From 9a743139afc23e03712f7febefbbcda189e19111 Mon Sep 17 00:00:00 2001
From: Damien Lejeune <damien.lejeune@amd.com>
Date: Wed, 21 Jan 2026 07:42:59 -0500
Subject: [PATCH] Fix kBlockSize

---
 .../ops/sinkhorn_knopp/pipeline/sinkhorn_knopp_problem.hpp   | 4 +---
 include/ck_tile/ops/sinkhorn_knopp/sinkhorn_knopp_kernel.hpp | 5 +++--
 2 files changed, 4 insertions(+), 5 deletions(-)

diff --git a/include/ck_tile/ops/sinkhorn_knopp/pipeline/sinkhorn_knopp_problem.hpp b/include/ck_tile/ops/sinkhorn_knopp/pipeline/sinkhorn_knopp_problem.hpp
index 80c6fa40b1..6d888c9fd2 100644
--- a/include/ck_tile/ops/sinkhorn_knopp/pipeline/sinkhorn_knopp_problem.hpp
+++ b/include/ck_tile/ops/sinkhorn_knopp/pipeline/sinkhorn_knopp_problem.hpp
@@ -24,7 +24,6 @@ namespace ck_tile {
 //     static constexpr index_t Repeat_N        = Repeat_N;
 // };
 
-
 template <typename BlockWarps, // num warps along seq<M, N>
           typename BlockTile,  // block size, seq<M, N>
           typename WarpTile,   // warp size, seq<M, N>
@@ -56,8 +55,7 @@ struct SinkhornKnoppShape
     static constexpr index_t Repeat_M = Block_M * RepeatInWarp_M / (WarpPerBlock_M * Warp_M);
     static constexpr index_t Repeat_N = Block_N * RepeatInWarp_N / (WarpPerBlock_N * Warp_N);
 
-    // static constexpr index_t BlockSize = ck_tile::get_warp_size();
-    static constexpr index_t BlockSize = 1; // TODO
+    static constexpr index_t BlockSize = ck_tile::get_warp_size();
 };
 
 template <typename _XDataType,
diff --git a/include/ck_tile/ops/sinkhorn_knopp/sinkhorn_knopp_kernel.hpp b/include/ck_tile/ops/sinkhorn_knopp/sinkhorn_knopp_kernel.hpp
index c1b27e0c1f..ae65e8c0a9 100644
--- a/include/ck_tile/ops/sinkhorn_knopp/sinkhorn_knopp_kernel.hpp
+++ b/include/ck_tile/ops/sinkhorn_knopp/sinkhorn_knopp_kernel.hpp
@@ -75,7 +75,7 @@ struct SinkhornKnoppKernelDummyNonStochastic
         return tensor;
     }
 
-    CK_TILE_DEVICE void operator()([[maybe_unused]]const SinkhornKnoppArgs& args) const
+    CK_TILE_DEVICE void operator()([[maybe_unused]] const SinkhornKnoppArgs& args) const
     {
         // using S = Problem::BlockShape;
 
@@ -83,7 +83,8 @@ struct SinkhornKnoppKernelDummyNonStochastic
 
         // const auto x_desc = make_naive_tensor_descriptor(make_tuple(args.input_m, args.input_m),
         //                                                  make_tuple(args.input_m, 1),
-        //                                                  number<4>{}, // TODO: Hardcoded vectorization, we should calculate it!
+        //                                                  number<4>{}, // TODO: Hardcoded
+        //                                                  vectorization, we should calculate it!
         //                                                  number<1>{});
 
         // auto buffer_view = make_buffer_view<address_space_enum::global>(