 #include "cutlass/matrix_coord.h"
 #include "cutlass/complex.h"
 #include "cutlass/semaphore.h"
+#include "cutlass/layout/pitch_linear.h"
 
 #include "cutlass/trace.h"