v4.1 release update v2. (#2481)

include/cutlass/gemm/warp/mma_complex_tensor_op.h

@@ -782,7 +782,7 @@ public:
       for (int n = 0; n < MmaIterations::kColumn; ++n) {
 
         // negate OperandB to accumulate  -(a.imag()*b.imag())
-        // negating OperandB emits less instrucitons than negating OperandA as OperandB has less elements
+        // negating OperandB emits less instructions than negating OperandA as OperandB has less elements
         negate<InstMmaOperandB> negate_op;
 
         // Real-valued accumulator part

include/cutlass/gemm/warp/mma_complex_tensor_op_fast_f32.h

@@ -598,7 +598,7 @@ public:
       for (int n = 0; n < MmaIterations::kColumn; ++n) {
 
         // negate OperandB to accumulate  -(a.imag()*b.imag())
-        // negating OperandB emits less instrucitons than negating OperandA as OperandB has less elements
+        // negating OperandB emits less instructions than negating OperandA as OperandB has less elements
         negate<InstMmaOperandB> negate_op;
 
         // Real-valued accumulator part

include/cutlass/gemm/warp/mma_mixed_input_tensor_op.h

@@ -427,7 +427,7 @@ public:
   using TransformedFragmentA =
       Array<ElementAMma, FragmentA::kElements>;
 
-  /// Underlying arch::Mma instruction operand fragement for matrix A
+  /// Underlying arch::Mma instruction operand fragment for matrix A
   using MmaOperandA = typename ArchMmaOperator::FragmentA;
 
   /// Iterates over the B operand in Shared Memory
@@ -443,7 +443,7 @@ public:
   using TransformedFragmentB =
       Array<ElementBMma, FragmentB::kElements>;
 
-  /// Underlying arch::Mma instruction operand fragement for matrix B
+  /// Underlying arch::Mma instruction operand fragment for matrix B
   using MmaOperandB = typename ArchMmaOperator::FragmentB;
 
   /// Iterates over the C operand in memory
@@ -454,7 +454,7 @@ public:
   /// Storage for C tile
   using FragmentC = typename IteratorC::Fragment;
 
-  /// Underlying arch::Mma instruction operand fragement for matrix C
+  /// Underlying arch::Mma instruction operand fragment for matrix C
   using MmaOperandC = typename ArchMmaOperator::FragmentC;
 
   /// Number of mma operations performed

include/cutlass/gemm/warp/mma_sparse_tensor_op.h

@@ -117,7 +117,7 @@ public:
   /// Shape of the warp in units of thread (concept: MmaLanePolicySimt)
   using Policy = Policy_;
 
-  /// Equivalant base dense mma
+  /// Equivalent base dense mma
   using Base = MmaTensorOp<Shape, ElementA, LayoutA, ElementB, LayoutB,
                            ElementC, LayoutC, Policy, PartitionsK_,
                            AccumulatorsInRowMajor, Enable>;

include/cutlass/gemm/warp/mma_tensor_op_fragment_iterator.h

@@ -33,7 +33,7 @@
     \brief This defines a "fragment" iterator for visiting the fragments of a warp tile
       that participate in one warp-level mma operation.
 
-      Typically, this is used to access the accumulator tile/fragement of a warp-level mma operation.
+      Typically, this is used to access the accumulator tile/fragment of a warp-level mma operation.
       The accumulator tile is then partitioned into smaller tiles/fragments that can be fed into 
       next warp-level mma operation. 
 

include/cutlass/gemm/warp/mma_tensor_op_tile_access_iterator.h

@@ -62,7 +62,7 @@ namespace warp {
 
 
 /// Tile access iterator
-/// Each iteration acess in the tile is
+/// Each iteration access in the tile is
 /// used as multiplicand for one
 /// warp-level matrix multiplication
 template <