数据依赖和端口数限制报错，导致II无法到1，Unable to enforce a carried dependence constraint……Unable to schedule 'load' operation

[demi]

需要对一张1000*1024的图像进行两次卷积操作，通过一些手动展开和数组分割的方法减少了一些II，但还是显示具有数据依赖以及端口数受限的问题，更改了几次数组分割的大小和方式，但没什么用，II最低只能是3，不知道之后还可以怎样优化。

#include <iostream>
using namespace std;

void erode(bool input[],bool output[])
{
	int i=0;
	int j=0;
	int k=1;
	//static int sumer_array[4] = {0};
	int sume1 = 0; int sume2 = 0;
	int sum1e1 = 0; int sum1e2 = 0;
	int sum2e1 = 0; int sum2e2 = 0;
	int sum3e1 = 0; int sum3e2 = 0;
	int sum4e1 = 0; int sum4e2 = 0;
	int sum5e1 = 0; int sum5e2 = 0;
	int sum6e1 = 0; int sum6e2 = 0;
	int sum7e1 = 0; int sum7e2 = 0;
	int sum8e1 = 0; int sum8e2 = 0;
	int sume = 0;int sum1e = 0;int sum2e = 0;int sum3e = 0;int sum4e = 0;int sum5e = 0;int sum6e = 0;int sum7e = 0;int sum8e = 0;
	int Q=0;int Q1=0;int Q2=0;int Q3=0;int Q4=0;int Q5=0;int Q6=0;int Q7=0;int Q8=0;
	//int sumd1 = 0; int sumd2 = 0;
	for (i = 0; i < ROW; i++) {
		for (j = 0; j < COL; j=j+6) {
#pragma HLS PIPELINE
			Q = i*COL + j;
			sume1 = input[Q] + input[Q + 1];
			sume2= input[Q + COL] + input[Q + COL + 1];
			sume=sume1+sume2;
			output[Q] = !(sume-4);

			Q1 = i*COL + j+1;
			sum1e1 = input[Q1] + input[Q1 + 1];
			sum1e2= input[Q1 + COL] + input[Q1 + COL + 1];
			sum1e=sum1e1+sum1e2;
			output[Q1] = !(sum1e-4);

			Q2 = i*COL + j+2;
			sum2e1 = input[Q2] + input[Q2 + 1];
			sum2e2= input[Q2 + COL] + input[Q2 + COL + 1];
			sum2e=sum2e1+sum2e2;
			output[Q2] = !(sum2e-4);

			Q3 = i*COL + j+3;
			sum3e1 = input[Q3] + input[Q3 + 1];
			sum3e2= input[Q3 + COL] + input[Q3 + COL + 1];
			sum3e=sum3e1+sum3e2;
			output[Q3] = !(sum3e-4);

			Q4 = i*COL + j+4;
			sum4e1 = input[Q4] + input[Q4 + 1];
			sum4e2= input[Q4 + COL] + input[Q4 + COL + 1];
			sum4e=sum4e1+sum4e2;
			output[Q4] = !(sum4e-4);

			Q5 = i*COL + j+5;
			sum5e1 = input[Q5] + input[Q5 + 1];
			sum5e2= input[Q5 + COL] + input[Q5 + COL + 1];
			sum5e=sum5e1+sum5e2;
			output[Q5] = !(sum5e-4);
		}
	}

}

void dilate(bool input[],bool output[])
{
	int sumd1 = 0; int sumd2 = 0;
	int sum1d1 = 0; int sum1d2 = 0;
	int sum2d1 = 0; int sum2d2 = 0;
	int sum3d1 = 0; int sum3d2 = 0;
	int sum4d1 = 0; int sum4d2 = 0;
	int sum5d1 = 0; int sum5d2 = 0;
	int sum6d1 = 0; int sum6d2 = 0;
	int sum7d1 = 0; int sum7d2 = 0;
	int sum8d1 = 0; int sum8d2 = 0;
	int sumd = 0;int sum1d = 0;int sum2d = 0;int sum3d = 0;int sum4d = 0;int sum5d = 0;int sum6d = 0;int sum7d = 0;int sum8d=0;
	int S=0;int S1=0;int S2=0;int S3=0;int S4=0;int S5=0;int S6=0;int S7=0;int S8=0;
	for (int i = 0; i < ROW; i++) {
		for (int j = 0; j < COL; j=j+6) {
#pragma HLS PIPELINE
			S = i*COL + j;
			sumd1 = input[S] + input[S + 1];
			sumd2= input[S + COL] + input[S + COL + 1];
			sumd=sumd1+sumd2;
			output[S] = max(sumd,0);

			S1 = i*COL + j+1;
			sum1d1 = input[S1] + input[S1 + 1];
			sum1d2= input[S1 + COL] + input[S1 + COL + 1];
			sum1d=sum1d1+sum1d2;
			output[S1] = max(sum1d,0);

			S2 = i*COL + j+2;
			sum2d1 = input[S2] + input[S2 + 1];
			sum2d2= input[S2 + COL] + input[S2 + COL + 1];
			sum2d=sum2d1+sum2d2;
			output[S2] = max(sum2d,0);

			S3 = i*COL + j+3;
			sum3d1 = input[S3] + input[S3 + 1];
			sum3d2= input[S3 + COL] + input[S3 + COL + 1];
			sum3d=sum3d1+sum3d2;
			output[S3] = max(sum3d,0);

			S4 = i*COL + j+4;
			sum4d1 = input[S4] + input[S4 + 1];
			sum4d2= input[S4 + COL] + input[S4 + COL + 1];
			sum4d=sum4d1+sum4d2;
			output[S4] = max(sum4d,0);

			S5 = i*COL + j+5;
			sum5d1 = input[S5] + input[S5 + 1];
			sum5d2= input[S5 + COL] + input[S5 + COL + 1];
			sum5d=sum5d1+sum5d2;
			output[S5] = max(sum5d,0);
		}
	}
}

void detect_position(AXI_VAL img_input[ROW*COL],AXI_VAL end_1_out[select_size] )
{
	static bool image_c[ROW*COL] = {0};
#pragma HLS ARRAY_PARTITION variable=image_c cyclic factor=9 dim=1
	static bool image_d[ROW*COL] = { 0 };
	static bool image_e[ROW*COL] = { 0 };
#pragma HLS ARRAY_PARTITION variable=image_e cyclic factor=9 dim=1

        dilate(image_c, image_e);
        erode(image_e, image_d);
}