Pedram Alizadeh 3f7c08648f 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

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

..

network/unpack

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

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

all_gather.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

all_reduce.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

alltoall_pivot.h

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

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

broadcast.h

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

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

common_kernel.h

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

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

common.cu

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

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

common.h

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

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

generate.py

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

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

msccl_kernel_impl.h

removed gfx940 and gfx941 (#1606)

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

onerank.cu

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

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

op128.h

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

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

primitives.h

moving the thread_fence to apply before atomic fetch (#1672)

2025-05-14 10:10:05 -05:00

prims_ll128.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

prims_ll.h

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

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

prims_simple.h

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

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

reduce_kernel.h

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

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

reduce_scatter.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

reduce.h

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

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

sendrecv.h

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

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