Add support for inter-node communication using sockets and InfiniBand/RoCE. Improve latency. Add support for aggregation. Improve LL/regular tuning. Remove tests as those are now at github.com/nvidia/nccl-tests .
44 lines
1.2 KiB
Makefile
44 lines
1.2 KiB
Makefile
#
|
|
# Copyright (c) 2015-2018, NVIDIA CORPORATION. All rights reserved.
|
|
#
|
|
# See LICENSE.txt for license information
|
|
#
|
|
|
|
include ../../makefiles/common.mk
|
|
include ../../makefiles/version.mk
|
|
BUILDDIR ?= $(abspath ../../build)
|
|
TXZPREPDIR := $(BUILDDIR)/txz
|
|
PKGDIR := $(BUILDDIR)/pkg/txz/
|
|
|
|
TXZGEN_IN := $(wildcard *.in)
|
|
TXZGEN := $(TXZGEN_IN:.in=)
|
|
TXZTARGETS := $(patsubst %, $(TXZPREPDIR)/%, $(TXZGEN))
|
|
|
|
PKG_ARCH := $(shell uname -m)
|
|
|
|
prep: $(TXZTARGETS)
|
|
$(MAKE) -C ../.. lic BUILDDIR=$(BUILDDIR)
|
|
|
|
build: prep
|
|
$(MAKE) -C ../.. src.build BUILDDIR=$(BUILDDIR)
|
|
@printf "Building tar.xz package\n"
|
|
(cd $(BUILDDIR); bash txz/create_txz.sh)
|
|
mkdir -p $(PKGDIR)
|
|
mv $(BUILDDIR)/../nccl*.txz $(PKGDIR)
|
|
|
|
clean:
|
|
rm -Rf $(TXZPREPDIR) $(PKGDIR)
|
|
|
|
$(TXZPREPDIR)/% : %.in
|
|
@printf "Generating %-35s > %s\n" $< $@
|
|
mkdir -p $(TXZPREPDIR)
|
|
sed -e "s/\$${nccl:Major}/$(NCCL_MAJOR)/g" \
|
|
-e "s/\$${nccl:Minor}/$(NCCL_MINOR)/g" \
|
|
-e "s/\$${nccl:Patch}/$(NCCL_PATCH)/g" \
|
|
-e "s/\$${nccl:Suffix}/$(NCCL_SUFFIX)/g" \
|
|
-e "s/\$${cuda:Major}/$(CUDA_MAJOR)/g" \
|
|
-e "s/\$${cuda:Minor}/$(CUDA_MINOR)/g" \
|
|
-e "s/\$${pkg:Revision}/$(PKG_REVISION)/g" \
|
|
-e "s/\$${pkg:Arch}/$(PKG_ARCH)/g" \
|
|
$< > $@
|