add script for oss build and add dev signature as suffix (NVIDIA#38)

Summary:

See title.

Differential Revision: D51420806

maint/oss_build.sh

@@ -0,0 +1,53 @@
+#!/bin/bash
+set -x
+
+hash=""
+# guess default DEV_SIGNATURE
+is_git=$(git rev-parse --is-inside-work-tree)
+if [ $is_git ]; then
+    hash="git-"$(git rev-parse --short HEAD)
+fi
+
+BUILDDIR=${BUILDDIR:=/tmp/nccl-exp/2.18.3/build}
+NVCC_ARCH=${NVCC_ARCH:="a100,h100"}
+CUDA_HOME=${CUDA_HOME:="/usr/local/cuda"}
+# TODO: automatically get hg id or git hash
+DEV_SIGNATURE=${DEV_SIGNATURE:="$hash"}
+
+echo "BUILDDIR=${BUILDDIR}"
+echo "NVCC_ARCH=${NVCC_ARCH}"
+echo "CUDA_HOME=${CUDA_HOME}"
+echo "DEV_SIGNATURE=${DEV_SIGNATURE}"
+
+IFS=',' read -ra arch_array <<< "$NVCC_ARCH"
+arch_gencode=""
+for arch in "${arch_array[@]}"
+do
+    case "$arch" in
+       "p100")
+       arch_gencode="$arch_gencode -gencode=arch=compute_60,code=sm_60"
+        ;;
+       "v100")
+       arch_gencode="$arch_gencode -gencode=arch=compute_70,code=sm_70"
+        ;;
+       "a100")
+       arch_gencode="$arch_gencode -gencode=arch=compute_80,code=sm_80"
+       ;;
+       "h100")
+        arch_gencode="$arch_gencode -gencode=arch=compute_90,code=sm_90"
+       ;;
+    esac
+done
+
+echo "NVCC_GENCODE=$arch_gencode"
+
+# build libnccl
+export BUILDDIR=$BUILDDIR
+make src.clean && make -j src.build NVCC_GENCODE="$arch_gencode" CUDA_HOME="$CUDA_HOME" DEV_SIGNATURE="$DEV_SIGNATURE"
+
+# sanity check
+pushd examples
+export NCCL_HOME=$BUILDDIR
+make clean && make all
+NCCL_DEBUG=WARN LD_LIBRARY_PATH=$BUILDDIR/lib $BUILDDIR/examples/HelloWorld
+popd

src/Makefile

@@ -83,6 +83,7 @@ $(INCDIR)/nccl.h : nccl.h.in ../makefiles/version.mk
 	    -e "s/\$${nccl:Patch}/$(NCCL_PATCH)/g" \
 	    -e "s/\$${nccl:Suffix}/$(NCCL_SUFFIX)/g" \
 	    -e "s/\$${nccl:Version}/$(NCCL_VERSION)/g" \
+	    -e "s/\$${nccl:DevSignature}/$(DEV_SIGNATURE)/g" \
 	    $< > $@
 
 $(LIBDIR)/$(LIBTARGET): $(LIBOBJ) $(DEVICELIB)

src/nccl.h.in

@@ -16,7 +16,7 @@
 #define NCCL_MAJOR ${nccl:Major}
 #define NCCL_MINOR ${nccl:Minor}
 #define NCCL_PATCH ${nccl:Patch}
-#define NCCL_SUFFIX "${nccl:Suffix}"
+#define NCCL_SUFFIX "${nccl:Suffix} ${nccl:DevSignature}"
 #define IS_NCCL_EXP
 
 #define NCCL_VERSION_CODE ${nccl:Version}