Small refinements.

include/ck/tensor_operation/gpu/device/impl/device_grouped_gemm_multiple_d_dl.hpp

@@ -744,7 +744,7 @@ struct DeviceGroupedGemmMultipleD_Dl : public DeviceGroupedGemm<ALayout,
             << K1 << ", "
             << M1PerThread << ", "
             << N1PerThread << ", "
-            << KPerThread
+            << KPerThread << ", "
             << getGemmSpecializationString(GemmSpec)
             << ">";
         // clang-format on

include/ck/tensor_operation/gpu/grid/gridwise_gemm_dl_multiple_d.hpp

@@ -315,7 +315,7 @@ struct GridwiseGemmDlMultipleD_km_kn_mn
         static_assert(a_block_desc_k0_m0_m1_k1.GetElementSpaceSize() ==
                               a_k0_m_k1_block_desc.GetElementSpaceSize() &&
                           b_block_desc_k0_n0_n1_k1.GetElementSpaceSize() ==
-                          b_k0_n_k1_block_desc.GetElementSpaceSize() &&
+                              b_k0_n_k1_block_desc.GetElementSpaceSize(),
                       "wrong!");
 
         // A matrix blockwise copy

profiler/include/profiler/profile_grouped_gemm_impl.hpp

@@ -253,7 +253,12 @@ bool profile_grouped_gemm_impl(int do_verification,
                                                               c_element_op);
 
                     ref_invoker.Run(ref_argument);
-                    pass = pass && ck::utils::check_err(c_m_n_device_results[i], c_m_n_host_result);
+                    bool group_pass =
+                        ck::utils::check_err(c_m_n_device_results[i], c_m_n_host_result);
+                    pass = pass && group_pass;
+
+                    std::cout << "group: " << i << " verification result: " << std::boolalpha
+                              << group_pass << std::endl;
 
                     if(do_log)
                     {
@@ -276,6 +281,11 @@ bool profile_grouped_gemm_impl(int do_verification,
         }
     }
 
+    if(do_verification)
+    {
+        std::cout << "Verification: " << (pass ? "SUCCESS" : "FAILURE") << std::endl;
+    }
+
     std::cout << "Best Perf: " << best_ave_time << " ms, " << best_tflops << " TFlops, "
               << best_gb_per_sec << " GB/s, " << best_gemm_name << std::endl;