2.5.6-1 (#255)

Add LL128 Protocol.

Rewrite the topology detection and tree/ring creation (#179). Improve
tree performance by sending/receiving from different GPUs. Add
model-based tuning to switch between the different algorithms and
protocols.

Rework P2P/SHM detection in containers (#155, #248).

Detect duplicated devices and return an error (#231).

Add tuning for GCP

[ROCm/rccl commit: 299c554dcc]

projects/rccl/makefiles/common.mk

@@ -25,8 +25,7 @@ CUDA_MINOR = $(shell echo $(CUDA_VERSION) | cut -d "." -f 2)
 
 # Better define NVCC_GENCODE in your environment to the minimal set
 # of archs to reduce compile time.
-CUDA8_GENCODE = -gencode=arch=compute_30,code=sm_30 \
-                -gencode=arch=compute_35,code=sm_35 \
+CUDA8_GENCODE = -gencode=arch=compute_35,code=sm_35 \
                 -gencode=arch=compute_50,code=sm_50 \
                 -gencode=arch=compute_60,code=sm_60 \
                 -gencode=arch=compute_61,code=sm_61
@@ -46,7 +45,10 @@ endif
 CXXFLAGS   := -DCUDA_MAJOR=$(CUDA_MAJOR) -DCUDA_MINOR=$(CUDA_MINOR) -fPIC -fvisibility=hidden
 CXXFLAGS   += -Wall -Wno-unused-function -Wno-sign-compare -std=c++11 -Wvla
 CXXFLAGS   += -I $(CUDA_INC)
-NVCUFLAGS  := -ccbin $(CXX) $(NVCC_GENCODE) -lineinfo -std=c++11 -Xptxas -maxrregcount=96 -Xfatbin -compress-all
+# Maxrregcount needs to be set accordingly to NCCL_MAX_NTHREADS (otherwise it will cause kernel launch errors)
+# 512 : 120, 640 : 96, 768 : 80, 1024 : 60
+# We would not have to set this if we used __launch_bounds__, but this only works on kernels, not on functions.
+NVCUFLAGS  := -ccbin $(CXX) $(NVCC_GENCODE) -std=c++11 -Xptxas -maxrregcount=96 -Xfatbin -compress-all
 # Use addprefix so that we can specify more than one path
 NVLDFLAGS  := -L${CUDA_LIB} -lcudart -lrt
 

projects/rccl/makefiles/version.mk

@@ -1,6 +1,6 @@
 ##### version
 NCCL_MAJOR   := 2
-NCCL_MINOR   := 4
-NCCL_PATCH   := 8
+NCCL_MINOR   := 5
+NCCL_PATCH   := 6
 NCCL_SUFFIX  :=
 PKG_REVISION := 1