fixed comments

include/ck/tensor_operation/gpu/device/device_batched_gemm_c_permute.hpp

@@ -32,7 +32,7 @@ struct DeviceBatchedGemmCPermute : public BaseOperator
                         AElementwiseOperation a_element_op,
                         BElementwiseOperation b_element_op,
                         CElementwiseOperation c_element_op,
-                        ck::index_t BatchCount = 1) = 0;
+                        ck::index_t BatchCount) = 0;
 
     virtual std::unique_ptr<BaseInvoker> MakeInvokerPointer() = 0;
 };

include/ck/tensor_operation/gpu/device/device_batched_gemm_c_permute_xdl.hpp

@@ -115,7 +115,7 @@ __global__ void
     ignore = c_element_op;
     ignore = compute_ptr_offset_of_batch;
     ignore = block_2_ctile_map;
-#endif // end of if (defined(__gfx908__) || defined(__gfx90a__))
+#endif
 }
 
 template <typename ALayout,
@@ -519,7 +519,7 @@ struct DeviceBatchedGemmCPermuteXdl : public DeviceBatchedGemmCPermute<AElementw
         {
             const index_t G1 = e_grid_desc_g0_g1_m_n_.GetLength(I1);
             index_t b0       = g_idx / G1;
-            index_t b1       = g_idx % G1;
+            index_t b1       = b0 - b0 * G1; // g_idx % G1
             return e_grid_desc_g0_g1_m_n_.CalculateOffset(make_multi_index(b0, b1, 0, 0));
         }