diff --git a/src/layer/arm/prelu_arm.cpp b/src/layer/arm/prelu_arm.cpp
index 72d9ae0fba6..04d06028380 100644
--- a/src/layer/arm/prelu_arm.cpp
+++ b/src/layer/arm/prelu_arm.cpp
@@ -72,7 +72,7 @@ int PReLU_arm::forward(const Mat& bottom_blob, Mat& top_blob) const
             "vdup.f32   q2, %6              \n"
             "0:                             \n"
             "pld        [%1, #128]          \n"
-            "vld1.f32   {d0-d1}, [%1 :128]  \n"
+            "vld1.f32   {d0-d1}, [%1 :128]! \n"
             "vcle.f32   q3, q0, q1          \n"
             "vmul.f32   q4, q0, q2          \n"
             "vbit.32    q0, q4, q3          \n"
diff --git a/src/layer/arm/relu_arm.cpp b/src/layer/arm/relu_arm.cpp
index 5477c37afe1..9000d365f11 100644
--- a/src/layer/arm/relu_arm.cpp
+++ b/src/layer/arm/relu_arm.cpp
@@ -130,7 +130,7 @@ int ReLU_arm::forward(const Mat& bottom_blob, Mat& top_blob) const
                 "vdup.f32   q2, %6              \n"
                 "0:                             \n"
                 "pld        [%1, #128]          \n"
-                "vld1.f32   {d0-d1}, [%1 :128]  \n"
+                "vld1.f32   {d0-d1}, [%1 :128]! \n"
                 "vcle.f32   q3, q0, q1          \n"
                 "vmul.f32   q4, q0, q2          \n"
                 "vbit.32    q0, q4, q3          \n"