| 13 | |---------------- |
| 14 | {{{#!td style="vertical-align:top;width: 50px" |
| 15 | 18/10/14 |
| 16 | }}} |
| 17 | {{{#!td style="vertical-align:top;width: 50px" |
| 18 | SR |
| 19 | }}} |
| 20 | {{{#!td style="vertical-align:top;width: 75px" |
| 21 | r1482 |
| 22 | }}} |
| 23 | {{{#!td style="vertical-align:top" |
| 24 | 3.10 |
| 25 | }}} |
| 26 | {{{#!td style="vertical-align:top" |
| 27 | N,C,B |
| 28 | }}} |
| 29 | {{{#!td style="vertical-align:top" |
| 30 | '''New:''' |
| 31 | |
| 32 | Adjustments for using CUDA-aware OpenMPI. The local MPI rank is determined from environment variable {{{OMPI_COMM_WORLD_LOCAL_RANK}}}. (palm) |
| 33 | |
| 34 | '''Changed:''' |
| 35 | |
| 36 | Use 2d-decomposition, if accelerator boards are used. (poisfft) |
| 37 | |
| 38 | '''Bugfixes:''' |
| 39 | |
| 40 | cudafft workaround for data declaration of {{{ar_tmp}}} because of PGI 14.1 bug. (fft_xy) \\ |
| 41 | Information about calculated or predefined virtual processor topology adjusted. (header) \\ |
| 42 | Missing {{{ngp_sums_ls}}} added in accelerator version. (flow_statistics) |
| 43 | }}} |