refine blockgemm pipeline version as base struct.

example/65_gemm_multiply_multiply/gemm_multiply_multiply_xdl_fp8_bpreshuffle.cpp

@@ -9,6 +9,7 @@
 #include "ck/ck.hpp"
 #include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
 #include "ck/tensor_operation/gpu/device/impl/device_gemm_multiple_d_xdl_cshuffle_v3_b_preshuffle.hpp"
+#include "ck/tensor_operation/gpu/device/impl/device_gemm_multiple_d_xdl_cshuffle_v3.hpp"
 #include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
 #include "ck/tensor_operation/gpu/element/unary_element_wise_operation.hpp"
 

include/ck/tensor_operation/gpu/block/blockwise_gemm_pipeline_xdlops_ab_scale_selector.hpp

@@ -9,13 +9,6 @@
 
 namespace ck {
 
-enum struct BlockGemmPipelineVersion
-{
-    v1, // Naive
-    v2, // Mem
-    v3, // Comp
-};
-
 template <BlockGemmPipelineVersion BlkGemmPipelineVer,
           BlockGemmPipelineScheduler BlkGemmPipeSche,
           index_t BlockSize,

include/ck/tensor_operation/gpu/block/blockwise_gemm_pipeline_xdlops_b_preshuffle_selector.hpp

@@ -7,12 +7,6 @@
 #include "ck/tensor_operation/gpu/block/blockwise_gemm_pipeline_xdlops_b_preshuffle_v2.hpp"
 namespace ck {
 
-enum struct BlockGemmPipelineVersion
-{
-    v1, // Single lds buffer
-    v2, // Double lds buffer
-};
-
 template <BlockGemmPipelineVersion BlkGemmPipelineVer,
           BlockGemmPipelineScheduler BlkGemmPipeSche,
           index_t BlockSize,

include/ck/tensor_operation/gpu/block/blockwise_gemm_pipeline_xdlops_selector.hpp

@@ -11,15 +11,6 @@
 
 namespace ck {
 
-enum struct BlockGemmPipelineVersion
-{
-    v1, // Naive
-    v2, // Mem
-    v3, // Comp
-    v4, // Comp, double lds buffer
-    v5, // Comp, double global prefetch register buffer
-};
-
 template <BlockGemmPipelineVersion BlkGemmPipelineVer,
           BlockGemmPipelineScheduler BlkGemmPipeSche,
           index_t BlockSize,

include/ck/utility/blkgemmpipe_scheduler.hpp

@@ -8,6 +8,19 @@
 
 namespace ck {
 
+enum struct BlockGemmPipelineVersion
+{
+    // For GEMM
+    v1, // Naive
+    v2, // Mem
+    v3, // Comp
+    v4, // Comp, double lds buffer
+    v5, // Comp, double global prefetch register buffer
+
+    // For GEMM with preshuffled weight
+    // v1, single lds buffer
+    // v2, double lds buffer
+};
 enum struct BlockGemmPipelineScheduler
 {
     Intrawave,