Optimize nearest_interp backward

paddle/fluid/operators/interpolate_v2_op.cu

@@ -199,32 +199,66 @@ __global__ void KeNearestNeighbor3DInterpFw(
   }
 }
 
+template <typename T>
+__global__ void KeNearestNeighborInterpNCHWBw(
+    T* in, const size_t in_img_h, const size_t in_img_w, const T* out,
+    const size_t out_img_h, const size_t out_img_w, const size_t nc,
+    const float ratio_h, const float ratio_w, const bool align_corners) {
+  int out_img_idx = threadIdx.x + blockIdx.x * blockDim.x;
+  int out_img_idy = threadIdx.y + blockIdx.y * blockDim.y;
+  int nc_id = threadIdx.z + blockIdx.z * blockDim.z;
+  int nc_stride = blockDim.z * gridDim.z;
+
+  // nearest_sampling by multiple read in_addr and write to out_addr
+  int in_img_idx = (align_corners)
+                       ? static_cast<int>(ratio_w * out_img_idx + 0.5)
+                       : static_cast<int>(ratio_w * out_img_idx);
+  int in_img_idy = (align_corners)
+                       ? static_cast<int>(ratio_h * out_img_idy + 0.5)
+                       : static_cast<int>(ratio_h * out_img_idy);
+
+  int in_index = (nc_id * in_img_h + in_img_idy) * in_img_w + in_img_idx;
+  int in_index_stride = nc_stride * in_img_h * in_img_w;
+
+  int out_index = (nc_id * out_img_h + out_img_idy) * out_img_w + out_img_idx;
+  int out_index_stride = nc_stride * out_img_h * out_img_w;
+
+  // prevent from multiple threads writing
+  if (out_img_idx < out_img_w && out_img_idy < out_img_h) {
+    while (nc_id < nc) {
+      T* in_pos = &in[in_index];
+      const T out_pos = out[out_index];
+      platform::CudaAtomicAdd(in_pos, out_pos);
+      in_index += in_index_stride;
+      out_index += out_index_stride;
+      nc_id += nc_stride;
+    }
+  }
+}
+
 template <typename T>
 __global__ void KeNearestNeighborInterpBw(
     T* in, const size_t in_img_h, const size_t in_img_w, const size_t input_h,
     const size_t input_w, const T* out, const size_t out_img_h,
     const size_t out_img_w, const size_t output_h, const size_t output_w,
     const size_t num_channels, const float ratio_h, const float ratio_w,
-    const bool align_corners, const DataLayout data_layout) {
+    const bool align_corners, FastDivModForInterpolate divmods) {
   int nthreads = output_h * output_w;
   int tid = blockIdx.x * blockDim.x + threadIdx.x;
   int stride = blockDim.x * gridDim.x;
+  int in_img_size = in_img_h * in_img_w;
+  int out_img_size = out_img_h * out_img_w;
+
   for (; tid < nthreads; tid += stride) {
-    int out_id_h = tid / output_w;
-    int out_id_w = tid % output_w;
-    int in_img_size = input_w / num_channels;
-    int out_img_size = output_w / num_channels;
+    auto out_id_divmod = divmods.output_w_div.Divmod(tid);
+    int out_id_h = out_id_divmod.val[0];
+    int out_id_w = out_id_divmod.val[1];
 
-    int channel_id, out_img_idy, out_img_idx;
-    if (data_layout == DataLayout::kNCHW) {
-      channel_id = out_id_w / out_img_size;
-      out_img_idy = (out_id_w % out_img_size) / out_img_w;
-      out_img_idx = tid % out_img_w;
-    } else {
-      out_img_idy = out_id_w / (out_img_w * num_channels);
-      out_img_idx = out_id_w % (out_img_w * num_channels) / num_channels;
-      channel_id = tid % num_channels;
-    }
+    int channel_id = divmods.channels_div.Divmod(tid).val[1];
+    auto outimg_id_divmod = divmods.output_wc_div.Divmod(out_id_w);
+    int out_img_idy = outimg_id_divmod.val[0];
+    int out_img_idx =
+        divmods.channels_div.Divmod(outimg_id_divmod.val[1]).val[0];
 
     int in_img_idy = (align_corners)
                          ? static_cast<int>(ratio_h * out_img_idy + 0.5)
@@ -233,15 +267,10 @@ __global__ void KeNearestNeighborInterpBw(
                          ? static_cast<int>(ratio_w * out_img_idx + 0.5)
                          : static_cast<int>(ratio_w * out_img_idx);
 
-    T* in_pos;
-    if (data_layout == DataLayout::kNCHW) {
-      in_pos = &in[out_id_h * input_w + channel_id * in_img_size +
-                   in_img_idy * in_img_w + in_img_idx];
-    } else {
-      in_pos = &in[out_id_h * input_w + in_img_idy * in_img_w * num_channels +
-                   in_img_idx * num_channels + channel_id];
-    }
-    const T out_pos = out[out_id_h * output_w + out_id_w];
+    T* in_pos = &in[out_id_h * input_w + in_img_idy * in_img_w * num_channels +
+                    in_img_idx * num_channels + channel_id];
+
+    const T out_pos = out[tid];
     platform::CudaAtomicAdd(in_pos, out_pos);
   }
 }
@@ -1812,11 +1841,26 @@ static void Interpolate2DCUDABwd(const framework::ExecutionContext& ctx,
       platform::GetGpuLaunchConfig1D(ctx.cuda_device_context(), pixelNum);
 
   if ("nearest" == interp_method) {
-    KeNearestNeighborInterpBw<
-        T><<<config.block_per_grid, config.thread_per_block, 0,
-             ctx.cuda_device_context().stream()>>>(
-        input_grad_data, in_h, in_w, n, in_chw, output_grad_data, out_h, out_w,
-        n, out_chw, c, ratio_h, ratio_w, align_corners, data_layout);
+    if (data_layout == DataLayout::kNCHW) {
+      // get launch 3D config
+      int nc = n * c;
+      platform::GpuLaunchConfig config_3d =
+          GetGpuLaunchConfig3D(ctx.cuda_device_context(), nc, out_h, out_w);
+      KeNearestNeighborInterpNCHWBw<
+          T><<<config_3d.block_per_grid, config_3d.thread_per_block, 0,
+               ctx.cuda_device_context().stream()>>>(
+          input_grad_data, in_h, in_w, output_grad_data, out_h, out_w, nc,
+          ratio_h, ratio_w, align_corners);
+    } else {
+      int64_t cw = c * out_w;
+      auto interp_divmods = FastDivModForInterpolate(c, out_chw, cw);
+      KeNearestNeighborInterpBw<
+          T><<<config.block_per_grid, config.thread_per_block, 0,
+               ctx.cuda_device_context().stream()>>>(
+          input_grad_data, in_h, in_w, n, in_chw, output_grad_data, out_h,
+          out_w, n, out_chw, c, ratio_h, ratio_w, align_corners,
+          interp_divmods);
+    }
   } else if ("bilinear" == interp_method) {
     const T align_type_value = (align_mode == 0 && !align_corners) ? 0.5f : 0;
     bool is_nchw = (data_layout == DataLayout::kNCHW) ? true : false;