Pedram Alizadeh 1ace5d05ed Reapplying PR #1641 [AG and RS channel tuning] Add thread work threshold to tuning models and precompute reg index in LL128 (#1713) ...

* Reapply "[AG and RS channel tuning] Add thread work threshold to tuning models and precompute reg index in LL128 (#1641)"

This reverts commit 943ad6f7820739385a0b54e81f823d0df1dbf71c.

* Decreasing NCCL_LL128_SHMEM_ELEMS_PER_THREAD from 16 to 8

[ROCm/rccl commit: 3f7c08648f]

2025-06-04 13:22:11 -04:00

..

msccl

SPLITCOMM design fix in src/misc/msccl (#1715)

2025-06-01 21:00:38 -05:00

mscclpp

fix scatter_perf crash (#1493)

2025-01-21 09:24:32 -06:00

npkit

NPKit: enable reduce scatter profiling (#1580)

2025-03-04 10:03:56 -08:00

nvtx3

Merge remote-tracking branch 'nccl/master' into develop

2025-04-30 13:31:41 -05:00

alloc.h

Merge remote-tracking branch 'nccl/master' into develop

2025-04-30 13:31:41 -05:00

alt_rsmi.h

add alternative to rocm_smi_lib

2024-05-14 13:51:41 -07:00

api_trace.h

SPLITCOMM design fix in src/misc/msccl (#1715)

2025-06-01 21:00:38 -05:00

archinfo.h

Fix gcnArchName bug in topology dump (#937)

2023-10-28 12:30:36 -05:00

argcheck.h

2.21.5-1

2024-04-02 01:53:21 -07:00

BfdBacktrace.hpp

Merge remote-tracking branch 'nccl/master' into develop

2022-09-09 16:05:38 +00:00

bitops.h

Merge remote-tracking branch 'nccl/master' into develop

2025-03-27 12:53:04 -05:00

bootstrap.h

2.23.4-1

2024-09-16 23:41:17 -07:00

channel.h

2.22.3-1

2024-06-19 01:57:16 -07:00

checks.h

2.23.4-1

2024-09-16 23:41:17 -07:00

coll_net.h

2.20.3-1

2024-02-13 04:22:38 -08:00

collectives.h

Use One Slice per Basic Primitive for AllReduce, ReduceScatter, AllGather (#1681) for Single Node on Some GFX9 Systems

2025-05-29 16:17:35 -07:00

comm.h

Use One Slice per Basic Primitive for AllReduce, ReduceScatter, AllGather (#1681) for Single Node on Some GFX9 Systems

2025-05-29 16:17:35 -07:00

core.h

Merge remote-tracking branch 'rccl/develop' into 2.19.4

2024-02-09 11:31:03 -06:00

cpuset.h

2.8.3-1

2020-11-17 11:08:52 -08:00

cudawrap.h

2.23.4-1

2024-09-16 23:41:17 -07:00

debug.h

2.24.3-1

2025-01-07 02:01:15 -08:00

device.h

Reapplying PR #1641 [AG and RS channel tuning] Add thread work threshold to tuning models and precompute reg index in LL128 (#1713)

2025-06-04 13:22:11 -04:00

enqueue.h

Merge pull request #1662 from BertanDogancay/2.25

2025-05-06 09:39:09 -04:00

gdrwrap.h

removed gfx940 and gfx941 (#1606)

2025-03-20 09:34:53 -06:00

git_version.h

Merge remote-tracking branch 'nccl/master' into develop

2022-09-09 16:05:38 +00:00

graph.h

Merge remote-tracking branch 'nccl/master' into develop

2025-04-30 13:31:41 -05:00

group.h

Merge remote-tracking branch 'nccl/master' into develop

2025-01-23 12:08:46 -06:00

hip_rocm_version_info.h

Improved version reporting in NCCL_DEBUG=VERSION (#1232)

