v4.3 update. (#2709)

* v4.3 update.

* Update the cute_dsl_api changelog's doc link

* Update version to 4.3.0

* Update the example link

* Update doc to encourage user to install DSL from requirements.txt

---------

Co-authored-by: Larry Wu <larwu@nvidia.com>

tools/library/include/cutlass/library/descriptions.h

@@ -348,6 +348,9 @@ struct BlockScaledGemmDescription : public OperationDescription {
   /// Describes the destination matrix
   TensorDescription D;
 
+  /// Describes the sparse meta matrices
+  TensorDescription E;
+
   /// Describes the SFA operand
   TensorDescription SFA;
 

tools/library/include/cutlass/library/library.h

@@ -392,10 +392,11 @@ struct BlockScaledGemmArguments {
 
   library::RuntimeDatatype runtime_input_datatype_a{library::RuntimeDatatype::kStatic}; 
   library::RuntimeDatatype runtime_input_datatype_b{library::RuntimeDatatype::kStatic}; 
-
+  int device_index{0};
   bool use_pdl{false};
 };
 
+
 /// Blockwise GEMM
 //
 // OperationKind: kBlockwiseGemm

tools/library/include/cutlass/library/types.h

@@ -209,6 +209,7 @@ enum class GemmKind {
   kPlanarComplex,
   kPlanarComplexArray,
   kGrouped,
+  kBlockScaledSparseGemm,
   kInvalid
 };