C Matrix data load store api and functionality

[aditya4d]

API

C Matrix Load

//
// Used to load C Matrix in an optimal way by using as few registers as possible
// Output: returns a new vector of c data elements
//
std::vector<llvm::Value*> LoadCMatrix(llvm::IRBuilder<>& builder, 
llvm::LLVMContext& context, llvm::Value* ptr, llvm::Value* workitem_index,
std::vector<size_t> indices, std::vector<bool> is_relative_indices);

C Matrix Store

//
// Used to store C Matrix in an optimal way
//
void StoreCMatrix(llvm::IRBuilder<>& builder,
llvm::LLVMContext& context, llvm::Value* ptr, llvm::Value* workitem_index,
std::vector<size_t> indices, std::vector<bool> is_relative_indices,
std::vector<llvm::Value*> c_values);

• API support for Load
• API support for Store
• Add support for different gfx architectures
• Use immediate mode addressing for gfx9 using offset field for global loads