2.12.7-1

Add network communication through another GPU connected with NVLink
(PXN).
Add aggregation of messages coming from different local GPUs through
PXN and going to the same destination.
Add new v5 plugin API with grouped receives and tags.
Add compat for v4 plugins.
Add naming of NCCL threads to help debugging.
Fix NVLink detection and avoid data corruption when some NVLinks are
down.
Add support for Relaxed Ordering for IB.
Add profiling and timing infrastructure.

makefiles/common.mk

@@ -1,5 +1,5 @@
 #
-# Copyright (c) 2015-2020, NVIDIA CORPORATION. All rights reserved.
+# Copyright (c) 2015-2022, NVIDIA CORPORATION. All rights reserved.
 #
 # See LICENSE.txt for license information
 #
@@ -23,7 +23,6 @@ CUDA_MAJOR = $(shell echo $(CUDA_VERSION) | cut -d "." -f 1)
 CUDA_MINOR = $(shell echo $(CUDA_VERSION) | cut -d "." -f 2)
 #$(info CUDA_VERSION ${CUDA_MAJOR}.${CUDA_MINOR})
 
-
 # You should define NVCC_GENCODE in your environment to the minimal set
 # of archs to reduce compile time.
 CUDA8_GENCODE = -gencode=arch=compute_35,code=sm_35 \
@@ -39,7 +38,7 @@ CUDA11_PTX    = -gencode=arch=compute_80,code=compute_80
 
 # Include Ampere support if we're using CUDA11 or above
 ifeq ($(shell test "0$(CUDA_MAJOR)" -ge 11; echo $$?),0)
-  NVCC_GENCODE ?= $(CUDA8_GENCODE) $(CUDA9_GENCODE) $(CUDA9_PTX) $(CUDA11_GENCODE) $(CUDA11_PTX)
+  NVCC_GENCODE ?= $(CUDA8_GENCODE) $(CUDA9_GENCODE) $(CUDA11_GENCODE) $(CUDA11_PTX)
 # Include Volta support if we're using CUDA9 or above
 else ifeq ($(shell test "0$(CUDA_MAJOR)" -ge 9; echo $$?),0)
   NVCC_GENCODE ?= $(CUDA8_GENCODE) $(CUDA9_GENCODE) $(CUDA9_PTX)

makefiles/version.mk

@@ -1,6 +1,6 @@
 ##### version
 NCCL_MAJOR   := 2
-NCCL_MINOR   := 11
-NCCL_PATCH   := 4
+NCCL_MINOR   := 12
+NCCL_PATCH   := 7
 NCCL_SUFFIX  :=
 PKG_REVISION := 1