a67ae11ce4
Merge pull request #124 from wenkaidu/upstream_sync
...
Upstream sync
2019-08-16 16:41:55 -07:00
7c38da0939
Merge remote-tracking branch 'remotes/nccl/master' into HEAD
2019-08-16 16:13:34 -07:00
72a64e27f3
Merge pull request #123 from wenkaidu/tune_unroll
...
Tune AUTOUNROLL for better performance
2019-08-16 11:15:49 -07:00
1faededc03
Tune AUTOUNROLL for better performance
...
Also remove all unused UNROLL defines
2019-08-16 10:34:53 -07:00
50c2202fe9
Merge pull request #121 from mhbliao/hliao/master/swdev-200061
...
Fix build with hip-clang.
2019-08-15 12:40:46 -07:00
9369f8d75d
Fix build with hip-clang.
...
- Add necessary function attribute for HIP programming model.
- Explicitly include hsa headers.
2019-08-15 14:56:04 -04:00
3f6662f837
Merge pull request #122 from wenkaidu/tune_ll
...
Tune LL threshold for VEGA
2019-08-15 10:33:17 -07:00
2223cccf15
Tune LL threshold for VEGA
...
Also move abort check after SPINS_BEFORE_CHECK_ABORT as NCCL
2019-08-15 09:16:11 -07:00
9af66195db
Merge pull request #120 from wenkaidu/rccl_2.4_update
...
RCCL 2.4 update
2019-08-14 15:21:30 -07:00
4b77a16f3f
Default to minimal 2 rings and improve LL loop
2019-08-14 14:12:56 -07:00
5782a8d857
Remove duplicate line
2019-08-14 13:22:43 -07:00
6827b174c0
Merge remote-tracking branch 'remotes/rccl/master' into rccl_2.4_update
2019-08-14 10:44:18 -07:00
f11c8f60cd
RCCL 2.4 update
2019-08-14 10:42:35 -07:00
ccb1298148
Merge branch 'lowintelligence-shm'
...
PR#196
2019-08-14 10:09:53 -07:00
fad079a8ae
Updated PR#196 to use a common hash function
2019-08-14 10:08:39 -07:00
01d1836668
Merge branch 'shm' of git://github.com/lowintelligence/nccl into lowintelligence-shm
2019-08-14 09:45:45 -07:00
7f2b337e70
Make use of SO_REUSEPORT conditional
...
Fixes : #244
SO_RESUEPORT was introduced in Linux 3.9 and later.
This change allows NCCL to compile against older releases.
The functionality is only required if the user is specifying
a NCCL bootstrap address via an environment variable.
2019-08-13 16:32:07 -07:00
40445c17d8
Adding linkinfo and srcGPU to destGPU info ( #114 )
...
* Adding linkinfo and srcGPU to destGPU info
2019-08-13 09:25:03 -07:00
b3a57dbb33
Merge pull request #116 from stanleytsang-amd/master
...
Removing unnecessary device collective source files.
2019-08-12 18:26:02 -04:00
3a61907182
Removing unnecessary device collective source files.
2019-08-12 18:23:23 +00:00
b8cf48fc16
Adding TransferBench tool ( #113 )
...
* Adding standalone TransferBench tool
2019-08-07 17:21:41 -06:00
f1c727d4ce
Merge pull request #112 from wenkaidu/hdp
...
Get HDP register address from hipDeviceGetAttribute API
2019-08-05 14:27:19 -07:00
84d3344796
Get HDP register address from hipDeviceGetAttribute API
2019-08-05 14:14:09 -07:00
4a9bdd8539
Merge pull request #108 from wenkaidu/xgmi_finegrain
...
Remove dependency to HSA_FORCE_FINE_GRAIN_PCIE flag for XGMI link
2019-08-02 10:00:48 -07:00
315f792f83
Merge pull request #110 from mhbliao/hliao/master/swdev-198268
...
Revise the previous fix to use the canonical path to HSA.
2019-08-01 12:46:25 -07:00
4f2aa06688
Revise the previous fix to use the canonical path to HSA.
...
- This fix the build failures under certain environments.
2019-08-01 14:50:44 -04:00
9189279220
Merge pull request #107 from mhbliao/hliao/master/swdev-198268
...
Fix build with hip-clang
2019-08-01 08:58:37 -07:00
e7022e9196
Remove dependency to HSA_FORCE_FINE_GRAIN_PCIE flag for XGMI link
2019-08-01 04:26:37 +00:00
41310144f6
Fix build with hip-clang
...
Two minor issues are solved:
+ Enclose the kernel function with parenthesis as hip-clang defines
`hipLaunchKernelGGL` as macro.
