Update linear_combination_generic.h (NVIDIA#472)

add `skip_elementwise_` to support serial splitk in linear_combination_generic.h`

include/cutlass/epilogue/thread/linear_combination_generic.h

@@ -126,6 +126,7 @@ class LinearCombinationGeneric {
 
   ElementCompute alpha_;
   ElementCompute beta_;
+  bool skip_elementwise_;
 
 public:
 
@@ -135,6 +136,7 @@ class LinearCombinationGeneric {
 
     alpha_ = (params.alpha_ptr ? *params.alpha_ptr : params.alpha);
     beta_ = (params.beta_ptr ? *params.beta_ptr : params.beta);
+    skip_elementwise_ = false;
   }
 
   /// Returns true if source is needed
@@ -155,6 +157,10 @@ class LinearCombinationGeneric {
     if (k_partition) {
       beta_ = ElementCompute(1);
     }
+
+    if (k_partition != k_partition_count - 1) {
+      skip_elementwise_ = true;
+    }
   }
 
   /// Computes linear scaling: D = alpha * accumulator + beta * source
@@ -188,7 +194,7 @@ class LinearCombinationGeneric {
       intermediate = mul_add_accumulator(alpha_, converted_accumulator, intermediate);    // D = alpha * Accum + X
     }
 
-    intermediate = activation(intermediate);
+    intermediate = skip_elementwise_ ? intermediate : activation(intermediate);
 
     // Convert to destination numeric type
     NumericArrayConverter<ElementOutput, ElementCompute, kCount, Round> destination_converter;
@@ -219,7 +225,7 @@ class LinearCombinationGeneric {
       intermediate = mul_add_accumulator(alpha_, converted_accumulator);    // D = alpha * Accum
     }
 
-    intermediate = activation(intermediate);
+    intermediate = skip_elementwise_ ? intermediate : activation(intermediate);
 
     // Convert to destination numeric type
     NumericArrayConverter<ElementOutput, ElementCompute, kCount, Round> destination_converter;