|
1 | 1 | Source: nccl |
2 | | -Section: universe/science |
| 2 | +Section: universe/libs |
3 | 3 | Maintainer: Boris Fomitchev < [email protected]> |
4 | 4 | Priority: optional |
5 | | -Build-depends: debhelper(>=9), nvidia-352-dev, cuda-toolkit-7-0, |
| 5 | +Build-depends: debhelper(>=9), cuda-toolkit-7-0, |
6 | 6 | cuda-curand-dev-7-0, cuda-cublas-dev-7-0, cuda-cudart-dev-7-0, cuda-ld-conf-7-0 |
7 | 7 | Standards-Version: 3.9.5 |
8 | 8 |
|
9 | | -Package: nccl |
10 | | -Section: universe/tools |
| 9 | +Package: libnccl1 |
| 10 | +Section: universe/libs |
11 | 11 | Architecture: amd64 |
12 | | -Depends: ${misc:Depends}, ${shlibs:Depends}, nvidia-352, |
| 12 | +Depends: ${misc:Depends}, ${shlibs:Depends}, |
13 | 13 | cuda-curand-7-0, cuda-cublas-7-0, cuda-cudart-7-0, cuda-ld-conf-7-0 |
14 | | -Description: Optimized primitives for collective multi-GPU communication |
15 | | - NCCL (pronounced "Nickel") is a stand-alone library of standard collective communication routines, |
16 | | - such as all-gather, reduce, broadcast, etc., that have been optimized to achieve high bandwidth over PCIe. |
17 | | - NCCL supports up to eight GPUs and can be used in either single- or multi-process (e.g., MPI) applications. |
| 14 | +Description: NVIDIA Communication Collectives Library (NCCL) Runtime |
| 15 | + NCCL (pronounced "Nickel") is a stand-alone library of standard collective |
| 16 | + communication routines,such as all-gather, reduce, broadcast, etc., that have |
| 17 | + been optimized to achieve high bandwidth over PCIe. NCCL supports up to eight |
| 18 | + GPUs and can be used in either single- or multi-process (e.g., MPI) |
| 19 | + applications. |
18 | 20 |
|
19 | | -Package: nccl-dev |
20 | | -Section: universe/tools |
| 21 | +Package: libnccl-dev |
| 22 | +Section: universe/libdevel |
21 | 23 | Architecture: amd64 |
22 | | -Depends: ${misc:Depends}, ${shlibs:Depends}, nccl (= ${binary:Version}) |
23 | | -Description: Dev package for NCCL library |
24 | | - NCCL (pronounced "Nickel") is a stand-alone library of standard collective communication routines, |
25 | | - such as all-gather, reduce, broadcast, etc., that have been optimized to achieve high bandwidth over PCIe. |
26 | | - NCCL supports up to eight GPUs and can be used in either single- or multi-process (e.g., MPI) applications. |
| 24 | +Depends: ${misc:Depends}, ${shlibs:Depends}, libnccl1 (= ${binary:Version}) |
| 25 | +Description: NVIDIA Communication Collectives Library (NCCL) Development Files |
| 26 | + NCCL (pronounced "Nickel") is a stand-alone library of standard collective |
| 27 | + communication routines,such as all-gather, reduce, broadcast, etc., that have |
| 28 | + been optimized to achieve high bandwidth over PCIe. NCCL supports up to eight |
| 29 | + GPUs and can be used in either single- or multi-process (e.g., MPI) |
| 30 | + applications. |
| 31 | + |
0 commit comments