AI/rocm-systems
2
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

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

Browse Source 
* Internal RCCL/NCCL functionality exposed when RCCL_EXPOSE_STATIC is enabled
* Algo/protocol/max channels can be obtained with the new RCCL API
* Introduce rccl_static and rccl_static_inline macros to work around invisible functions in core source files like enqueue.cc
* Add usage example in topo-explorer tool
This commit is contained in:
Mustafa Abduljabbar
2025-04-23 15:44:56 -04:00
committed by GitHub
parent 45e1c3f3e2
commit 82afb2bcfe
8 changed files with 110 additions and 12 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/device/common.h
-1
View File
@@ -14,7 +14,6 @@
#include "reduce_kernel.h"
#include "device_table.h"
#include "network/unpack/unpack_defs.h"
#define NCCL_MAX_DEV_ARITY (NCCL_MAX_TREE_ARITY-1) // Using balanced tree instead of split tree
#define __syncwarp()