80.

2023210/main.cpp

@@ -1,7 +1,7 @@
 #include <stdio.h>
 
 struct CalcResult {
-    int dim_sizes[32];
+    int dim_status;
     int min_multiply_times;
 };
 
@@ -11,83 +11,133 @@ int n = 0, k = 0;
 CalcResult *m;
 // CalcResult m[10000];
 // int* m;
+int dim_sizes[32] = {0};
+
+int precalc[2][1048576] = {0};
 
 int tensor_pos(int num, int dim) {
     return k * num + dim;
 }
 
-int setm(int i, int j, int num, int *dims) {
-    if (j < i) {
+int setm(int i, int j, int num) {
+    if (j <= i) {
         return 0;
     }
 
-    m[(2 * n - i + 1) * i / 2 + j - i].min_multiply_times = num;
-    for (int i = 0; i < k - 2; i++) {
-        m[(2 * n - i + 1) * i / 2 + j - i].dim_sizes[i] = dims[i];
+    return m[(2 * n - i + 1) * i / 2 + j - i].min_multiply_times = num;
+}
+
+int getm(int i, int j) {
+    if (j <= i) {
+        return 0;
     }
 
     return m[(2 * n - i + 1) * i / 2 + j - i].min_multiply_times;
 }
 
-CalcResult getm(int i, int j) {
-    if (j <= i) {
-        return CalcResult{{0}, 0};
+int setmdim(int i, int j, int d) {
+    if (j < i) {
+        return 0;
     }
 
-    return m[(2 * n - i + 1) * i / 2 + j - i];
+    return m[(2 * n - i + 1) * i / 2 + j - i].dim_status = d;
 }
 
-CalcResult find_min(int start, int end) {
-    if (start == end || getm(start, end).min_multiply_times != 0) {
+int getmdim(int i, int j) {
+    if (j < i) {
+        return 0;
+    }
+
+    return m[(2 * n - i + 1) * i / 2 + j - i].dim_status;
+}
+
+int find_min(int start, int end) {
+    if (start == end || getm(start, end) != 0) {
         return getm(start, end);
     }
 
     int min = 2147483647;
     int needed_multiply_by_d = 1;
+    int dim_status = getmdim(start, end);
 
     // prepare the dims
-    int segment_dim_size[32] = {0};
-    for (int dim = 0; dim < k - 2; dim++) {
-        for (int tens = start; tens <= end; tens++) {
-            int d_k = tensors[tensor_pos(tens, dim)];
-            if (d_k != 1) {
-                needed_multiply_by_d *= d_k;
-                segment_dim_size[dim] = d_k;
-                break;
-            }
-            segment_dim_size[dim] = 1;
-        }
+    // for (int i = 0; i < k - 2; i++) {
+    //     if ((dim_status >> i) & 1) {
+    //         needed_multiply_by_d *= dim_sizes[i];
+    //     }
+    // }
+    for (int i = 0; i <= (k - 2) / 16; i++) {
+        needed_multiply_by_d *= precalc[i][(dim_status >> (16 * i)) & 0xffff];
     }
 
     for (int split_pos = start; split_pos < end; split_pos++) {
-        // int a = tensors[tensor_pos(start, k - 2)];
-        // a = tensors[tensor_pos(split_pos, k - 1)];
-        // a = tensor_pos(end, k - 1);
         int total_needed = tensors[tensor_pos(start, k - 2)] *
                            tensors[tensor_pos(split_pos, k - 1)] * tensors[tensor_pos(end, k - 1)] *
                            needed_multiply_by_d;
-        CalcResult left = find_min(start, split_pos), right = find_min(split_pos + 1, end);
-        total_needed += left.min_multiply_times + right.min_multiply_times;
+        int left = find_min(start, split_pos), right = find_min(split_pos + 1, end);
+        total_needed += left + right;
         if (total_needed < min) {
             min = total_needed;
         }
     }
 
-    setm(start, end, min, segment_dim_size);
-    return getm(start, end);
+    setm(start, end, min);
+    return min;
+}
+
+void preparedim(int length) {
+    if (length == 1) {
+        for (int i = 0; i < n; i++) {
+            int cur_stat = 0;
+            for (int j = 0; j < k - 2; j++) {
+                if (tensors[tensor_pos(i, j)] != 1) {
+                    cur_stat |= 1 << j;
+                }
+            }
+            setmdim(i, i, cur_stat);
+        }
+        return;
+    }
+
+    preparedim(length - 1);
+    for (int i = length - 1; i < n; i++) {
+        setmdim(i - length + 1, i, getmdim(i - length + 1, i - 1) | getmdim(i, i));
+    }
 }
 
 int main() {
     scanf("%d %d", &n, &k);
     tensors = new unsigned short[n * k];
     m = new CalcResult[(n + 1) * n / 2];
+
+    for (int i = 0; i < 32; i++) {
+        dim_sizes[i] = 1;
+    }
+    // m = new int[(n + 1) * n / 2];
     for (int tensor_count = 0; tensor_count < n; tensor_count++) {
         for (int dim = 0; dim < k; dim++) {
             scanf("%hu", &tensors[tensor_pos(tensor_count, dim)]);
+            if (tensors[tensor_pos(tensor_count, dim)] != 1) {
+                dim_sizes[dim] = tensors[tensor_pos(tensor_count, dim)];
+            }
         }
     }
 
-    printf("%d\n", find_min(0, n - 1).min_multiply_times);
+    for (int i = 0; i < 2; i++) {
+        for (int j = 0; j < 0xffff; j++) {
+            int prod = 1;
+            for (int r = 0; r < 16; r++) {
+                if ((j >> r) & 1) {
+                    prod *= dim_sizes[i * 16 + r];
+                }
+            }
+            precalc[i][j] = prod;
+        }
+    }
+
+    preparedim(n);
+
+    printf("%d\n", find_min(0, n - 1));
 
     return 0;
 }