AI/rocm-systems
2
0
Форкнуть 0
Код Задачи Запросы на слияние Действия Пакеты Проекты Релизы Вики Активность
76 333 коммитов 3 Ветки 0 Теги
develop
Граф коммитов

48 Коммитов

Автор SHA1 Сообщение Дата
David Addison 17c8317cb1 NCCL 2.4.6-1 
Added detection of IBM/Power NVLink bridge device.
    Add NUMA support to PCI distance calculations.
    Added NCCL_IGNORE_CPU_AFFINITY env var.
    Fix memory leaks; GithubIssue#180
    Compiler warning fix; GithubIssue#178
    Replace non-standard variable length arrays. GithubIssue#171
    Fix Tree+Shared Memory crash. GithubPR#185
    Fix LL cleanup hang during long running DL jobs.
    Fix NCCL_RINGS environment variable handling.
    Added extra checks to catch repeat calls to ncclCommDestroy() GithubIssue#191
    Improve bootstrap socket connection reliability at scale.
    Fix hostname hashing issue. GithubIssue#187
    Code cleanup to rename all non device files from *.cu to *.cc


[ROCm/rccl commit: f40ce73e89]
 2019-04-05 13:05:45 -07:00
Sylvain Jeaugey 1cc88accd5 Add install target 
Fix issue #145

[ROCm/rccl commit: bed43524cc]
 2018-11-05 09:53:59 -08:00
Sylvain Jeaugey 8ffcfac437 2.3.5-5 
Add support for inter-node communication using sockets and InfiniBand/RoCE.
Improve latency.
Add support for aggregation.
Improve LL/regular tuning.
Remove tests as those are now at github.com/nvidia/nccl-tests .


[ROCm/rccl commit: f93fe9bfd9]
 2018-09-25 14:12:01 -07:00
Sylvain Jeaugey 338f6eaae1 Add support for CUDA9 half semantics 
[ROCm/rccl commit: 29a1a916dc]
 2017-06-14 11:20:24 -07:00
Boris Fomitchev 12e9519ed0 Added Pascal nvcc flags, bumped version 
[ROCm/rccl commit: 649f04d077]
 2017-03-24 11:58:14 -07:00
Sylvain Jeaugey 4644095088 Bumping version to 1.3.3 
[ROCm/rccl commit: 7fef264bfa]
 2017-03-01 16:44:27 -08:00
Sylvain Jeaugey db3103d96a 1.3.2 release 
Broadcast tuning
Better checking of inputs
Copy/reduce code simplification


[ROCm/rccl commit: 34d27771c6]
 2016-12-01 15:17:50 -08:00
Peter Jin 95860f0721 Add a static library target "staticlib" to the Makefile. 
Rename the static library "libnccl_static.a" to disambiguate from the
dynamic libraries.


[ROCm/rccl commit: 5765d608cc]
 2016-11-24 11:31:03 -08:00
Kyle Fernandes, ne Jacobs a35c827aa6 Add Fortran bindings 
[ROCm/rccl commit: 5f2b32e45b]
 2016-11-17 15:33:34 -08:00
Sylvain Jeaugey 5d496d605d Bump to 1.3.1 
[ROCm/rccl commit: 534b9a1697]
 2016-10-13 10:33:05 -07:00
Sylvain Jeaugey d4c7d8014a Add scan tests 
[ROCm/rccl commit: ca330b110a]
 2016-09-22 11:58:33 -07:00
Sylvain Jeaugey b540c2b5cc Heavy code refactoring to remove a lot of code in collectives (~1000 lines). 
Have all collectives use the same args, the same ring, and the same primitives for synchronization between threads with the same pattern.


[ROCm/rccl commit: cabd6848e4]
 2016-09-22 11:57:56 -07:00
Sylvain Jeaugey 996a743a44 Merge pull request #41 from jia-kai/master 
Some minor fixes for compile/usage

[ROCm/rccl commit: 9ee6189bf9]
 2016-09-15 09:45:52 -07:00
Sylvain Jeaugey a59d93cf7a Updated LICENCE.txt 
[ROCm/rccl commit: 75bad643bd]
 2016-08-26 15:08:20 -07:00
jiakai d74f2f154d link library with -lrt; otherwise there is undefined reference to shm_open 
[ROCm/rccl commit: ed401cc29b]
 2016-08-19 18:58:56 +08:00
Sylvain Jeaugey 5779f12272 Improved Deb generation 
[ROCm/rccl commit: 7a1aa6b563]
 2016-07-07 16:31:57 +02:00
Sylvain Jeaugey b865b44417 Fix version number 
[ROCm/rccl commit: 9ae84f5d6b]
 2016-06-16 17:07:42 -07:00
Sylvain Jeaugey 2a9f6574a5 Add a debug level to NCCL and CUDA versions at init 
[ROCm/rccl commit: e51e922924]
 2016-06-16 17:04:41 -07:00
Sylvain Jeaugey cabb9878d2 Increased version to 1.2.3 
[ROCm/rccl commit: 9fcc523485]
 2016-06-15 19:18:13 -07:00
Sylvain Jeaugey 4d7076f7dc Packaging : Generate shlibs.local 
[ROCm/rccl commit: 67d1ab9106]
 2016-06-15 19:03:08 -07:00
Sylvain Jeaugey 5749822385 Move deb to build directory 
[ROCm/rccl commit: da6d2009e0]
 2016-06-15 18:20:10 -07:00
Sylvain Jeaugey d51c6b0f89 Fix make install to use BUILDDIR 
[ROCm/rccl commit: 155132d336]
 2016-06-15 18:20:02 -07:00
Sylvain Jeaugey 0ff5aa2cff Rework debian packaging 
[ROCm/rccl commit: 08ddfe03d2]
 2016-06-15 18:18:44 -07:00
Boris Fomitchev 6e4066f5c5 Gencodes changed to NV recommended 
[ROCm/rccl commit: 177505b757]
 2016-06-06 00:06:18 -07:00
Sylvain Jeaugey fbd1a386db Bump to 1.2.2 
[ROCm/rccl commit: 9d9d8cd59f]
 2016-06-03 17:21:53 -07:00
Sylvain Jeaugey 7ac49d9cdc Better name for GENCODE 
[ROCm/rccl commit: 1657af1567]
 2016-06-03 10:25:37 -07:00
Sylvain Jeaugey 778bbbce93 Makefile improvements 
- Use standard CXX env var
 - Permit redefinition of more env
 - Separate lib from tests


