alex-breslow-amd 455d516dc4 [gfx950] Make bypassing __threadfence the default for multinode. (#1947) ...

* Gate based on ROCM version, safe for ROCm 7.0.2 and beyond.
* Updates naming to gfx9CheapFenceOff since we use this for gfx942 and gfx950.  Thanks Nilesh.
* Add info logging statement to NCCL_INIT to print whether enabled when INFO logging is enabled.

[ROCm/rccl commit: c70f5b4621]

2025-10-15 09:15:36 -07:00

..

device

[gfx950] Make bypassing __threadfence the default for multinode. (#1947)

2025-10-15 09:15:36 -07:00

graph

gfx950 channel tuning for ReduceScatter and AllGather (#1940)

2025-10-14 09:50:44 -05:00

include

[gfx950] Make bypassing __threadfence the default for multinode. (#1947)

2025-10-15 09:15:36 -07:00

misc

Reverse logic of context tracking enablement from #1927 (#1971)

2025-10-09 10:24:09 +02:00

plugin

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

2025-10-06 18:36:49 -04:00

ras

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

2025-08-28 15:46:28 -05:00

register

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

2025-08-28 15:46:28 -05:00

transport

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

2025-10-06 18:36:49 -04:00

allocator.cc

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

2025-08-28 15:46:28 -05:00

bootstrap.cc

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

2025-08-28 15:46:28 -05:00

channel.cc

NCCL 2.27.3-1

2025-05-29 20:56:40 -07:00

collectives.cc

Update direct AG and single node LL threshold (#1944)

2025-10-09 10:48:50 -05:00

debug.cc

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

2025-08-28 15:46:28 -05:00

enhcompat.cc

2.12.7-1

2022-03-02 20:48:56 +01:00

enqueue.cc

[gfx950] Make bypassing __threadfence the default for multinode. (#1947)

2025-10-15 09:15:36 -07:00

group.cc

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

2025-08-28 15:46:28 -05:00

init_nvtx.cc

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

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

init.cc

[gfx950] Make bypassing __threadfence the default for multinode. (#1947)

2025-10-15 09:15:36 -07:00

Makefile

NCCL 2.27.3-1

2025-05-29 20:56:40 -07:00

mnnvl.cc

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

2025-08-28 15:46:28 -05:00

msccl.cc

Deprecated MSCCL API functions (#1740)

2025-06-11 17:52:09 -06:00

nccl.h.in

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

2025-08-28 15:46:28 -05:00

nccl.pc.in

Add pkgconfig file (#190)

2019-04-08 09:16:54 -07:00

proxy.cc

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

2025-08-28 15:46:28 -05:00

rccl_wrap.cc

gfx950 channel tuning for ReduceScatter and AllGather (#1940)

2025-10-14 09:50:44 -05:00

symmetric.cc

Added an implementation of ncclSymGetKernelPtr for when GENERATE_SYM_KERNELS is not defined, as it is normally generated code. (#1925)

2025-09-19 07:52:33 -06:00

transport.cc

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

2025-08-28 15:46:28 -05:00