kernels.h

浏览该文件的文档.

/******************************************************************************
 * Copyright 2020 The Apollo Authors. All Rights Reserved.
 *
 * Licensed under the Apache License, Version 2.0 (the "License");
 * you may not use this file except in compliance with the License.
 * You may obtain a copy of the License at
 *
 * http://www.apache.org/licenses/LICENSE-2.0
 *
 * Unless required by applicable law or agreed to in writing, software
 * distributed under the License is distributed on an "AS IS" BASIS,
 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 * See the License for the specific language governing permissions and
 * limitations under the License.
 *****************************************************************************/
 
#pragma once
 
#include <iostream>
 
#include "modules/perception/common/base/common.h"
 
namespace apollo {
namespace perception {
namespace inference {
 
#define NUM_2D_BOX_CORNERS_MACRO 4
#define NUM_THREADS_MACRO 64
 
#define CUDA_KERNEL_LOOP(i, n)                                 \
  for (int i = blockIdx.x * blockDim.x + threadIdx.x; i < (n); \
       i += blockDim.x * gridDim.x)
 
#define DIVUP(m, n) ((m) / (n) + ((m) % (n) > 0))
 
void bbox_transform_inv_cuda(int block_size, int thread_size, int shared_mem,
                             cudaStream_t stream, const int nthreads,
                             const float *boxes, const float *deltas,
                             const int num_box, const int num_channel,
                             float *out_boxes);
 
void clip_boxes_cuda(int block_size, int thread_size, int shared_mem,
                     cudaStream_t stream, const int nthreads, float *boxes,
                     const float height, const float width);
 
void filter_boxes_cuda(
    int block_size, int thread_size, int shared_mem, cudaStream_t stream,
    const int nthreads, const float *boxes, const float *scores,
    const float *all_probs, const int num_box, const int num_channel,
    const int num_class, const int num_prob, const int filter_channel,
    const int filter_class, const int min_size_mode, const float min_size_h,
    const float min_size_w, const float threshold_score, float *filtered_boxes,
    float *filtered_scores, float *filtered_all_probs, int *filtered_count);
 
void keep_topN_boxes_cuda(int block_size, int thread_size, int shared_mem,
                          cudaStream_t stream, const int nthreads,
                          const float *boxes, const float *scores,
                          const float *all_probs, const int *indexes,
                          const int *count, const bool keep_score,
                          const int num_box, const int num_prob, const int topN,
                          float *out_boxes, float *out_scores,
                          float *out_all_probs);
 
void repeatedly_add_cuda(int block_size, int thread_size, int shared_mem,
                         cudaStream_t stream, const int nthreads,
                         const float *in_data, float *out_data,
                         const float *add_vec, int add_vec_size);
 
void repeatedly_mul_cuda(int block_size, int thread_size, int shared_mem,
                         cudaStream_t stream, const int nthreads,
                         const float *in_data, float *out_data,
                         const float *mul_vec, int mul_vec_size);
 
void slice2d_cuda(int block_size, int thread_size, int shared_mem,
                  cudaStream_t stream, const int nthreads, const float *in_data,
                  float *out_data, const int *slice_axises, int slice_axis_num,
                  int input_axis_size);
 
void NmsForward(bool rpn_proposal_output_score, int host_filter_count,
                int num_box_corners, float nms_overlap_threshold,
                int num_candidate, int top_n, int batch_id, int num_prob,
                float *dev_sorted_box_for_nms, float *scores, float *all_probs,
                float *out_boxes, int *acc_box_num, cudaStream_t stream);
 
}  // namespace inference
}  // namespace perception
}  // namespace apollo