NCCL 2.4.6-1
Added detection of IBM/Power NVLink bridge device.
Add NUMA support to PCI distance calculations.
Added NCCL_IGNORE_CPU_AFFINITY env var.
Fix memory leaks; GithubIssue#180
Compiler warning fix; GithubIssue#178
Replace non-standard variable length arrays. GithubIssue#171
Fix Tree+Shared Memory crash. GithubPR#185
Fix LL cleanup hang during long running DL jobs.
Fix NCCL_RINGS environment variable handling.
Added extra checks to catch repeat calls to ncclCommDestroy() GithubIssue#191
Improve bootstrap socket connection reliability at scale.
Fix hostname hashing issue. GithubIssue#187
Code cleanup to rename all non device files from *.cu to *.cc
[ROCm/rccl commit: f40ce73e89]
This commit is contained in:
David Addison
@@ -1,5 +1,5 @@
|
||||
#
|
||||
# Copyright (c) 2015-2018, NVIDIA CORPORATION. All rights reserved.
|
||||
# Copyright (c) 2015-2019, NVIDIA CORPORATION. All rights reserved.
|
||||
#
|
||||
# See LICENSE.txt for license information
|
||||
#
|
||||
@@ -9,10 +9,10 @@ include ../makefiles/version.mk
|
||||
|
||||
##### src files
|
||||
INCEXPORTS := nccl.h nccl_net.h
|
||||
LIBSRCFILES := init.cu channel.cu bootstrap.cu transport.cu enqueue.cu \
|
||||
misc/group.cu misc/nvmlwrap.cu misc/ibvwrap.cu misc/rings.cu misc/utils.cu misc/checks.cu misc/trees.cu \
|
||||
transport/p2p.cu transport/shm.cu transport/net.cu transport/net_socket.cu transport/net_ib.cu \
|
||||
collectives/all_reduce.cu collectives/all_gather.cu collectives/broadcast.cu collectives/reduce.cu collectives/reduce_scatter.cu
|
||||
LIBSRCFILES := init.cc channel.cc bootstrap.cc transport.cc enqueue.cc \
|
||||
misc/group.cc misc/nvmlwrap.cc misc/ibvwrap.cc misc/rings.cc misc/utils.cc misc/argcheck.cc misc/trees.cc misc/topo.cc \
|
||||
transport/p2p.cc transport/shm.cc transport/net.cc transport/net_socket.cc transport/net_ib.cc \
|
||||
collectives/all_reduce.cc collectives/all_gather.cc collectives/broadcast.cc collectives/reduce.cc collectives/reduce_scatter.cc
|
||||
|
||||
##### lib files
|
||||
LIBNAME := libnccl.so
|
||||
@@ -27,7 +27,7 @@ INCTARGETS := $(INCEXPORTS:%=$(INCDIR)/%)
|
||||
LIBSONAME := $(LIBNAME:%=%.$(NCCL_MAJOR))
|
||||
LIBTARGET := $(LIBNAME:%=%.$(NCCL_MAJOR).$(NCCL_MINOR).$(NCCL_PATCH))
|
||||
STATICLIBTARGET := $(STATICLIBNAME)
|
||||
LIBOBJ := $(LIBSRCFILES:%.cu=$(OBJDIR)/%.o)
|
||||
LIBOBJ := $(LIBSRCFILES:%.cc=$(OBJDIR)/%.o)
|
||||
DEPFILES := $(LIBOBJ:%.o=%.d)
|
||||
LDFLAGS += -L${CUDA_LIB} -lcudart_static -lpthread -lrt -ldl
|
||||
|
||||
@@ -87,11 +87,11 @@ $(INCDIR)/nccl_%.h : include/nccl_%.h
|
||||
mkdir -p $(INCDIR)
|
||||
cp -f $< $@
|
||||
|
||||
$(OBJDIR)/%.o : %.cu
|
||||
$(OBJDIR)/%.o : %.cc
|
||||
@printf "Compiling %-35s > %s\n" $< $@
|
||||
mkdir -p `dirname $@`
|
||||
$(NVCC) -I. -I$(INCDIR) -Iinclude -c $(NVCUFLAGS) --compiler-options "$(CXXFLAGS)" $< -o $@
|
||||
@$(NVCC) -I. -I$(INCDIR) -Iinclude -M $(NVCUFLAGS) --compiler-options "$(CXXFLAGS)" $< > $(@:%.o=%.d.tmp)
|
||||
$(CXX) -I. -I$(INCDIR) $(CXXFLAGS) -Iinclude -c $< -o $@
|
||||
@$(CXX) -I. -I$(INCDIR) $(CXXFLAGS) -Iinclude -M $< > $(@:%.o=%.d.tmp)
|
||||
@sed "0,/^.*:/s//$(subst /,\/,$@):/" $(@:%.o=%.d.tmp) > $(@:%.o=%.d)
|
||||
@sed -e 's/.*://' -e 's/\\$$//' < $(@:%.o=%.d.tmp) | fmt -1 | \
|
||||
sed -e 's/^ *//' -e 's/$$/:/' >> $(@:%.o=%.d)
|
||||
@@ -107,7 +107,7 @@ install : lib
|
||||
cp -P -v $(BUILDDIR)/lib/* $(PREFIX)/lib/
|
||||
cp -v $(BUILDDIR)/include/* $(PREFIX)/include/
|
||||
|
||||
FILESTOFORMAT := $(shell find . -name ".\#*" -prune -o \( -name "*.cu" -o -name "*.h" \) -print | grep -v -E 'ibvwrap.h|nvmlwrap.h|nccl.h')
|
||||
FILESTOFORMAT := $(shell find . -name ".\#*" -prune -o \( -name "*.cc" -o -name "*.h" \) -print | grep -v -E 'ibvwrap.h|nvmlwrap.h|nccl.h')
|
||||
# Note that formatting.mk defines a new target so in order to not overwrite the default target,
|
||||
# it shouldn't be included at the top. Also, it uses the above definition of FILESTOFORMAT as well
|
||||
# as the BUILDDIR variable.
|
||||
|
||||
مرجع در شماره جدید
Block a user