initial work on producing a reference that explains each pipeline policy class

docs/reference/pipelines/ck_tile-v1.rst

@@ -0,0 +1,19 @@
+ck_tile - AGmemBGmemCReg - V1 - [SCHEDULER]
+--------------------------------------------
+
+**loop: buffer_load(i+1)**
+
+.. code-block::
+
+    elementwise_As_res = load_tile_with_elementwise(as_copy_dram_window, a_element_func);
+    elementwise_Bs_res = load_tile_with_elementwise(bs_copy_dram_window, b_element_func);    
+
+The ``load_tile_with_elementwise`` function calls the ``load`` method on ``tile_window_with_static_distribution`` instances.  
+
+- The ``tile_window_with_static_distribution::load`` method calls ``make_static_distributed_tensor`` to create distributed tensors for storing loaded data.
+- The ``static_distributed_tensor`` class uses ``make_naive_tensor_descriptor_packed`` from ``ck_tile/core/tensor/tensor_descriptor.hpp`` in its tensor slicing operations (via ``set_y_sliced_thread_data``).
+
+
+**full class definition**
+
+.. literalinclude:: ./gemm_pipeline_agmem_bgmem_creg_v1.hpp

docs/reference/pipelines/xdl-v1.rst

@@ -0,0 +1,35 @@
+
+Xdl - v1 - Intrawave
+----------------------
+
+**loop: buffer_load(i+1)**
+
+.. code-block::
+
+    static_for<0, KRepeat, 1>{}([&](auto k) {
+        static_for<0, MRepeat, 1>{}([&](auto m0) {
+            a_thread_copy_.Run(a_block_desc_m0_m1_m2_k,
+                                make_tuple(m0, I0, I0, Number<k * AMmaKStride>{}),
+                                a_block_buf,
+                                a_thread_desc_,
+                                make_tuple(m0, I0, k, I0),
+                                a_thread_buf);
+            static_for<0, NRepeat, 1>{}([&](auto n0) {
+                b_thread_copy_.Run(b_block_desc_n0_n1_n2_k,
+                                    make_tuple(n0, I0, I0, Number<k * BMmaKStride>{}),
+                                    b_block_buf,
+                                    b_thread_desc_,
+                                    make_tuple(n0, I0, k, I0),
+                                    b_thread_buf);
+            });
+        });
+    });
+
+The ``a_thread_copy_.Run`` function is a member of the ``ThreadwiseTensorSliceTransfer_v4`` class.  The ``ThreadwiseTensorSliceTransfer_v4`` class uses,
+
+- ``make_tensor_coordinate`` from ``tensor_descriptor.hpp`` 
+- ``make_naive_tensor_descriptor_packed`` from ``tensor_descriptor_helper.hpp``
+
+**full class definition**
+
+.. literalinclude:: ./blockwise_gemm_pipeline_xdlops_v1.hpp