d95a71b2ef
Switch to hipLaunchHostFunc for HIP graph ( #667 )
...
[ROCm/rccl commit: 2288e9ae80
2022-12-15 10:16:46 -08:00
431c7d2cb9
Add MSCCL Support ( #658 )
...
* Add MSCCL support
* Add alignment and message size checking
* Fix nRanks checking, in-place and out-of-place tests and group call handling
* Fix hipGraph unit test
* Change MSCCL init warning to INFO
* Revise license info
[ROCm/rccl commit: adafc0f759
2022-12-12 15:51:04 -08:00
bb27f5c8ba
Fix typo in detecting Intel platforms ( #661 )
...
[ROCm/rccl commit: b953544a59
2022-12-07 13:36:11 -08:00
90de5e3a84
decreased warp size for gfx110x ( #655 )
...
[ROCm/rccl commit: eca623df07
2022-12-01 12:19:21 -07:00
416b754b31
Query DMABuf support through HSA runtime API ( #654 )
...
[ROCm/rccl commit: fb9938cffa
2022-11-30 08:53:03 -08:00
2ce8946622
2.16.2-1
...
Add support for CUDA 12.0, drop Kepler (sm_35).
Support for H100 features.
Make socket code more robust and protected. Solves #555 .
Improve performance on large CUDA graphs, reducing dependencies.
Reduce inter-socket bandwidth on AMD CPUs to favor better paths.
Various fixes to ncclCommAbort.
Make service thread polling resistant to EINTR.
Compile with profiling API by default.
Extend NVTX instrumentation with call arguments.
[ROCm/rccl commit: 28189e2df8
2022-11-30 02:31:59 -08:00
55e1008756
Adjust P2P channels on Intel platform ( #653 )
...
[ROCm/rccl commit: 9594bbee3b
2022-11-29 13:57:10 -08:00
cb382a0a6e
Fix incorrect rocm-smi ID conversion ( #648 )
...
[ROCm/rccl commit: 57764f8152
2022-11-21 19:44:39 -08:00
b72619495d
Fix collective trace timestamp format ( #647 )
...
[ROCm/rccl commit: 9cb72a3d0f
2022-11-21 08:11:12 -08:00
ba18b77b07
Fix typo in previous hipify change ( #645 )
...
[ROCm/rccl commit: cf3c32a626
2022-11-15 11:51:47 -08:00
c15a10a9d2
Move hipify to cmake stage
...
Add minimal ROCm/HIP version requirements for Graph support
[ROCm/rccl commit: 562dd87036
2022-11-14 18:10:45 +00:00
ffecb74b1e
Update tuning table and fix topo_expl
...
[ROCm/rccl commit: 94ad7f6f51
2022-11-07 18:24:24 +00:00
4c9c1d41ee
Merge remote-tracking branch 'nccl/master' into develop
...
[ROCm/rccl commit: 9a077e6947
2022-11-03 21:17:42 +00:00
4630365f2d
Fix P2P scheduling
...
[ROCm/rccl commit: 72ef100050
2022-10-31 08:54:34 -07:00
775f1b59ba
Merge tag 'v2.15.5-1'
...
[ROCm/rccl commit: 2f4cb874ba
2022-10-25 01:15:22 -07:00
0b20e8b7e9
2.15.5-1
...
Fix crash with CollnetChain on some node topologies
Fix hang when interleaving the capture of different graphs
Fix hang during init in multi-threaded mode
Fix potential data corruption with LL128 protocol on unaligned buffers.
Fix CPU usage during preconnect
Fixes double-free in the error path for ncclCommInitAll
Workaround hang on H100 with Ring/LL128 on 2 GPUs.
[ROCm/rccl commit: cb111f764a
2022-10-25 00:55:55 -07:00
36e5e02e46
Merge remote-tracking branch 'nccl/master' into develop
...
[ROCm/rccl commit: 4f0e223db4
2022-10-20 15:41:29 +00:00
cf1d4a62e8
Fix missing initialization due to merge error ( #640 )
...
[ROCm/rccl commit: bc8ef779df
2022-10-19 21:20:11 -07:00
76414d3230
Support P2P with invisible devices ( #636 )
...
* Support P2P with invisible devices
* Update copyright year
[ROCm/rccl commit: 9ddf0e0649
2022-10-17 10:24:59 -07:00
7cb9d872ab
Merge pull request #634 from yzygitzh/ziyyang/npkit-fix
...
Apply several fixes to NPKit
[ROCm/rccl commit: 9916a09818
2022-10-17 08:01:24 -07:00
0ca30fb88a
Updating files for missing licenses ( #637 )
...
[ROCm/rccl commit: ebb8b5bf63
2022-10-14 13:49:16 -06:00
d54574b0eb
apply npkit
...
[ROCm/rccl commit: 7d6bbc19d4
2022-10-14 01:28:17 +00:00
37dc333d42
Merge tag 'v2.15.1-1'
...
[ROCm/rccl commit: d128d62238
2022-10-07 11:00:26 -07:00
c9cd7243ed
Fixes a double-free in the error path of ncclCommInitAll.
...
Fixes https://github.com/NVIDIA/nccl/issues/726
[ROCm/rccl commit: 2401f4a918
2022-10-03 17:12:32 -07:00
f2736a4fb3
introduce a hw topology aware bintree
...
for hayabusa architecture.