2024-07-12 08:14:29 -05:00

ibvcore.h

2.19.1-1

2023-09-26 05:50:33 -07:00

ibvsymbols.h

2.19.1-1

2023-09-26 05:50:33 -07:00

ibvwrap.h

2.24.3-1

2025-01-07 02:01:15 -08:00

info.h

Merge remote-tracking branch 'nccl/master' into develop

2025-01-23 12:08:46 -06:00

ipcsocket.h

2.20.3-1

2024-02-13 04:22:38 -08:00

mnnvl.h

NCCL 2.25.1-1

2025-01-27 03:33:57 -08:00

nccl_common.h

Re-apply unroll=1 and 112 channels for gfx950 (#1706)

2025-05-28 14:58:10 -05:00

nccl_net.h

2.24.3-1

2025-01-07 02:01:15 -08:00

nccl_profiler.h

2.24.3-1

2025-01-07 02:01:15 -08:00

nccl_tuner.h

2.24.3-1

2025-01-07 02:01:15 -08:00

net_device.h

2.24.3-1

2025-01-07 02:01:15 -08:00

net.h

2.22.3-1

2024-06-19 01:57:16 -07:00

nvmlwrap.h

2.24.3-1

2025-01-07 02:01:15 -08:00

nvtx_payload_schemas.h

Merge remote-tracking branch 'nccl/master' into develop

2025-04-30 13:31:41 -05:00

nvtx_stub.h

Merge remote-tracking branch 'nccl/master' into develop

2025-04-30 13:31:41 -05:00

nvtx.h

Merge remote-tracking branch 'nccl/master' into develop

2025-04-30 13:31:41 -05:00

p2p.h

2.23.4-1

2024-09-16 23:41:17 -07:00

param.h

Merge remote-tracking branch 'nccl/v2.19' into develop

2024-01-24 15:25:33 -08:00

profiler.h

2.24.3-1

2025-01-07 02:01:15 -08:00

proxy.h

Merge remote-tracking branch 'nccl/master' into develop

2025-04-30 13:31:41 -05:00

ras.h

2.24.3-1

2025-01-07 02:01:15 -08:00

rccl_common.h

Reapplying PR #1641 [AG and RS channel tuning] Add thread work threshold to tuning models and precompute reg index in LL128 (#1713)

2025-06-04 13:22:11 -04:00

rccl_float8.h

Switched to using the hip_fp8 header instead of rccl_float8, resolving compatibility issues. (#1546)

2025-05-14 15:33:03 -05:00

rccl_vars.h

Expose production tuning table in topo_explorer using internal RCCL/NCCL logic (#1628)

2025-04-23 15:44:56 -04:00

recorder.h

RCCL Replayer update (#1603)

2025-04-19 00:21:27 -04:00

register.h

2.24.3-1

2025-01-07 02:01:15 -08:00

rocm_smi_wrap.h

Support P2P with invisible devices (#636)

2022-10-17 10:24:59 -07:00

rocmwrap.h

Merge remote-tracking branch 'nccl/master' into develop

2025-04-30 13:31:41 -05:00

roctx.h

Add ncclDataType_t as type to ROCTX (#1512)

2025-01-30 13:46:48 -05:00

shm.h

2.23.4-1

2024-09-16 23:41:17 -07:00

shmutils.h

2.24.3-1

2025-01-07 02:01:15 -08:00

signals.h

Merge remote-tracking branch 'nccl/master' into develop

2022-09-09 16:05:38 +00:00

socket.h

2.24.3-1

2025-01-07 02:01:15 -08:00

strongstream.h

Use new HIP graph API compatible with CUDA 11030 (#991)

2024-01-21 19:00:50 -08:00

timer.h

2.23.4-1

2024-09-16 23:41:17 -07:00

transport.h

Merge remote-tracking branch 'nccl/master' into develop

2025-04-23 20:47:43 -07:00

trees.h

2.8.3-1

2020-11-17 11:08:52 -08:00

tuner.h

2.22.3-1

2024-06-19 01:57:16 -07:00

utils.h

RCCL Multinode DMA Buffer crash fix (#1682)

2025-05-08 19:17:39 -05:00