[ROCm/rccl commit: 889ad3d4e6]
 2016-06-02 15:01:03 -07:00
Boris Fomitchev 8265e45d65 Merge pull request #21 from borisfom/master 
Fixed install location, new .deb version

[ROCm/rccl commit: 0629fb62d7]
 2016-04-21 14:46:41 -07:00
Boris Fomitchev 14b094e21b Fixed install location, new .deb version 
[ROCm/rccl commit: 0177cf3ea4]
 2016-04-21 14:10:31 -07:00
Nathan Luehr d3b27aa92d Merge pull request #17 from Hopobcn/master 
Enable compilation with specific g++

[ROCm/rccl commit: 658aca1469]
 2016-04-21 13:25:18 -07:00
Nathan Luehr 6d6120cd1e Moved no-as-needed flag to link rule. 
Avoids link errors for tests linked with nvcc.


[ROCm/rccl commit: 03df4c7759]
 2016-04-19 14:51:03 -07:00
Adam Paszke 3baec150aa Add --no-as-needed to make sure that cudart library gets liked 
[ROCm/rccl commit: c0c959b1be]
 2016-04-13 10:04:38 -04:00
Pau Farré c23edf3d49 Enable compilation with old g++ when the default g++ is not supported (+5.0) 
[ROCm/rccl commit: e30bf95989]
 2016-04-12 12:49:13 +02:00
Boris Fomitchev 3070897cb8 Removing Tegra 
[ROCm/rccl commit: e6f4a83da6]
 2016-03-17 17:25:27 -07:00
Boris Fomitchev 801d962bfe Use arch=5.3 as well 
[ROCm/rccl commit: 62b551798f]
 2016-03-16 23:09:36 -07:00
Boris Fomitchev cf44496d6e Added compute 5.3 
[ROCm/rccl commit: 92d2123d8d]
 2016-03-16 19:24:48 -07:00
Boris Fomitchev 07a4274a66 Preparing for pbuild 
[ROCm/rccl commit: ec3de28ae5]
 2016-03-16 19:23:49 -07:00
Boris Fomitchev 711444221f Updated package version, added manpage 
[ROCm/rccl commit: 941d9da08c]
 2016-02-29 12:10:34 -08:00
Nathan Luehr 16379ec27b Added NCCL error checking to tests. 
Also cleaned up makefile so that tests and lib are not built unnecessarily.

Change-Id: Ia0c596cc2213628de2f066be97615c09bb1bb262
Reviewed-on: http://git-master/r/999627
Reviewed-by: Przemek Tredak <ptredak@nvidia.com>
Tested-by: Przemek Tredak <ptredak@nvidia.com>


[ROCm/rccl commit: 2758353380]
 2016-01-29 11:09:05 -08:00
Sylvain Jeaugey 42c3edc35a Moved tests to separate dir and improved MPI test 
test sources moved to test/ directory.
MPI test displays PASS/FAIL and returns code accordingly.

Change-Id: I058ebd1bd5202d8f38cc9787898b2480100c102b
Reviewed-on: http://git-master/r/936086
Reviewed-by: Przemek Tredak <ptredak@nvidia.com>
Tested-by: Przemek Tredak <ptredak@nvidia.com>


[ROCm/rccl commit: c05312f151]
 2016-01-28 12:56:36 -08:00
Nathan Luehr bb5df3dbb4 Merge pull request #6 from lukeyeager/deb 
Deb packaging

[ROCm/rccl commit: 90af7c73ef]
 2016-01-07 13:06:28 -08:00
Luke Yeager d529f63072 Update deb packaging scripts 
[ROCm/rccl commit: c9da89254b]
 2015-12-18 14:23:34 -08:00
Nathan Luehr 8915ecb588 Merge pull request #5 from lukeyeager/tests-nvml 
Don't link tests with NVML

[ROCm/rccl commit: eb2d869f71]
 2015-12-18 13:36:20 -08:00
Boris Fomitchev b3db23e837 Don't link tests with NVML 
[ROCm/rccl commit: b5400c54df]
 2015-12-18 13:27:55 -08:00
Nathan Luehr 87be580f28 Merge pull request #4 from lukeyeager/build-sm50 
Build SM 5.0 code

[ROCm/rccl commit: a4de6016f8]
 2015-12-18 13:23:48 -08:00
Luke Yeager eff4cd1cb3 Build SM 5.0 code 
[ROCm/rccl commit: dd0884b707]
 2015-12-18 13:19:50 -08:00
Luke Yeager a0d7e36891 Use semantic versioning 
[ROCm/rccl commit: e1634ca6cb]
 2015-12-18 12:02:17 -08:00
Nathan Luehr 529702aec6 Initial release. 
[ROCm/rccl commit: 0673d5f44f]
 2015-11-17 11:30:40 -08:00