MD simulation of HAdV3 hexon homotrimer in water 总体系:368,286 atoms Gromacs 4.5.3 软件包64 bit 第一阶段:从1 cores增加到32 cores 第二阶段:从8个节点(64 cores)到24个节点(192 cores)进行统计 曙光高性能集群系统 Intel(R) Xeon(R) CPU E5430 @ 2.66GHz 结论: Gromacs4.5的并行效率很高,在约30万个原子的模拟体系计算中,一直到192个核心也没有出现效能平台,并行效率一直是线性增长(再多的核心数没有尝试),这也可能与体系的大小有关,体系越大,核心数就越多越好。对于一定大小的的体系,理论上当核心数达到一定规模的时候会出现效能平台期。 _________________________________________________________ 1 cores: NODE (s) Real (s) (%) Time: 318.370 318.414 100.0 5:18 (Mnbf/s) (GFlops) (ns/day) (hour/ns) Performance: 42.231 2.689 0.093 258.585 gcq#127: "It's Not Your Fault" (Pulp Fiction) __________________________________________________________ 2 cores: Parallel run - timing based on wallclock. NODE (s) Real (s) (%) Time: 126.752 126.752 100.0 2:06 (Mnbf/s) (GFlops) (ns/day) (hour/ns) Performance: 99.875 6.352 0.219 109.345 __________________________________________________________ 3 cores: Parallel run - timing based on wallclock. NODE (s) Real (s) (%) Time: 124.612 124.612 100.0 2:04 (Mnbf/s) (GFlops) (ns/day) (hour/ns) Performance: 145.687 9.264 0.320 74.923 __________________________________________________________ 4 cores: Parallel run - timing based on wallclock. NODE (s) Real (s) (%) Time: 98.066 98.066 100.0 1:38 (Mnbf/s) (GFlops) (ns/day) (hour/ns) Performance: 201.129 12.791 0.442 54.264 __________________________________________________________ 5 cores: Parallel run - timing based on wallclock. NODE (s) Real (s) (%) Time: 949.204 949.204 100.0 15:49 (Mnbf/s) (GFlops) (ns/day) (hour/ns) Performance: 227.260 14.451 0.501 47.922 __________________________________________________________ 6 cores: Parallel run - timing based on wallclock. NODE (s) Real (s) (%) Time: 95.109 95.109 100.0 1:35 (Mnbf/s) (GFlops) (ns/day) (hour/ns) Performance: 265.130 16.859 0.583 41.152 __________________________________________________________ 8 cores: Parallel run - timing based on wallclock. NODE (s) Real (s) (%) Time: 125.620 125.620 100.0 2:05 (Mnbf/s) (GFlops) (ns/day) (hour/ns) Performance: 338.097 21.504 0.744 32.250 gcq#96: "Proceed, With Fingers Crossed" (TeX) __________________________________________________________ 12 cores: Parallel run - timing based on wallclock. NODE (s) Real (s) (%) Time: 76.288 76.288 100.0 1:16 (Mnbf/s) (GFlops) (ns/day) (hour/ns) Performance: 515.582 32.780 1.135 21.149 gcq#54: "And It Goes a Little Something Like This" (Tag Team) __________________________________________________________ 16 cores: Parallel run - timing based on wallclock. NODE (s) Real (s) (%) Time: 313.087 313.087 100.0 5:13 (Mnbf/s) (GFlops) (ns/day) (hour/ns) Performance: 656.646 41.762 1.447 16.591 gcq#151: "I Ripped the Cord Right Out Of the Phone" (Capt. Beefheart) __________________________________________________________ 64 cores:5 x 4 x 2 NODE (s) Real (s) (%) Time: 61.191 61.191 100.0 1:01 (Mnbf/s) (GFlops) (ns/day) (hour/ns) Performance: 2470.244 156.950 5.425 4.424 gcq#44: "You Try to Run the Universe" (Tricky) __________________________________________________________ 80 cores:5 x 2 x 5 NODE (s) Real (s) (%) Time: 56.262 56.262 100.0 (Mnbf/s) (GFlops) (ns/day) (hour/ns) Performance: 2799.056 177.866 6.146 3.905 gcq#308: "I believe in miracles cause I'm one" (The Ramones) __________________________________________________________ 96 cores:8 x 4 x 2 NODE (s) Real (s) (%) Time: 58.090 58.090 100.0 (Mnbf/s) (GFlops) (ns/day) (hour/ns) Performance: 3333.504 211.844 7.321 3.278 gcq#61: "Would You Like to Be the Monster Tonight ?" (Captain Beefheart) __________________________________________________________ 112 cores:8 x 3 x 3 NODE (s) Real (s) (%) Time: 76.179 76.179 100.0 1:16 (Mnbf/s) (GFlops) (ns/day) (hour/ns) Performance: 3731.124 237.090 8.191 2.930 gcq#314: "Don't You Wish You Never Met Her, Dirty Blue Gene?" (Captain Beefheart) __________________________________________________________ 144 cores:8 x 6 x 2 NODE (s) Real (s) (%) Time: 50.006 50.006 100.0 (Mnbf/s) (GFlops) (ns/day) (hour/ns) Performance: 4675.533 297.155 10.267 2.338 gcq#58: "I Could Take You Home and Abuse You" (Magnapop) __________________________________________________________ 160 cores:4 x 9 x 3 NODE (s) Real (s) (%) Time: 54.993 54.993 100.0 (Mnbf/s) (GFlops) (ns/day) (hour/ns) Performance: 4709.033 299.315 10.341 2.321 gcq#336: "Ohne Arbeit wird das Leben Oede" (Wir Sind Helden) __________________________________________________________ 176 cores:8 x 3 x 5 NODE (s) Real (s) (%) Time: 40.981 40.981 100.0 (Mnbf/s) (GFlops) (ns/day) (hour/ns) Performance: 5108.852 324.781 11.220 2.139 gcq#62: "Meet Me At the Coffee Shop" (Red Hot Chili Peppers) __________________________________________________________ 192 cores:8 x 8 x 2 NODE (s) Real (s) (%) Time: 42.517 42.517 100.0 (Mnbf/s) (GFlops) (ns/day) (hour/ns) Performance: 5924.472 376.706 13.010 1.845 gcq#230: "Encountered Subspace Anomaly" (Star Trek) __________________________________________________________