use gemm to s1p0 instead of s1p1

src/operators/kernel/central-arm-func/conv_add_arm_func.h

@@ -121,7 +121,8 @@ void ConvAddCompute(const FusionConvAddParam<CPU> &param) {
   if (param.Groups() == param.Input()->dims()[1] &&
       param.Input()->dims()[1] == param.Output()->dims()[1] &&
       param.Filter()->dims()[2] == param.Filter()->dims()[3] &&
-      param.Filter()->dims()[2] == 3 && param.Strides()[0] == 1) {
+      param.Filter()->dims()[2] == 3 && param.Strides()[0] == 1 &&
+      param.paddings_[0] == 1) {
     math::DepthwiseConv3x3s1p1(param.Input(), param.Filter(), param.Output(),
                                param.Bias(), true, false);
   } else if (param.Groups() == param.Input()->dims()[1] &&

src/operators/kernel/central-arm-func/conv_add_relu_arm_func.h

@@ -125,7 +125,8 @@ void ConvAddReluCompute(const FusionConvAddReluParam<CPU> &param) {
   if (param.Groups() == param.Input()->dims()[1] &&
       param.Input()->dims()[1] == param.Output()->dims()[1] &&
       param.Filter()->dims()[2] == param.Filter()->dims()[3] &&
-      param.Filter()->dims()[2] == 3 && param.Strides()[0] == 1) {
+      param.Filter()->dims()[2] == 3 && param.Strides()[0] == 1 &&
+      param.paddings_[0] == 1) {
     math::DepthwiseConv3x3s1p1(param.Input(), param.Filter(), param.Output(),
                                param.Bias(), true, true);
   } else if (param.Groups() == param.Input()->dims()[1] &&

src/operators/kernel/central-arm-func/conv_bn_add_relu_arm_func.h

@@ -122,7 +122,8 @@ void ConvBNAddReluCompute(const FusionConvBNAddReluParam<CPU> &param) {
   if (param.Groups() == param.Input()->dims()[1] &&
       param.Input()->dims()[1] == param.Output()->dims()[1] &&
       param.Filter()->dims()[2] == param.Filter()->dims()[3] &&
-      param.Filter()->dims()[2] == 3 && param.Strides()[0] == 1) {
+      param.Filter()->dims()[2] == 3 && param.Strides()[0] == 1 &&
+      param.paddings_[0] == 1) {
     math::DepthwiseConvAddBNRelu3x3s1p1(param.Input(), param.Filter(),
                                         param.Output(), param.NewScale(),
                                         param.NewBias(), true);

src/operators/kernel/central-arm-func/conv_bn_relu_arm_func.h

@@ -120,7 +120,8 @@ void ConvBNReluCompute(const FusionConvBNReluParam<CPU> &param) {
   if (param.Groups() == param.Input()->dims()[1] &&
       param.Input()->dims()[1] == param.Output()->dims()[1] &&
       param.Filter()->dims()[2] == param.Filter()->dims()[3] &&
-      param.Filter()->dims()[2] == 3 && param.Strides()[0] == 1) {
+      param.Filter()->dims()[2] == 3 && param.Strides()[0] == 1 &&
+      param.paddings_[0] == 1) {
     math::DepthwiseConvAddBNRelu3x3s1p1(param.Input(), param.Filter(),
                                         param.Output(), param.NewScale(),
                                         param.NewBias(), true);