[ROCm/rccl commit: e645b02cd8
2022-10-03 15:26:21 +00:00
dbbdf65020
Added new gpu targets ( #631 )
...
[ROCm/rccl commit: ef71550738
2022-09-29 14:53:55 -06:00
07a0adf1d6
Another threadfence and flags rework ( #629 )
...
[ROCm/rccl commit: a523b37ac7
2022-09-28 16:49:29 -07:00
f6da79844a
Add LL128 tuning ( #630 )
...
[ROCm/rccl commit: 021932b3c8
2022-09-27 09:39:09 -07:00
b4bac0d15a
2.15.1-1
...
Add support for H100 (sm90).
Make sure NCCL kernel honor user stream priorities.
[ROCm/rccl commit: da8152e57a
2022-09-27 02:31:13 -07:00
8761a6c2fc
Merge remote-tracking branch 'origin/master'
...
[ROCm/rccl commit: 99c28f2e75
2022-09-27 02:24:41 -07:00
37ccaf1f82
Use compatibility shim only with static cudart
...
Closes issue 658
[ROCm/rccl commit: 78313a6d21
2022-09-27 02:22:48 -07:00
27d27e971b
Merge pull request #619 from whchung/exp_reduce_code_size
...
Only use split tree algorithm to reduce kernel code size.
[ROCm/rccl commit: e8af0716c4
2022-09-26 10:06:27 -05:00
fda4362c9e
Fix potential deadlock during init in multi-thread mode.
...
Make sure all calls calling cudaMalloc (including devCommSetup) are
called before the last bootstrapBarrier. That way, we avoid calls to
cudaMalloc be blocked by a NCCL kernel launched on another GPU by
another thread which completed init faster.
Resolve #623 .
[ROCm/rccl commit: ecab28a7c9
2022-09-26 02:13:10 -07:00
bd19566413
Merge pull request #621 from whchung/exp_reduce_sleep_cycles
...
Reduce s_sleep cycles
[ROCm/rccl commit: 35f1fe3434
2022-09-23 15:31:16 -05:00
183f1e6b32
Merge pull request #624 from whchung/exp_tweak_unroll_factors
...
Tweak unroll factors.
[ROCm/rccl commit: a80cc7e6e1
2022-09-23 11:30:05 -05:00
7cde92deff
Tweak unroll factors.
...
[ROCm/rccl commit: 84054c3b30
2022-09-22 13:03:04 -05:00
abdc365a05
Only use split tree algorithm to reduce kernel code size.
...
[ROCm/rccl commit: 02929cffb6
2022-09-22 12:01:53 -05:00
0b56e397cc
Reduce s_sleep cycles
...
[ROCm/rccl commit: a3c8ef8c03
2022-09-22 12:01:12 -05:00
975642e7ee
Abolish syncthreads and only wait on LDS traffic.
...
[ROCm/rccl commit: b9ae02d4ad
2022-09-22 12:00:37 -05:00
e4d46a0f64
Rework threadfence and flag setting ( #627 )
...
[ROCm/rccl commit: 49c811ecf9
2022-09-22 08:35:42 -07:00
81c71aeb67
Revert changes to gfx1030 ( #622 )
...
[ROCm/rccl commit: d9216af48b
2022-09-20 20:06:17 -07:00
98609a7b92
Define ncclShmem as global shared ( #618 )
...
* Use global defined shared memory
* Add --hipcc-func-supp to compile option
* Force inline some device functions
* Add back threadfence
[ROCm/rccl commit: 9e6c87a2bf
2022-09-20 09:00:20 -07:00
554b03bed1
address review comments
...
[ROCm/rccl commit: f89fd4777d
2022-09-20 11:58:33 +02:00
e742734b82
Fix intermittent 11.6 builds: generate unique .cu file for each object file
...
[ROCm/rccl commit: 79fb0326ac
2022-09-20 11:58:33 +02:00
95d6ed2154
make binary tree work on 2.13.4
...
[ROCm/rccl commit: 8f3219dbd4
2022-09-15 00:01:54 +00:00
ba6e2db70d
Misc fixes and disable binTree
...
[ROCm/rccl commit: a06e14e39b
2022-09-14 00:26:19 +00:00
4a86adcaba
Update init.cc
...
[ROCm/rccl commit: e5d2dfed34
2022-09-13 17:29:32 -05:00
4c17f4dcc1
Merge branch 'develop' into 2.13.4
...
[ROCm/rccl commit: be935d7ce7
2022-09-13 17:19:04 -05:00
7148c0aa7b
add binary tree
...
In addition, introduce the ability to have 2 trees at the same time.
Only for allreduce at the moment.
[ROCm/rccl commit: 65e2ae20e5
2022-09-13 20:52:32 +00:00
1d24c476f4
Merge branch 'develop' into 2.13.4
...
[ROCm/rccl commit: 009e79623f
2022-09-09 23:07:04 +00:00
Wenkai Du
Ziyue Yang
akolliasAMD
Sylvain Jeaugey
Wenkai Du
gilbertlee-amd
John Bachan
Edgar Gabriel
Cliff Woolley
Wen-Heng (Jack) Chung
Wen-Heng (Jack) Chung
Jane Xu
Edgar Gabriel
Gilbert Lee