稀疏矩陣的矩陣向量乘法的并行算法性能

1、 附錄一：數(shù)據(jù)結(jié)構(gòu)的 C 語(yǔ)言定義 typedef struct int row; float val; Column; typedef struct int col; int nonzeros; int colsize; Column *index; ColumnIndex; typedef struct int rowsize,colsize; int nonzeros; int col,from; ColumnIndex *column; Matrix; typedef struct int size; float *element; Vector; 附錄二：ACI 機(jī)群的配置 CPU

2、: PIII 733MHz Cache : 256K Networks: Ethernet 附錄三：ACI 機(jī)群上的實(shí)驗(yàn)結(jié)果 size 9282 9282 9282 9282 16482 16482 16482 16482 25210 25210 25210 25210 222490 222490 nonzeros 63926 63926 63926 63926 114010 114010 114010 114010 174726 174726 174726 174726 1552294 1552294 numprocs 1 2 4 8 1 2 4 8 1 2 4 8 1 2 time (se

3、c MPI 0.004680 0.009800 0.015596 0.02522 0.008337 0.017426 0.027885 0.044943 0.013355 0.027088 0.042559 0.068129 0.145393 0.256043 PART 0.005573 0.004983 0.007401 0.010398 0.009939 0.009491 0.021572 0.016142 0.015433 0.015500 0.027846 0.022796 0.164053 0.174237 同步 0.004765 0.010043 0.014210 0.062890

4、 0.008339 0.014981 0.035804 0.069530 0.013431 0.024147 0.060149 0.113717 0.146757 0.255025 交替 0.004713 0.009949 0.015427 0.024222 0.008278 0.017355 0.027700 0.039580 0.013409 0.026250 0.042073 0.059829 0.145358 0.252532 輪轉(zhuǎn) 0.004651 0.009766 0.022323 0.050307 0.008273 0.017535 0.039853 0.088247 0.013

5、328 0.026463 0.060868 0.135810 0.145497 0.252111 第 11 頁(yè) 共 11 頁(yè) 222490 1552294 4 0.410589 0.386611 0.550810 0.395246 0.212587 參考文獻(xiàn) Bruce Hendrickson, Robert Leland and Steve Plimpton, An Efficient Parallel Algorithm for Matrix-Vector Multiplication, Sandia National Laboratories, Albuquerque, NM 87185 L.M. Romero and E.L. Zapata, Data Distributions for Sparse Matrix Vector Multiplication, Un