diff --git a/src/nvfp4_megamoe_kernel/cutlass_nvfp4_gemm/cutlass_nvfp4_gemm.cu b/src/nvfp4_megamoe_kernel/cutlass_nvfp4_gemm/cutlass_nvfp4_gemm.cu
index 6367e297..6c0178bf 100644
--- a/src/nvfp4_megamoe_kernel/cutlass_nvfp4_gemm/cutlass_nvfp4_gemm.cu
+++ b/src/nvfp4_megamoe_kernel/cutlass_nvfp4_gemm/cutlass_nvfp4_gemm.cu
@@ -117,6 +117,14 @@ __global__ void remap_sf_to_cutlass_kernel(
 
     constexpr int flat_rank = cute::rank_v<decltype(flat)>;
 
+    // Debug: print flat_rank and first few coordinates (only for idx 0)
+    if (dst_idx == 0) {
+        printf("[remap] flat_rank=%d, MN=%d, K_sf=%d, total=%d\n", flat_rank, MN, K_sf, total);
+        printf("[remap] layout shape: ");
+        cute::print(layout_sf.shape());
+        printf("\n");
+    }
+
     int m, k_sf_out;
 
     if constexpr (flat_rank == 6) {