From 059711633041b73ac3ed3c3b287eee8667092f3f Mon Sep 17 00:00:00 2001
From: Chao Liu <lc.roy86@gmail.com>
Date: Wed, 9 Jan 2019 19:11:45 -0600
Subject: [PATCH] refactor

---
 src/include/blockwise_tensor_op.cuh | 1 +
 src/include/common.cuh              | 2 ++
 2 files changed, 3 insertions(+)

diff --git a/src/include/blockwise_tensor_op.cuh b/src/include/blockwise_tensor_op.cuh
index 6cfff6023d..0649f6ba77 100644
--- a/src/include/blockwise_tensor_op.cuh
+++ b/src/include/blockwise_tensor_op.cuh
@@ -76,6 +76,7 @@ blockwise_4d_tensor_pointwise_operation_unary(DstDesc, Float* __restrict__ p_dst
     }
 }
 
+// Function: p_dst[reorder[i0], reorder[i1], reorder[i2], reorder[i3]] = p_src[i0,i1,i2,i3]
 // TODO: in order to optimize mem access for different mem type,
 // need to write specialized version
 template <unsigned BlockSize,
diff --git a/src/include/common.cuh b/src/include/common.cuh
index d5c975bf05..dc8e2c5fed 100644
--- a/src/include/common.cuh
+++ b/src/include/common.cuh
@@ -11,3 +11,5 @@ struct is_same<T, T>
 {
     static const bool value = true;
 };
+
+__device__ unsigned get_thread_local_id() { return threadIdx.x; }