bev not good, suggest one about sfa3d HOT 4 CLOSED

cudaMalloc((void**)&cnt_gpu_, params_->bev_size_ * sizeof(int));
cudaMalloc((void**)&bev_gpu_, params_->bev_size_ * 3 * sizeof(float));

if(pc[id*4] <= min_x || pc[id*4] >= max_x
	|| pc[id*4+1] <= min_y || pc[id*4+1] >= max_y
	|| pc[id*4+2] <= min_z || pc[id*4+2] >= max_z)
{
	pc[id*4] = 0;
	pc[id*4+1] = 0;
	pc[id*4+2] = 0;
	pc[id*4+3] = 0;
}

int x = __float2int_rd(floorf(pc[id * 4] / xgrid) + xoffset);
int y = __float2int_rd(floorf(pc[id * 4 + 1] / ygrid) + yoffset);
if(x >= bev_height || y >= bev_width)
{
	return ;
}

int map_id = x * bev_width + y;

float intensity = (pc[id * 4 + 3] / 255.0f);
float zcoor = (pc[id * 4 + 2] - z_min) / z_size;

atomicAdd(bev_map+map_id, intensity);
atomicAdd(bev_map+bev_size+map_id, zcoor);
atomicAdd(cnt+map_id, (int)1);

int max_cnt = max(1, cnt[id]);
float zcoor = bev_map[id+bev_size] / (float)max_cnt;
float intensity = bev_map[id] / (float)max_cnt;

float rate = logf((float)cnt[id] + 1.0f) / logf(max_unique);
float map_rate = min(1.0f, rate);

bev_map[bev_size*2+id] = map_rate;
bev_map[bev_size+id] = zcoor;
bev_map[id] = intensity;

printf("%f, %f, %f, %f\n", params_->bound_grid_x_, params_->bound_grid_y_, 
	(float)params_->offset_x_, (float)params_->offset_y_);

dim3 thread1d; dim3 block1d; thread1d.x = 1024;
block1d.x = (points_num + thread1d.x - 1) / thread1d.x;
cuda_range_filter<<<block1d, thread1d>>>(points_gpu, params_->bound_min_x_, params_->bound_max_x_,
	params_->bound_min_y_, params_->bound_max_y_, params_->bound_min_z_, params_->bound_max_z_, points_num);
cudaDeviceSynchronize();

cuda_generate_info<<<block1d, thread1d>>>(points_gpu, params_->bound_grid_x_, params_->bound_grid_y_, 
	(float)params_->offset_x_, (float)params_->offset_y_, params_->bound_min_z_, params_->bound_size_z_, 
	bev_gpu_, params_->bev_height_, params_->bev_width_, params_->bev_size_, cnt_gpu_, points_num);
cudaDeviceSynchronize();

block1d.x = (params_->bev_size_ + thread1d.x - 1) / thread1d.x;
cuda_generate_bev<<<block1d, thread1d>>>(bev_gpu_, cnt_gpu_, (float)params_->max_unique_, params_->bev_size_);
cudaDeviceSynchronize();