The results of the GraphHPC-2015
As a part of the GraphHPC-2015 workshop a contest was held on the most efficient implementation of the Minimum Spanning Tree (MST) problem.
The contest was conducted using an automatic contest system. The task is described on the Problem page. The following computer systems were provided to run the tasks:
Multi-node cluster with 8 computing nodes connected by dual-port Infiniband FDR x4 interconnect (fat tree topology), consisting of x2 6-core Intel Xeon E5-2630 2.3 GHz (HT-enabled) processors and 78GB of DDR3 DRAM.
Single computing node IBM SystemX 3650 consisting of x2 Intel Xeon E5-2690 2.9 GHz (HT-enabled), NVIDIA GPU K20x (ECC is disabled), 128GB of DDR3 DRAM.
Dates: Contest started on January 20 and ended on March 2, 2015. The results were announced on March 5, 2015 at the GraphHPC-2015 workshop.
Final results
Alexander Daryin from T-Platforms won the first place in the GraphHPC contest with a score of 3080 MTEPS. Fedor Minyuk from Europlan took the second place with a score of 2488 MTEPS. Alexander Kolganov was third with a score of 1655 MTEPS. In the student competition the winner was Artem Mazeev from CMC MSU with a score of 1651 MTEPS, Denis Kuplyakov and Vadim Zaytsev won the second and the third places respectively.
The first place winners (in student and general classification) were awarded with SSD disks (Intel 530 Series, 240 GB). The second place winner was awarded with NVIDIA Jetson TK1 DevKit. Other winners received “Concrete Mathematics: A Foundation for Computer Science” books. All participants received certificates and souvenirs from the GraphHPC-2015 partners: NVIDIA, IBM, T-Platforms, Intel, AMD, and Sapphire.
Results
Node
| Participant | Organization | Student | System | Scale | RMAT MTEPS | SSCA2 MTEPS | Average MTEPS | |
|---|---|---|---|---|---|---|---|---|
| 1 | Дарьин Александр | Т-Платформы | 2x Intel Xeon E5-2690 NVIDIA Tesla K20x | 17 | 4 491.52 | 1 669.02 | 3 080.27 | |
| 2 | Минюк Фёдор | ЗАО «Анти-Плагиат» | 2x Intel Xeon E5-2690 | 16 | 1 830.38 | 3 145.58 | 2 487.98 | |
| 3 | Дарьин Александр | Т-Платформы | 2x Intel Xeon E5-2690 | 17 | 2 834.77 | 1 630.92 | 2 232.84 | |
| 4 | Колганов Александр | ВМК МГУ | Intel Xeon E5-1660v2 NVIDIA GTX Titan | 24 | 1 479.00 | 2 873.00 | 2 176.00 | |
| 5 | Колганов Александр | ВМК МГУ | 2x Intel Xeon E5-2690 NVIDIA Tesla K20x | 22 | 1 289.64 | 2 021.02 | 1 655.33 | |
| 6 | Мазеев Артём | ВМК МГУ | yes | 2x Intel Xeon E5-2690 | 16 | 1 865.46 | 1 436.20 | 1 650.83 |
| 7 | Купляков Денис | ВМК МГУ | yes | 2x Intel Xeon E5-2690 | 15 | 1 766.85 | 1 439.08 | 1 602.97 |
| 8 | Зайцев Вадим | Новосибирский ГУ | yes | 2x Intel Xeon E5-2690 | 22 | 1 627.51 | 1 508.87 | 1 568.19 |
| 9 | Жуков Владимир | ВМК МГУ | yes | 2x Intel Xeon E5-2690 NVIDIA Tesla K20x | 20 | 938.00 | 1 303.20 | 1 120.60 |
| 10 | Попов Александр | Positive Technologies | 2x Intel Xeon E5-2690 | 12 | 49.74 | 90.05 | 69.89 | |
| 11 | Аладышев Константин | НИЦЭВТ | 2x Intel Xeon E5-2690 NVIDIA Tesla K20x | 17 | 58.04 | 39.72 | 48.88 | |
| 12 | Груздев Арсений | ВМК МГУ | yes | 2x Intel Xeon E5-2690 | 10 | 26.22 | 26.35 | 26.28 |
| 13 | Пример реализации | DISLab | 2x Intel Xeon E5-2690 | 10 | 0.42 | 8.12 | 4.27 |
Cluster
| Participant | Organization | Student | System | # nodes | Scale | RMAT MTEPS | SSCA2 MTEPS | Average MTEPS | |
|---|---|---|---|---|---|---|---|---|---|
| 1 | Дарьин Александр | Т-Платформы | Nicevt Cluster | 8 | 15 | 112.89 | 179.54 | 146.21 | |
| 2 | Пример реализации | DISLab | Nicevt Cluster | 8 | 10 | 0.07 | 0.07 | 0.07 |

