add gl pipeline

example/ck_tile/16_fused_moe_general/instances/fused_moegemm_api_internal.hpp

@@ -38,7 +38,7 @@ float fused_moegemm_(const ck_tile::stream_config& s, fused_moegemm_args a)
                                              f_traits>;
 
     // using f_pipeline    = ck_tile::FusedMoeGemmPipeline_FlatmmEx<f_problem>;
-    using f_pipeline    = ck_tile::FusedMoeGemmPipeline_FlatmmUk<f_problem>;
+    using f_pipeline    = ck_tile::FusedMoeGemmPipeline_FlatmmGl<f_problem>;
     using f_partitioner = ck_tile::FusedMoeGemmTilePartitioner_Linear<f_shape>;
     using f_kernel      = ck_tile::FusedMoeGemmKernel<f_partitioner, f_pipeline, void>;
 

example/ck_tile/16_fused_moe_general/instances/fused_moegemm_api_traits.hpp

@@ -34,11 +34,11 @@ struct fmoe_ // traits, ugly name, only used for internal
     using TopkWeightDataType   = ck_tile::remove_cvref_t<typename TypeConfig::TopkWeightDataType>;
     using IndexDataType        = ck_tile::remove_cvref_t<typename TypeConfig::IndexDataType>;
 
-    static constexpr ck_tile::index_t BT_ = BlockTIle_::at(ck_tile::number<0>{}); // block token
+    static constexpr ck_tile::index_t BT_ = BlockTIle_::at(ck_tile::number<0>{}); // block token(block_m0, block_m1)
     static constexpr ck_tile::index_t BI_ =
-        BlockTIle_::at(ck_tile::number<1>{}); // block intermediate
-    static constexpr ck_tile::index_t BH_ = BlockTIle_::at(ck_tile::number<2>{}); // block hidden
-    static constexpr ck_tile::index_t BD_ = BlockTIle_::at(ck_tile::number<3>{}); // block down
+        BlockTIle_::at(ck_tile::number<1>{}); // block intermediate  (block_n0, block_k1)
+    static constexpr ck_tile::index_t BH_ = BlockTIle_::at(ck_tile::number<2>{}); // block hidden(block_k0)
+    static constexpr ck_tile::index_t BD_ = BlockTIle_::at(ck_tile::number<3>{}); // block down(block_n1)
 
     using BlockTile_0    = ck_tile::sequence<BT_, BI_, BH_>;
     using WarpPerBlock_0 = ck_tile::remove_cvref_t<WarpPerBlock_>;

include/ck_tile/ops/fused_moe.hpp

@@ -11,6 +11,7 @@
 #include "ck_tile/ops/fused_moe/pipeline/fused_moegemm_pipeline_flatmm_policy.hpp"
 #include "ck_tile/ops/fused_moe/pipeline/fused_moegemm_pipeline_flatmm_uk.hpp"
 #include "ck_tile/ops/fused_moe/pipeline/fused_moegemm_pipeline_problem.hpp"
+#include "ck_tile/ops/fused_moe/pipeline/fused_moegemm_pipeline_flatmm_gl.hpp"
 #include "ck_tile/ops/fused_moe/pipeline/fused_moegemm_traits.hpp"
 #include "ck_tile/ops/fused_moe/pipeline/moe_sorting_pipeline.hpp"
 #include "ck_tile/ops/fused_moe/pipeline/moe_sorting_policy.hpp"