+ Need to explicitly include <hsa.h> for hip-clang.
2019-07-31 15:07:36 -04:00
bfb3921519
Refine RPM package building spec file.
...
Add /sbin/ldconfig into RPM package install operations.
2019-07-31 10:36:22 -07:00
1969e89003
Add gfx908 target ( #106 )
2019-07-30 13:56:45 -07:00
1fee6f9d50
Remove extra "." from version string ( #104 )
2019-07-25 15:25:02 -07:00
fdee095dd3
Changing to rocm-cmake new style versioning ( #103 )
2019-07-22 23:40:13 +00:00
0522041fac
Use hipExtLaunchMultiKernelMultiDevice API ( #100 )
...
Depends on HIP version with this pull request:
https://github.com/ROCm-Developer-Tools/HIP/pull/1232
2019-07-18 09:02:37 -07:00
4d579e51cc
Fix NIC distances for 11+ NICs
2019-07-17 06:32:33 -07:00
920ae57c14
Fix #224 : prevent number of IB devices from going out of bound
2019-07-17 06:32:33 -07:00
dc1908e944
Increase debug print of ring topology to 64 ranks ( #99 )
2019-07-16 14:54:17 -07:00
43bd6f5fbf
Allocate transport memory based on numa node ( #97 )
2019-07-15 11:45:38 -07:00
c8c68fb5f7
Size up IPC buffers to multiples of 2MB
...
Avoid potential CUDA error in concurrent communicator initialization
2019-07-12 09:50:17 -07:00
7b6332d3d0
Adding explicit HDP flush when using RDMA via Infiniband ( #95 )
...
* Adding explicit HDP flush when using RDMA via Infiniband
2019-07-10 16:29:02 -06:00
0b192d2299
Add the exact matching modifier support "=" to the NCCL_IB_HCA variable ( #236 )
...
Perform exact matching when the prefix "=" is specified in the NCCL_IB_HCA variable to exclude HCAs mlx5_X[0-9]+ when mlx5_X is specified.
2019-07-09 14:45:41 -07:00
70804da15b
Refactor primitive test to support multiple GPUs in rings ( #94 )
...
* Refactor primitive test to support multiple GPUs in rings
* Make GPUs sync before transfer optional
* Use same ring format as RCCL
* Extend to 8 GPUs and report errors if there is no P2P access
* Control GPUs sync before ops from command line with "-s" option
* Change buffer size through command line option "-n"
Rename iterations command line option to "-i"
2019-07-05 14:29:20 -07:00
949d680e49
Fix share memory collision in multi-communicator case. ( #93 )
...
Current SHM object name would only use pidHash and ranks as
identification, which would collide each other when program runs with
multiple communicators. Here we added commId info into pidHash, it makes
'pidHash'es of different communicators keeping in same process will be
distincted with each other.
Ported from original commit: https://github.com/lowintelligence/nccl/commits/shm
2019-07-02 09:27:16 -07:00
e6a0da444f
Match primitives unroll counts with latest RCCL ( #91 )
2019-06-26 15:09:13 -07:00
329a62a01f
Fixing install script to actually install library when requested ( #88 )
...
* Fixing install script to actually install library when requested. Cleaning up unused code.
Removing unused arguments from install script.
Fixing weird whitespacing
* Fixing install script to install to correct location /opt/rocm, now creates symlink in /opt/rocm/lib
* Updates and corrections to README and install script
2019-06-25 17:25:21 -06:00
8e04d80382
Merge branch 'master' into HEAD
2019-06-25 13:39:08 -07:00
7c72dee660
2.4.8-1
...
Fix #209 : improve socket transport performance
Split transfers over multiple sockets
Launch multiple threads to drive sockets
Detect AWS NICs and set nsockets/nthreads accordingly
2019-06-25 13:22:47 -07:00
840f8715ef
Changing maintainer to no-reply to fix deb generation ( #86 )
2019-06-24 17:13:57 -06:00
0de9051ace
Merge pull request #83 from ROCmSoftwarePlatform/devel
...
Devel to Master
2019-06-24 14:25:18 -06:00
1c7b0bd878
Fixing Centos 7 Packaging and package versioning/maintainer ( #82 )
...
- Fixing Centos 7 Packaging
- standardizing version numbers for release to use rocm versioning
- removing maintainer email based on legal's input
2019-06-24 14:22:16 -06:00
Wenkai Du
Wenkai Du
Michael LIAO
David Addison
rpathani
Stanley Tsang
gilbertlee-amd
Cao Zongyan
saadrahim
Ke Wen
Hirochika Asai