稀疏对称矩阵的LDLT分解在GPU上的高效实现
陈鑫峰,王武

An Effective Implementation of LDLT Decomposition of Sparse Symmetric Matrix on GPU
Chen Xinfeng,Wang Wu
算法2. 数据依赖图的划分
1: 为lev[]分配内存,初值设为0
2: nlevel = 1
3: for k = 0 to n - 1 do
4: if parent[k]≠ -1 then
5: lev[parent[k]] = max(lev[parent[k]], lev[k]+1);
6: end if
7: nlevel = max(nlevel, lev[k]+1);
8: end for
9: 为c[]分配内存, 大小为nlevel,初值设为0
10: for k = 0 to n- 1 do
11: c[lev[k]]++;
12: end for
13: 为level_p[]分配内存, 大小为nlevel+1
14: level_p[0] = 0;
15: for k = 1 to nlevel do
16: level_p[k] = level_p[k-1]+c[k-1];
17: c[k-1] = level_p[k-1]
18: end for
19: 为level_i 分配内存空间, 大小为n
20: for k = 0 to n - 1 do
21: level_i[c[lev[k]]++] = k;
22: end for
23: 释放lev[]和c[]