Revert "use bgemm_naive instead of bconv_naive as fallback due to the…

… 128-align weight"

This reverts commit be606fa.

dabnn/layers/BinConv.cpp

@@ -25,6 +25,7 @@ BinConv::BinConv(NetCP net, const std::string &name, css input, css weight,
       stride_h(stride_h),
       stride_w(stride_w) {
     auto &mat_map = net.lock()->mat_map_;
+<<<<<<< HEAD
     if (direct_conv_compatible()) {
         const auto binaized_name = "binaized_for_" + output + "_cal";
         if (mat_map.find(binaized_name) == mat_map.end()) {
@@ -34,7 +35,16 @@ BinConv::BinConv(NetCP net, const std::string &name, css input, css weight,
                 binaized_name);
         }
         binarized_mat = mat(binaized_name);
+=======
+    const auto binaized_name = "binaized_for_" + output + "_cal";
+    if (mat_map.find(binaized_name) == mat_map.end()) {
+        auto &input_mat = *mat_map[input];
+        mat_map[binaized_name] =
+            std::make_shared<Mat>(input_mat.h, input_mat.w, input_mat.elem_c,
+                                  DataType::Bit, binaized_name);
+>>>>>>> parent of be606fa... use bgemm_naive instead of bconv_naive as fallback due to the 128-align weight
     }
+    binarized_mat = mat(binaized_name);
 
     const auto pad_name = "pad_for_" + output + "_cal";
     if (mat_map.find(pad_name) == mat_map.end()) {
@@ -45,6 +55,7 @@ BinConv::BinConv(NetCP net, const std::string &name, css input, css weight,
     }
     padded_mat = mat(pad_name);
 
+<<<<<<< HEAD
     if (net.lock()->optimize && !direct_conv_compatible()) {
         const auto col_mat_name = "col_for_" + output + "_cal";
         if (mat_map.find(col_mat_name) == mat_map.end()) {
@@ -56,7 +67,20 @@ BinConv::BinConv(NetCP net, const std::string &name, css input, css weight,
                 std::make_shared<Mat>(1, 1, len, bnn::DataType::Bit);
         }
         col_mat = mat(col_mat_name);
+=======
+    const auto col_mat_name = "col_for_" + output + "_cal";
+    if (mat_map.find(col_mat_name) == mat_map.end()) {
+        const auto len =
+            output_mat->h * output_mat->w *
+            align_to(weight_mat->h * weight_mat->w * input_mat->elem_c, 128);
+        mat_map[col_mat_name] =
+            std::make_shared<Mat>(1, 1, len, bnn::DataType::Bit);
+    }
+    col_mat = mat(col_mat_name);
+>>>>>>> parent of be606fa... use bgemm_naive instead of bconv_naive as fallback due to the 128-align weight
 
+    if (net.lock()->optimize && !direct_conv_compatible() &&
+        gemm_compatible()) {
         const auto trans_weight_mat_name = "trans_" + weight;
         // transpose the weight for bgemm
         const int m = weight_mat->n;
@@ -123,6 +147,7 @@ bool BinConv::gemm_compatible() const {
 }
 
 void BinConv::forward_impl() const {
+<<<<<<< HEAD
     if (net_.lock()->optimize && direct_conv_compatible()) {
         pack_mat(*input_mat, *binarized_mat);
         pad(*binarized_mat, pad_h, pad_w, *padded_mat);
@@ -138,14 +163,38 @@ void BinConv::forward_impl() const {
         const int n = output_mat->h * output_mat->w;
         const int k = weight_mat->total() / weight_mat->n;
         if (net_.lock()->optimize && gemm_compatible()) {
+=======
+    if (net_.lock()->optimize) {
+        if (direct_conv_compatible()) {
+            pack_mat(*input_mat, *binarized_mat);
+            pad(*binarized_mat, pad_h, pad_w, *padded_mat);
+            bconv_3x3(*padded_mat, *weight_mat, *output_mat, stride_h);
+        } else if (gemm_compatible()) {
+            output_mat->fill<float>(0.f);
+
+            bnn::fused_binarize_im2col(*input_mat, weight_mat->h, weight_mat->w,
+                                       pad_h, pad_w, stride_h, stride_w, 1, 1,
+                                       *col_mat);
+
+            const int m = weight_mat->n;
+            const int n = output_mat->h * output_mat->w;
+            const int k = weight_mat->total() / weight_mat->n;
+>>>>>>> parent of be606fa... use bgemm_naive instead of bconv_naive as fallback due to the 128-align weight
             bgemm(m, n, k, static_cast<uint64_t *>(transposed_weight_mat->data),
                   m, static_cast<uint64_t *>(col_mat->data), k,
                   static_cast<float *>(output_mat->data), m);
         } else {
+<<<<<<< HEAD
             bgemm_naive(m, n, k,
                         static_cast<uint64_t *>(transposed_weight_mat->data), m,
                         static_cast<uint64_t *>(col_mat->data), k,
                         static_cast<float *>(output_mat->data), m);
+=======
+            pack_mat(*input_mat, *binarized_mat);
+            baseline_bconv(*binarized_mat, *weight_mat, weight_mat->h,
+                           weight_mat->w, pad_h, pad_w, stride_h, stride_w, 1,
+                           1, output_mat->c, *output_mat);
+>>>>>>> parent of be606fa... use bgemm_naive instead of bconv_naive as fallback due to the 128-align weight
         }
     }
 }