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

NCCL 2.25.1-1

Browse Source 
Add Blackwell/SM100 support
 * Add compilation for sm100
 * Add graph search speeds for Blackwell
 * Optimize graph search to converge on large NVLink domains
 * Limit NVLS heads to 32
 * Increase various limits to fit large NVLink domains
 * Add extra checks for IMEX setup, needed for MNNVL
 * Increase MAXCHANNELS to 64

Extend NVTX instrumentation to track NCCL communicators
 * Add communicator ID to NVTX traces to allow for correlation
   between ranks.

RAS fixes
This commit is contained in:
Sylvain Jeaugey
2025-01-27 03:30:22 -08:00
parent 1672c85781
commit 80f6bda437
30 changed files with 603 additions and 330 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/include/enqueue.h
+1 -1
View File
@@ -17,7 +17,7 @@
#define NCCL_SIMPLE_ALIGNMENT (WARP_SIZE * 8LL * 16LL)
#define NCCL_BYTES_ALIGNMENT 16
ncclResult_t ncclInitKernelsForDevice(int cudaArch, size_t* maxStackSize);
ncclResult_t ncclInitKernelsForDevice(int cudaArch, int maxSharedMem, size_t* maxStackSize);
ncclResult_t ncclEnqueueCheck(struct ncclInfo* info);
ncclResult_t ncclLaunchPrepare(struct ncclComm* comm);
ncclResult_t ncclLaunchKernelBefore_NoUncapturedCuda(struct ncclComm* comm, struct ncclKernelPlan* plan);