add description in example code

example/32_batched_gemm_scale_softmax_gemm/batched_multihead_attention_backward_fp16.cpp

@@ -5,13 +5,21 @@ Backprop for Gemm + Softmax + Gemm fused operation, where forward prop is define
 
   Y_g_m_o = Softmax(alpha * Q_g_m_k * K_g_k_n) * V_g_n_o
 
-Input:
+Computation graph:
 
-  Q, K, V, Y, dY, and per-row softmax stats computed beforehand during forward prop
+          K^T                   V
+          |                     |
+          |                     |
+    Q --- * ----- Softmax ----- * --> Y
+              S             P
 
-Outputs:
+Kernel inputs:
 
-  dQ, dK, dV
+    Q, K, V, Y, dY, per-row softmax stats (LSE)
+
+Kernel outputs:
+
+    dQ, dK, dV
 
 */
 
@@ -37,6 +45,7 @@ Outputs:
 #include "ck/library/reference_tensor_operation/cpu/reference_batched_gemm.hpp"
 #include "ck/library/reference_tensor_operation/cpu/reference_softmax.hpp"
 
+
 template <ck::index_t... Is>
 using S = ck::Sequence<Is...>;